style: fix import sorting (ruff I001)

Addresses reviewer suggestion to keep imports conventional.

.gitignore

@@ -1,26 +1,87 @@
+# Project-specific
 .worktrees/
 .assets
 .docs
 .env
 .web
+.orion
+
+# Python bytecode & caches
 *.pyc
-dist/
-build/
-*.egg-info/
-*.egg
-*.pycs
 *.pyo
 *.pyd
 *.pyw
 *.pyz
-*.pywz
-*.pyzz
+__pycache__/
+*.egg-info/
+*.egg
 .venv/
 venv/
-__pycache__/
-poetry.lock
 .pytest_cache/
-botpy.log
-nano.*.save
-.DS_Store
+.mypy_cache/
+.ruff_cache/
+.pytype/
+.dmypy.json
+dmypy.json
+.tox/
+.nox/
+.hypothesis/
+
+# Build & packaging
+dist/
+build/
+*.manifest
+*.spec
+pip-wheel-metadata/
+share/python-wheels/
+
+# Test & coverage
+.coverage
+.coverage.*
+htmlcov/
+coverage.xml
+*.cover
+
+# Lock files (project policy)
+poetry.lock
 uv.lock
+
+# Jupyter
+.ipynb_checkpoints/
+
+# macOS
+.DS_Store
+.AppleDouble
+.LSOverride
+
+# Windows
+Thumbs.db
+ehthumbs.db
+Desktop.ini
+
+# Linux
+.directory
+
+# Editors & IDEs (local workspace / user settings)
+.vscode/
+.cursor/
+.idea/
+.fleet/
+*.code-workspace
+*.sublime-project
+*.sublime-workspace
+*.swp
+*.swo
+*~
+nano.*.save
+
+# Environment & secrets (keep examples tracked if needed)
+.env.*
+!.env.example
+
+# Logs & temp
+*.log
+logs/
+tmp/
+temp/
+*.tmp

README.md

@@ -21,7 +21,20 @@
 
 ## 📢 News
 
+- **2026-04-14** 🚀 Released **v0.1.5.post1** — Dream skill discovery, mid-turn follow-up injection, WebSocket channel, and deeper channel integrations. Please see [release notes](https://github.com/HKUDS/nanobot/releases/tag/v0.1.5.post1) for details.
+- **2026-04-13** 🛡️ Agent turn hardened — user messages persisted early, auto-compact skips active tasks.
+- **2026-04-12** 🔒 Lark global domain support, Dream learns discovered skills, shell sandbox tightened.
+- **2026-04-11** ⚡ Context compact shrinks sessions on the fly; Kagi web search; QQ & WeCom full media.
+- **2026-04-10** 📓 Notebook editing tool, multiple MCP servers, Feishu streaming & done-emoji.
+- **2026-04-09** 🔌 WebSocket channel, unified cross-channel session, `disabled_skills` config.
+- **2026-04-08** 📤 API file uploads, OpenAI reasoning auto-routing with Responses fallback.
+- **2026-04-07** 🧠 Anthropic adaptive thinking, MCP resources & prompts exposed as tools.
+- **2026-04-06** 🛰️ Langfuse observability, unified Whisper transcription, email attachments.
 - **2026-04-05** 🚀 Released **v0.1.5** — sturdier long-running tasks, Dream two-stage memory, production-ready sandboxing and programming Agent SDK. Please see [release notes](https://github.com/HKUDS/nanobot/releases/tag/v0.1.5) for details.
+
+<details>
+<summary>Earlier news</summary>
+
 - **2026-04-04** 🚀 Jinja2 response templates, Dream memory hardened, smarter retry handling.
 - **2026-04-03** 🧠 Xiaomi MiMo provider, chain-of-thought reasoning visible, Telegram UX polish.
 - **2026-04-02** 🧱 Long-running tasks run more reliably — core runtime hardening.
@@ -31,11 +44,6 @@
 - **2026-03-29** 💬 WeChat voice, typing, QR/media resilience; fixed-session OpenAI-compatible API.
 - **2026-03-28** 📚 Provider docs refresh; skill template wording fix.
 - **2026-03-27** 🚀 Released **v0.1.4.post6** — architecture decoupling, litellm removal, end-to-end streaming, WeChat channel, and a security fix. Please see [release notes](https://github.com/HKUDS/nanobot/releases/tag/v0.1.4.post6) for details.
-
-
-<details>
-<summary>Earlier news</summary>
-
 - **2026-03-26** 🏗️ Agent runner extracted and lifecycle hooks unified; stream delta coalescing at boundaries.
 - **2026-03-25** 🌏 StepFun provider, configurable timezone, Gemini thought signatures.
 - **2026-03-24** 🔧 WeChat compatibility, Feishu CardKit streaming, test suite restructured.
@@ -112,26 +120,57 @@
 
 ## Table of Contents
 
-- [News](#-news)
-- [Key Features](#key-features-of-nanobot)
-- [Architecture](#️-architecture)
-- [Features](#-features)
-- [Install](#-install)
-- [Quick Start](#-quick-start)
-- [Chat Apps](#-chat-apps)
-- [Agent Social Network](#-agent-social-network)
-- [Configuration](#️-configuration)
-- [Multiple Instances](#-multiple-instances)
-- [Memory](#-memory)
-- [CLI Reference](#-cli-reference)
-- [In-Chat Commands](#-in-chat-commands)
-- [Python SDK](#-python-sdk)
-- [OpenAI-Compatible API](#-openai-compatible-api)
-- [Docker](#-docker)
-- [Linux Service](#-linux-service)
-- [Project Structure](#-project-structure)
-- [Contribute & Roadmap](#-contribute--roadmap)
-- [Star History](#-star-history)
+- [📢 News](#-news)
+- [Key Features of nanobot:](#key-features-of-nanobot)
+- [🏗️ Architecture](#️-architecture)
+- [Table of Contents](#table-of-contents)
+- [✨ Features](#-features)
+- [📦 Install](#-install)
+  - [Update to latest version](#update-to-latest-version)
+- [🚀 Quick Start](#-quick-start)
+- [💬 Chat Apps](#-chat-apps)
+- [🌐 Agent Social Network](#-agent-social-network)
+- [⚙️ Configuration](#️-configuration)
+  - [Environment Variables for Secrets](#environment-variables-for-secrets)
+  - [Providers](#providers)
+  - [Channel Settings](#channel-settings)
+    - [Retry Behavior](#retry-behavior)
+  - [Web Search](#web-search)
+    - [`tools.web.search`](#toolswebsearch)
+  - [MCP (Model Context Protocol)](#mcp-model-context-protocol)
+  - [Security](#security)
+  - [Auto Compact](#auto-compact)
+  - [Timezone](#timezone)
+  - [Unified Session](#unified-session)
+  - [Disabled Skills](#disabled-skills)
+- [🧩 Multiple Instances](#-multiple-instances)
+  - [Quick Start](#quick-start)
+  - [Path Resolution](#path-resolution)
+  - [How It Works](#how-it-works)
+  - [Minimal Setup](#minimal-setup)
+  - [Common Use Cases](#common-use-cases)
+  - [Notes](#notes)
+- [🧠 Memory](#-memory)
+- [💻 CLI Reference](#-cli-reference)
+- [💬 In-Chat Commands](#-in-chat-commands)
+- [🐍 Python SDK](#-python-sdk)
+- [🔌 OpenAI-Compatible API](#-openai-compatible-api)
+  - [Behavior](#behavior)
+  - [Endpoints](#endpoints)
+  - [curl](#curl)
+  - [File Upload (JSON base64)](#file-upload-json-base64)
+  - [File Upload (multipart/form-data)](#file-upload-multipartform-data)
+  - [Python (`requests`)](#python-requests)
+  - [Python (`openai`)](#python-openai)
+- [🐳 Docker](#-docker)
+  - [Docker Compose](#docker-compose)
+  - [Docker](#docker)
+- [🐧 Linux Service](#-linux-service)
+- [📁 Project Structure](#-project-structure)
+- [🤝 Contribute \& Roadmap](#-contribute--roadmap)
+  - [Branching Strategy](#branching-strategy)
+  - [Contributors](#contributors)
+- [⭐ Star History](#-star-history)
 
 ## ✨ Features
 
@@ -276,6 +315,7 @@ Connect nanobot to your favorite chat platform. Want to build your own? See the
 | **Email** | IMAP/SMTP credentials |
 | **QQ** | App ID + App Secret |
 | **Wecom** | Bot ID + Bot Secret |
+| **Microsoft Teams** | App ID + App Password + public HTTPS endpoint |
 | **Mochat** | Claw token (auto-setup available) |
 
 <details>
@@ -394,7 +434,9 @@ If you prefer to configure manually, add the following to `~/.nanobot/config.jso
       "enabled": true,
       "token": "YOUR_BOT_TOKEN",
       "allowFrom": ["YOUR_USER_ID"],
-      "groupPolicy": "mention"
+      "allowChannels": [],
+      "groupPolicy": "mention",
+      "streaming": true
     }
   }
 }
@@ -405,6 +447,8 @@ If you prefer to configure manually, add the following to `~/.nanobot/config.jso
 > - `"open"` — Respond to all messages
 > DMs always respond when the sender is in `allowFrom`.
 > - If you set group policy to open create new threads as private threads and then @ the bot into it. Otherwise the thread itself and the channel in which you spawned it will spawn a bot session.
+> `allowChannels` restricts the bot to specific Discord channel IDs. Empty (default) means respond in every channel the bot can see. Example: `["1234567890", "0987654321"]`. The filter applies after `allowFrom`, so both must pass.
+> `streaming` defaults to `true`. Disable it only if you explicitly want non-streaming replies.
 
 **5. Invite the bot**
 - OAuth2 → URL Generator
@@ -560,6 +604,7 @@ Uses **WebSocket** long connection — no public IP required.
       "groupPolicy": "mention",
       "reactEmoji": "OnIt",
       "doneEmoji": "DONE",
+      "toolHintPrefix": "🔧",
       "streaming": true,
       "domain": "feishu"
     }
@@ -573,6 +618,7 @@ Uses **WebSocket** long connection — no public IP required.
 > `groupPolicy`: `"mention"` (default — respond only when @mentioned), `"open"` (respond to all group messages). Private chats always respond.
 > `reactEmoji`: Emoji for "processing" status (default: `OnIt`). See [available emojis](https://open.larkoffice.com/document/server-docs/im-v1/message-reaction/emojis-introduce).
 > `doneEmoji`: Optional emoji for "completed" status (e.g., `DONE`, `OK`, `HEART`). When set, bot adds this reaction after removing `reactEmoji`.
+> `toolHintPrefix`: Prefix for inline tool hints in streaming cards (default: `🔧`).
 > `domain`: `"feishu"` (default) for China (open.feishu.cn), `"lark"` for international Lark (open.larksuite.com).
 
 **3. Run**
@@ -857,6 +903,56 @@ nanobot gateway
 
 </details>
 
+<details>
+<summary><b>Microsoft Teams</b> (MVP — DM only)</summary>
+
+> Direct-message text in/out, tenant-aware OAuth, conversation reference persistence.
+> Uses a public HTTPS webhook — no WebSocket; you need a tunnel or reverse proxy.
+
+**1. Install the optional dependency**
+
+```bash
+pip install nanobot-ai[msteams]
+```
+
+**2. Create a Teams / Azure bot app registration**
+
+Create or reuse a Microsoft Teams / Azure bot app registration. Set the bot messaging endpoint to a public HTTPS URL ending in `/api/messages`.
+
+**3. Configure**
+
+```json
+{
+  "channels": {
+    "msteams": {
+      "enabled": true,
+      "appId": "YOUR_APP_ID",
+      "appPassword": "YOUR_APP_SECRET",
+      "tenantId": "YOUR_TENANT_ID",
+      "host": "0.0.0.0",
+      "port": 3978,
+      "path": "/api/messages",
+      "allowFrom": ["*"],
+      "replyInThread": true,
+      "mentionOnlyResponse": "Hi — what can I help with?",
+      "validateInboundAuth": true
+    }
+  }
+}
+```
+
+> - `replyInThread: true` replies to the triggering Teams activity when a stored `activity_id` is available.
+> - `mentionOnlyResponse` controls what Nanobot receives when a user sends only a bot mention (`<at>Nanobot</at>`). Set to `""` to ignore mention-only messages.
+> - `validateInboundAuth: true` enables inbound Bot Framework bearer-token validation (signature, issuer, audience, lifetime, `serviceUrl`). This is the safe default for public deployments. Only set it to `false` for local development or tightly controlled testing.
+
+**4. Run**
+
+```bash
+nanobot gateway
+```
+
+</details>
+
 ## 🌐 Agent Social Network
 
 🐈 nanobot is capable of linking to the agent social network (agent community). **Just send one message and your nanobot joins automatically!**
@@ -918,6 +1014,7 @@ IMAP_PASSWORD=your-password-here
 > - **Voice transcription**: Voice messages (Telegram, WhatsApp) are automatically transcribed using Whisper. By default Groq is used (free tier). Set `"transcriptionProvider": "openai"` under `channels` to use OpenAI Whisper instead — the API key is picked from the matching provider config.
 > - **MiniMax Coding Plan**: Exclusive discount links for the nanobot community: [Overseas](https://platform.minimax.io/subscribe/coding-plan?code=9txpdXw04g&source=link) · [Mainland China](https://platform.minimaxi.com/subscribe/token-plan?code=GILTJpMTqZ&source=link)
 > - **MiniMax (Mainland China)**: If your API key is from MiniMax's mainland China platform (minimaxi.com), set `"apiBase": "https://api.minimaxi.com/v1"` in your minimax provider config.
+> - **MiniMax thinking mode**: Use `providers.minimaxAnthropic` when you want `reasoningEffort` / thinking mode. MiniMax exposes that capability through its Anthropic-compatible endpoint, so nanobot keeps it as a separate provider instead of guessing MiniMax-specific thinking parameters on the generic OpenAI-compatible `minimax` endpoint. It uses the same `MINIMAX_API_KEY`. Default Anthropic-compatible base URL: `https://api.minimax.io/anthropic`; for mainland China use `https://api.minimaxi.com/anthropic`.
 > - **VolcEngine / BytePlus Coding Plan**: Use dedicated providers `volcengineCodingPlan` or `byteplusCodingPlan` instead of the pay-per-use `volcengine` / `byteplus` providers.
 > - **Zhipu Coding Plan**: If you're on Zhipu's coding plan, set `"apiBase": "https://open.bigmodel.cn/api/coding/paas/v4"` in your zhipu provider config.
 > - **Alibaba Cloud BaiLian**: If you're using Alibaba Cloud BaiLian's OpenAI-compatible endpoint, set `"apiBase": "https://dashscope.aliyuncs.com/compatible-mode/v1"` in your dashscope provider config.
@@ -935,6 +1032,7 @@ IMAP_PASSWORD=your-password-here
 | `deepseek` | LLM (DeepSeek direct) | [platform.deepseek.com](https://platform.deepseek.com) |
 | `groq` | LLM + Voice transcription (Whisper, default) | [console.groq.com](https://console.groq.com) |
 | `minimax` | LLM (MiniMax direct) | [platform.minimaxi.com](https://platform.minimaxi.com) |
+| `minimax_anthropic` | LLM (MiniMax Anthropic-compatible endpoint, thinking mode) | [platform.minimaxi.com](https://platform.minimaxi.com) |
 | `gemini` | LLM (Gemini direct) | [aistudio.google.com](https://aistudio.google.com) |
 | `aihubmix` | LLM (API gateway, access to all models) | [aihubmix.com](https://aihubmix.com) |
 | `siliconflow` | LLM (SiliconFlow/硅基流动) | [siliconflow.cn](https://siliconflow.cn) |
@@ -943,6 +1041,7 @@ IMAP_PASSWORD=your-password-here
 | `zhipu` | LLM (Zhipu GLM) | [open.bigmodel.cn](https://open.bigmodel.cn) |
 | `mimo` | LLM (MiMo) | [platform.xiaomimimo.com](https://platform.xiaomimimo.com) |
 | `ollama` | LLM (local, Ollama) | — |
+| `lm_studio` | LLM (local, LM Studio) | — |
 | `mistral` | LLM | [docs.mistral.ai](https://docs.mistral.ai/) |
 | `stepfun` | LLM (Step Fun/阶跃星辰) | [platform.stepfun.com](https://platform.stepfun.com) |
 | `ovms` | LLM (local, OpenVINO Model Server) | [docs.openvino.ai](https://docs.openvino.ai/2026/model-server/ovms_docs_llm_quickstart.html) |
@@ -1030,7 +1129,7 @@ nanobot agent -c ~/.nanobot-telegram/config.json -w /tmp/nanobot-telegram-test -
 <details>
 <summary><b>Custom Provider (Any OpenAI-compatible API)</b></summary>
 
-Connects directly to any OpenAI-compatible endpoint — LM Studio, llama.cpp, Together AI, Fireworks, Azure OpenAI, or any self-hosted server. Model name is passed as-is.
+Connects directly to any OpenAI-compatible endpoint — llama.cpp, Together AI, Fireworks, Azure OpenAI, or any self-hosted server. Model name is passed as-is.
 
 ```json
 {
@@ -1048,7 +1147,31 @@ Connects directly to any OpenAI-compatible endpoint — LM Studio, llama.cpp, To
 }
 ```
 
-> For local servers that don't require a key, set `apiKey` to any non-empty string (e.g. `"no-key"`).
+> For local servers that don't require authentication, set `apiKey` to `null`.
+>
+> `custom` is the right choice for providers that expose an OpenAI-compatible **chat completions** API. It does **not** force third-party endpoints onto the OpenAI/Azure **Responses API**.
+>
+> If your proxy or gateway is specifically Responses-API-compatible, use the `azure_openai` provider shape instead and point `apiBase` at that endpoint:
+>
+> ```json
+> {
+>   "providers": {
+>     "azure_openai": {
+>       "apiKey": "your-api-key",
+>       "apiBase": "https://api.your-provider.com",
+>       "defaultModel": "your-model-name"
+>     }
+>   },
+>   "agents": {
+>     "defaults": {
+>       "provider": "azure_openai",
+>       "model": "your-model-name"
+>     }
+>   }
+> }
+> ```
+>
+> In short: **chat-completions-compatible endpoint → `custom`**; **Responses-compatible endpoint → `azure_openai`**.
 
 </details>
 
@@ -1083,6 +1206,40 @@ ollama run llama3.2
 
 </details>
 
+<details>
+<summary><b>LM Studio (local)</b></summary>
+
+[LM Studio](https://lmstudio.ai/) provides a local OpenAI-compatible server for running LLMs. Download models through the LM Studio UI, then start the local server.
+
+**1. Start LM Studio server:**
+- Launch LM Studio
+- Go to the "Local Server" tab
+- Load a model (e.g., Llama, Mistral, Qwen)
+- Click "Start Server" (default port: 1234)
+
+**2. Add to config** (partial — merge into `~/.nanobot/config.json`):
+```json
+{
+  "providers": {
+    "lm_studio": {
+      "apiKey": null,
+      "apiBase": "http://localhost:1234/v1"
+    }
+  },
+  "agents": {
+    "defaults": {
+      "provider": "lm_studio",
+      "model": "local-model"
+    }
+  }
+}
+```
+
+> **Note:** Set `apiKey` to `null` for LM Studio since it runs locally and doesn't require authentication. The model name should match what's shown in the LM Studio UI.
+> `provider: "auto"` also works when `providers.lm_studio.apiBase` is configured, but setting `"provider": "lm_studio"` is the clearest option.
+
+</details>
+
 <details>
 <summary><b>OpenVINO Model Server (local / OpenAI-compatible)</b></summary>
 
@@ -1170,12 +1327,12 @@ vllm serve meta-llama/Llama-3.1-8B-Instruct --port 8000
 
 **2. Add to config** (partial — merge into `~/.nanobot/config.json`):
 
-*Provider (key can be any non-empty string for local):*
+*Provider (set API key to null for local servers):*
 ```json
 {
   "providers": {
     "vllm": {
-      "apiKey": "dummy",
+      "apiKey": null,
       "apiBase": "http://localhost:8000/v1"
     }
   }
@@ -1310,6 +1467,7 @@ If you need to allow trusted private ranges such as Tailscale / CGNAT addresses,
 | `brave` | `apiKey` | `BRAVE_API_KEY` | No |
 | `tavily` | `apiKey` | `TAVILY_API_KEY` | No |
 | `jina` | `apiKey` | `JINA_API_KEY` | Free tier (10M tokens) |
+| `kagi` | `apiKey` | `KAGI_API_KEY` | No |
 | `searxng` | `baseUrl` | `SEARXNG_BASE_URL` | Yes (self-hosted) |
 | `duckduckgo` (default) | — | — | Yes |
 
@@ -1366,6 +1524,20 @@ If you need to allow trusted private ranges such as Tailscale / CGNAT addresses,
 }
 ```
 
+**Kagi:**
+```json
+{
+  "tools": {
+    "web": {
+      "search": {
+        "provider": "kagi",
+        "apiKey": "your-kagi-api-key"
+      }
+    }
+  }
+}
+```
+
 **SearXNG** (self-hosted, no API key needed):
 ```json
 {
@@ -1503,13 +1675,13 @@ MCP tools are automatically discovered and registered on startup. The LLM can us
 
 ### Auto Compact
 
-When a user is idle for longer than a configured TTL, nanobot **proactively** compresses the session context into a summary. This reduces token cost and first-token latency when the user returns — instead of re-processing a long stale context with an expired KV cache, the model receives a compact summary and fresh input.
+When a user is idle for longer than a configured threshold, nanobot **proactively** compresses the older part of the session context into a summary while keeping a recent legal suffix of live messages. This reduces token cost and first-token latency when the user returns — instead of re-processing a long stale context with an expired KV cache, the model receives a compact summary, the most recent live context, and fresh input.
 
 ```json
 {
   "agents": {
     "defaults": {
-      "sessionTtlMinutes": 15
+      "idleCompactAfterMinutes": 15
     }
   }
 }
@@ -1517,15 +1689,22 @@ When a user is idle for longer than a configured TTL, nanobot **proactively** co
 
 | Option | Default | Description |
 |--------|---------|-------------|
-| `agents.defaults.sessionTtlMinutes` | `0` (disabled) | Minutes of idle time before auto-compaction. Set to `0` to disable. Recommended: `15` — matches typical LLM KV cache expiration, so compacted sessions won't waste cache on cold entries. |
+| `agents.defaults.idleCompactAfterMinutes` | `0` (disabled) | Minutes of idle time before auto-compaction starts. Set to `0` to disable. Recommended: `15` — close to a typical LLM KV cache expiry window, so stale sessions get compacted before the user returns. |
+
+`sessionTtlMinutes` remains accepted as a legacy alias for backward compatibility, but `idleCompactAfterMinutes` is the preferred config key going forward.
 
 How it works:
 1. **Idle detection**: On each idle tick (~1 s), checks all sessions for expiration.
-2. **Background compaction**: Expired sessions are summarized via LLM, then cleared.
-3. **Summary injection**: When the user returns, the summary is injected as runtime context (one-shot, not persisted).
+2. **Background compaction**: Idle sessions summarize the older live prefix via LLM and keep the most recent legal suffix (currently 8 messages).
+3. **Summary injection**: When the user returns, the summary is injected as runtime context (one-shot, not persisted) alongside the retained recent suffix.
+4. **Restart-safe resume**: The summary is also mirrored into session metadata so it can still be recovered after a process restart.
 
-> [!TIP]
-> The summary survives bot restarts — it's stored in session metadata and recovered on the next message.
+> [!NOTE]
+> Mental model: "summarize older context, keep the freshest live turns, **and overwrite the session file with the compact form.**" It is not a full `session.clear()`, but it is a write — not a soft cursor move.
+>
+> Concretely, auto compact rewrites `sessions/<key>.jsonl` in place: older messages (including their structured `tool_calls` / `tool_call_id` / `reasoning_content`) are replaced by just the retained recent suffix (currently 8 messages), while the archived prefix is preserved only as a plain-text summary appended to `memory/history.jsonl` (or a `[RAW] ...` flattened dump if LLM summarization fails). The original structured JSON of those turns is no longer recoverable from the session file.
+>
+> This differs from the **token-driven soft consolidation** that fires when a prompt exceeds the context budget: that path only advances an internal `last_consolidated` cursor and leaves the session file untouched, so the raw tool-call trail stays on disk and can still be replayed or audited. If you rely on that trail for debugging or auditing, leave `idleCompactAfterMinutes` at the default `0` and let only the token-driven path run.
 
 ### Timezone
 
@@ -1549,6 +1728,52 @@ Common examples: `UTC`, `America/New_York`, `America/Los_Angeles`, `Europe/Londo
 
 > Need another timezone? Browse the full [IANA Time Zone Database](https://en.wikipedia.org/wiki/List_of_tz_database_time_zones).
 
+### Unified Session
+
+By default, each channel × chat ID combination gets its own session. If you use nanobot across multiple channels (e.g. Telegram + Discord + CLI) and want them to share the same conversation, enable `unifiedSession`:
+
+```json
+{
+  "agents": {
+    "defaults": {
+      "unifiedSession": true
+    }
+  }
+}
+```
+
+When enabled, all incoming messages — regardless of which channel they arrive on — are routed into a single shared session. Switching from Telegram to Discord (or any other channel) continues the same conversation seamlessly.
+
+| Behavior | `false` (default) | `true` |
+|----------|-------------------|--------|
+| Session key | `channel:chat_id` | `unified:default` |
+| Cross-channel continuity | No | Yes |
+| `/new` clears | Current channel session | Shared session |
+| `/stop` finds tasks | By channel session | By shared session |
+| Existing `session_key_override` (e.g. Telegram thread) | Respected | Still respected — not overwritten |
+
+> This is designed for single-user, multi-device setups. It is **off by default** — existing users see zero behavior change.
+
+### Disabled Skills
+
+nanobot ships with built-in skills, and your workspace can also define custom skills under `skills/`. If you want to hide specific skills from the agent, set `agents.defaults.disabledSkills` to a list of skill directory names:
+
+```json
+{
+  "agents": {
+    "defaults": {
+      "disabledSkills": ["github", "weather"]
+    }
+  }
+}
+```
+
+Disabled skills are excluded from the main agent's skill summary, from always-on skill injection, and from subagent skill summaries. This is useful when some bundled skills are unnecessary for your deployment or should not be exposed to end users.
+
+| Option | Default | Description |
+|--------|---------|-------------|
+| `agents.defaults.disabledSkills` | `[]` | List of skill directory names to exclude from loading. Applies to both built-in skills and workspace skills. |
+
 ## 🧩 Multiple Instances
 
 Run multiple nanobot instances simultaneously with separate configs and runtime data. Use `--config` as the main entrypoint. Optionally pass `--workspace` during `onboard` when you want to initialize or update the saved workspace for a specific instance.
@@ -1635,6 +1860,7 @@ Example config:
     }
   },
   "gateway": {
+    "host": "127.0.0.1",
     "port": 18790
   }
 }
@@ -1647,6 +1873,14 @@ nanobot gateway --config ~/.nanobot-telegram/config.json
 nanobot gateway --config ~/.nanobot-discord/config.json
 ```
 
+Each gateway instance also exposes a lightweight HTTP health endpoint on
+`gateway.host:gateway.port`. By default, the gateway binds to `127.0.0.1`,
+so the endpoint stays local unless you explicitly set `gateway.host` to a
+public or LAN-facing address.
+
+- `GET /health` returns `{"status":"ok"}`
+- Other paths return `404`
+
 Override workspace for one-off runs when needed:
 
 ```bash
@@ -1674,6 +1908,7 @@ time.
 
 - `memory/history.jsonl` stores append-only summarized history
 - `SOUL.md`, `USER.md`, and `memory/MEMORY.md` store long-term knowledge managed by Dream
+- `Dream` can also promote repeated workflows into reusable workspace skills under `skills/`
 - `Dream` runs on a schedule and can also be triggered manually
 - memory changes can be inspected and restored with built-in commands
 
@@ -1788,7 +2023,21 @@ By default, the API binds to `127.0.0.1:8900`. You can change this in `config.js
 - Session isolation: pass `"session_id"` in the request body to isolate conversations; omit for a shared default session (`api:default`)
 - Single-message input: each request must contain exactly one `user` message
 - Fixed model: omit `model`, or pass the same model shown by `/v1/models`
-- No streaming: `stream=true` is not supported
+- Streaming: set `stream=true` to receive Server-Sent Events (`text/event-stream`) with OpenAI-compatible delta chunks, terminated by `data: [DONE]`; omit or set `stream=false` for a single JSON response
+- **File uploads**: supports images, PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx) via JSON base64 or `multipart/form-data` (max 10MB per file)
+- API requests run in the synthetic `api` channel, so the `message` tool does **not** automatically deliver to Telegram/Discord/etc. To proactively send to another chat, call `message` with an explicit `channel` and `chat_id` for an enabled channel.
+
+Example tool call for cross-channel delivery from an API session:
+
+```json
+{
+  "content": "Build finished successfully.",
+  "channel": "telegram",
+  "chat_id": "123456789"
+}
+```
+
+If `channel` points to a channel that is not enabled in your config, nanobot will queue the outbound event but no platform delivery will occur.
 
 ### Endpoints
 
@@ -1807,6 +2056,44 @@ curl http://127.0.0.1:8900/v1/chat/completions \
   }'
 ```
 
+### File Upload (JSON base64)
+
+Send images inline using the OpenAI multimodal content format:
+
+```bash
+curl http://127.0.0.1:8900/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "messages": [{"role": "user", "content": [
+      {"type": "text", "text": "Describe this image"},
+      {"type": "image_url", "image_url": {"url": "data:image/png;base64,iVBOR..."}}
+    ]}]
+  }'
+```
+
+### File Upload (multipart/form-data)
+
+Upload any supported file type (images, PDF, Word, Excel, PPT) via multipart:
+
+```bash
+# Single file
+curl http://127.0.0.1:8900/v1/chat/completions \
+  -F "message=Summarize this report" \
+  -F "files=@report.docx"
+
+# Multiple files with session isolation
+curl http://127.0.0.1:8900/v1/chat/completions \
+  -F "message=Compare these files" \
+  -F "files=@chart.png" \
+  -F "files=@data.xlsx" \
+  -F "session_id=my-session"
+```
+
+Supported file types:
+- **Images**: PNG, JPEG, GIF, WebP (sent to AI as base64 for vision analysis)
+- **Documents**: PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx) (text extracted and sent to AI)
+- **Text**: TXT, Markdown, CSV, JSON, etc. (read directly)
+
 ### Python (`requests`)
 
 ```python

docs/CHANNEL_PLUGIN_GUIDE.md

@@ -43,18 +43,33 @@ from typing import Any
 
 from aiohttp import web
 from loguru import logger
+from pydantic import Field
 
 from nanobot.channels.base import BaseChannel
 from nanobot.bus.events import OutboundMessage
+from nanobot.bus.queue import MessageBus
+from nanobot.config.schema import Base
+
+
+class WebhookConfig(Base):
+    """Webhook channel configuration."""
+    enabled: bool = False
+    port: int = 9000
+    allow_from: list[str] = Field(default_factory=list)
 
 
 class WebhookChannel(BaseChannel):
     name = "webhook"
     display_name = "Webhook"
 
+    def __init__(self, config: Any, bus: MessageBus):
+        if isinstance(config, dict):
+            config = WebhookConfig(**config)
+        super().__init__(config, bus)
+
     @classmethod
     def default_config(cls) -> dict[str, Any]:
-        return {"enabled": False, "port": 9000, "allowFrom": []}
+        return WebhookConfig().model_dump(by_alias=True)
 
     async def start(self) -> None:
         """Start an HTTP server that listens for incoming messages.
@@ -63,7 +78,7 @@ class WebhookChannel(BaseChannel):
         If it returns, the channel is considered dead.
         """
         self._running = True
-        port = self.config.get("port", 9000)
+        port = self.config.port
 
         app = web.Application()
         app.router.add_post("/message", self._on_request)
@@ -214,7 +229,7 @@ nanobot channels login <channel_name> --force  # re-authenticate
 | Method / Property | Description |
 |-------------------|-------------|
 | `_handle_message(sender_id, chat_id, content, media?, metadata?, session_key?)` | **Call this when you receive a message.** Checks `is_allowed()`, then publishes to the bus. Automatically sets `_wants_stream` if `supports_streaming` is true. |
-| `is_allowed(sender_id)` | Checks against `config["allowFrom"]`; `"*"` allows all, `[]` denies all. |
+| `is_allowed(sender_id)` | Checks against `config.allow_from`; `"*"` allows all, `[]` denies all. |
 | `default_config()` (classmethod) | Returns default config dict for `nanobot onboard`. Override to declare your fields. |
 | `transcribe_audio(file_path)` | Transcribes audio via Groq Whisper (if configured). |
 | `supports_streaming` (property) | `True` when config has `"streaming": true` **and** subclass overrides `send_delta()`. |
@@ -275,7 +290,6 @@ async def send_delta(self, chat_id: str, delta: str, metadata: dict[str, Any] |
 |------|---------|
 | `_stream_delta: True` | A content chunk (delta contains the new text) |
 | `_stream_end: True` | Streaming finished (delta is empty) |
-| `_resuming: True` | More streaming rounds coming (e.g. tool call then another response) |
 
 ### Example: Webhook with Streaming
 
@@ -284,7 +298,9 @@ class WebhookChannel(BaseChannel):
     name = "webhook"
     display_name = "Webhook"
 
-    def __init__(self, config, bus):
+    def __init__(self, config: Any, bus: MessageBus):
+        if isinstance(config, dict):
+            config = WebhookConfig(**config)
         super().__init__(config, bus)
         self._buffers: dict[str, str] = {}
 
@@ -333,12 +349,48 @@ When `streaming` is `false` (default) or omitted, only `send()` is called — no
 
 ## Config
 
-Your channel receives config as a plain `dict`. Access fields with `.get()`:
+### Why Pydantic model is required
+
+`BaseChannel.is_allowed()` reads the permission list via `getattr(self.config, "allow_from", [])`. This works for Pydantic models where `allow_from` is a real Python attribute, but **fails silently for plain `dict`** — `dict` has no `allow_from` attribute, so `getattr` always returns the default `[]`, causing all messages to be denied.
+
+Built-in channels use Pydantic config models (subclassing `Base` from `nanobot.config.schema`). Plugin channels **must do the same**.
+
+### Pattern
+
+1. Define a Pydantic model inheriting from `nanobot.config.schema.Base`:
+
+```python
+from pydantic import Field
+from nanobot.config.schema import Base
+
+class WebhookConfig(Base):
+    """Webhook channel configuration."""
+    enabled: bool = False
+    port: int = 9000
+    allow_from: list[str] = Field(default_factory=list)
+```
+
+`Base` is configured with `alias_generator=to_camel` and `populate_by_name=True`, so JSON keys like `"allowFrom"` and `"allow_from"` are both accepted.
+
+2. Convert `dict` → model in `__init__`:
+
+```python
+from typing import Any
+from nanobot.bus.queue import MessageBus
+
+class WebhookChannel(BaseChannel):
+    def __init__(self, config: Any, bus: MessageBus):
+        if isinstance(config, dict):
+            config = WebhookConfig(**config)
+        super().__init__(config, bus)
+```
+
+3. Access config as attributes (not `.get()`):
 
 ```python
 async def start(self) -> None:
-    port = self.config.get("port", 9000)
-    token = self.config.get("token", "")
+    port = self.config.port
+    token = self.config.token
 ```
 
 `allowFrom` is handled automatically by `_handle_message()` — you don't need to check it yourself.
@@ -348,9 +400,11 @@ Override `default_config()` so `nanobot onboard` auto-populates `config.json`:
 ```python
 @classmethod
 def default_config(cls) -> dict[str, Any]:
-    return {"enabled": False, "port": 9000, "allowFrom": []}
+    return WebhookConfig().model_dump(by_alias=True)
 ```
 
+> **Note:** `default_config()` returns a plain `dict` (not a Pydantic model) because it's used to serialize into `config.json`. The recommended way is to instantiate your config model and call `model_dump(by_alias=True)` — this automatically uses camelCase keys (`allowFrom`) and keeps defaults in a single source of truth.
+
 If not overridden, the base class returns `{"enabled": false}`.
 
 ## Naming Convention

docs/MSTEAMS.md

@@ -1,68 +0,0 @@
-# Microsoft Teams (MVP)
-
-This repository includes a built-in `msteams` channel MVP for Microsoft Teams direct messages.
-
-## Current scope
-
-- Direct-message text in/out
-- Tenant-aware OAuth token acquisition
-- Conversation reference persistence for replies
-- Public HTTPS webhook support through a tunnel or reverse proxy
-
-## Not yet included
-
-- Group/channel handling
-- Attachments and cards
-- Polls
-- Richer Teams activity handling
-
-## Example config
-
-```json
-{
-  "channels": {
-    "msteams": {
-      "enabled": true,
-      "appId": "YOUR_APP_ID",
-      "appPassword": "YOUR_APP_SECRET",
-      "tenantId": "YOUR_TENANT_ID",
-      "host": "0.0.0.0",
-      "port": 3978,
-      "path": "/api/messages",
-      "allowFrom": ["*"],
-      "replyInThread": true,
-      "mentionOnlyResponse": "Hi — what can I help with?",
-      "validateInboundAuth": false,
-      "restartNotifyEnabled": false,
-      "restartNotifyPreMessage": "Nanobot agent initiated a gateway restart. I will message again when the gateway is back online.",
-      "restartNotifyPostMessage": "Nanobot gateway is back online."
-    }
-  }
-}
-```
-
-## Behavior notes
-
-- `replyInThread: true` replies to the triggering Teams activity when a stored `activity_id` is available.
-- `replyInThread: false` posts replies as normal conversation messages.
-- If `replyInThread` is enabled but no `activity_id` is stored, Nanobot falls back to a normal conversation message.
-- `mentionOnlyResponse` controls what Nanobot receives when a user sends only a bot mention such as `<at>Nanobot</at>`.
-- Set `mentionOnlyResponse` to an empty string to ignore mention-only messages.
-- `validateInboundAuth: true` enables inbound Bot Framework bearer-token validation.
-- `validateInboundAuth: false` leaves inbound auth unenforced, which is safer while first validating a new relay, tunnel, or proxy path.
-- When enabled, Nanobot validates the inbound bearer token signature, issuer, audience, token lifetime, and `serviceUrl` claim when present.
-- `restartNotifyEnabled: true` enables optional Teams restart-notification configuration for external wrapper-script driven restarts.
-- `restartNotifyPreMessage` and `restartNotifyPostMessage` control the before/after announcement text used by that external wrapper.
-
-## Setup notes
-
-1. Create or reuse a Microsoft Teams / Azure bot app registration.
-2. Set the bot messaging endpoint to a public HTTPS URL ending in `/api/messages`.
-3. Forward that public endpoint to `http://localhost:3978/api/messages`.
-4. Start Nanobot with:
-
-```bash
-nanobot gateway
-```
-
-5. Optional: if you use an external restart wrapper (for example a script that stops and restarts the gateway), you can enable Teams restart announcements with `restartNotifyEnabled: true` and have the wrapper send `restartNotifyPreMessage` before restart and `restartNotifyPostMessage` after the gateway is back online.

docs/MY_TOOL.md

@@ -0,0 +1,207 @@
+# My Tool
+
+Let the agent sense and adjust its own runtime state — like asking a coworker "are you busy? can you switch to a bigger monitor?"
+
+## Why You Need It
+
+Normal tools let the agent operate on the outside world (read/write files, search code). But the agent knows nothing about itself — it doesn't know which model it's running on, how many iterations are left, or how many tokens it has consumed.
+
+My tool fills this gap. With it, the agent can:
+
+- **Know who it is**: What model am I using? Where is my workspace? How many iterations remain?
+- **Adapt on the fly**: Complex task? Expand the context window. Simple chat? Switch to a faster model.
+- **Remember across turns**: Store notes in your scratchpad that persist into the next conversation turn.
+
+## Configuration
+
+Enabled by default (read-only mode). The agent can check its state but not set it.
+
+```yaml
+tools:
+  my:
+    enable: true       # default: true
+    allow_set: false   # default: false (read-only)
+```
+
+To allow the agent to set its configuration (e.g. switch models, adjust parameters), set `tools.my.allow_set: true`.
+
+Legacy `tools.myEnabled` / `tools.mySet` keys are auto-migrated on load, and
+rewritten in-place the next time `nanobot onboard` refreshes the config.
+
+All modifications are held in memory only — restart restores defaults.
+
+---
+
+## check — Check "my" current state
+
+Without parameters, returns a key config overview:
+
+```
+my(action="check")
+# → max_iterations: 40
+#   context_window_tokens: 65536
+#   model: 'anthropic/claude-sonnet-4-20250514'
+#   workspace: PosixPath('/tmp/workspace')
+#   provider_retry_mode: 'standard'
+#   max_tool_result_chars: 16000
+#   _current_iteration: 3
+#   _last_usage: {'prompt_tokens': 45000, 'completion_tokens': 8000}
+#   Note: prompt_tokens is cumulative across all turns, not current context window occupancy.
+```
+
+With a key parameter, drill into a specific config:
+
+```
+my(action="check", key="_last_usage.prompt_tokens")
+# → How many prompt tokens I've used so far
+
+my(action="check", key="model")
+# → What model I'm currently running on
+
+my(action="check", key="web_config.enable")
+# → Whether web search is enabled
+```
+
+### What you can do with it
+
+| Scenario | How |
+|----------|-----|
+| "What model are you using?" | `check("model")` |
+| "How many more tool calls can you make?" | `check("max_iterations")` minus `check("_current_iteration")` |
+| "How many tokens has this conversation used?" | `check("_last_usage")` — cumulative across all turns |
+| "Where is your working directory?" | `check("workspace")` |
+| "Show me your full config" | `check()` |
+| "Are there any subagents running?" | `check("subagents")` — shows phase, iteration, elapsed time, tool events |
+
+---
+
+## set — Runtime tuning
+
+Changes take effect immediately, no restart required.
+
+```
+my(action="set", key="max_iterations", value=80)
+# → Bump iteration limit from 40 to 80
+
+my(action="set", key="model", value="fast-model")
+# → Switch to a faster model
+
+my(action="set", key="context_window_tokens", value=131072)
+# → Expand context window for long documents
+```
+
+You can also store custom state in your scratchpad:
+
+```
+my(action="set", key="current_project", value="nanobot")
+my(action="set", key="user_style_preference", value="concise")
+my(action="set", key="task_complexity", value="high")
+# → These values persist into the next conversation turn
+```
+
+### Protected parameters
+
+These parameters have type and range validation — invalid values are rejected:
+
+| Parameter | Type | Range | Purpose |
+|-----------|------|-------|---------|
+| `max_iterations` | int | 1–100 | Max tool calls per conversation turn |
+| `context_window_tokens` | int | 4,096–1,000,000 | Context window size |
+| `model` | str | non-empty | LLM model to use |
+
+Other parameters (e.g. `workspace`, `provider_retry_mode`, `max_tool_result_chars`) can be set freely, as long as the value is JSON-safe.
+
+---
+
+## Practical Scenarios
+
+### "This task is complex, I need more room"
+
+```
+Agent: This codebase is large, let me expand my context window to handle it.
+→ my(action="set", key="context_window_tokens", value=131072)
+```
+
+### "Simple question, don't waste compute"
+
+```
+Agent: This is a straightforward question, let me switch to a faster model.
+→ my(action="set", key="model", value="fast-model")
+```
+
+### "Remember user preferences across turns"
+
+```
+Turn 1: my(action="set", key="user_prefers_concise", value=True)
+Turn 2: my(action="check", key="user_prefers_concise")
+# → True (still remembers the user likes concise replies)
+```
+
+### "Self-diagnosis"
+
+```
+User: "Why aren't you searching the web?"
+Agent: Let me check my web config.
+→ my(action="check", key="web_config.enable")
+# → False
+Agent: Web search is disabled — please set web.enable: true in your config.
+```
+
+### "Token budget management"
+
+```
+Agent: Let me check how much budget I have left.
+→ my(action="check", key="_last_usage")
+# → {"prompt_tokens": 45000, "completion_tokens": 8000}
+Agent: I've used ~53k tokens total so far. I'll keep my remaining replies concise.
+```
+
+### "Subagent monitoring"
+
+```
+Agent: Let me check on the background tasks.
+→ my(action="check", key="subagents")
+# → 2 subagent(s):
+#   [task-1] 'Code review'
+#     phase: running, iteration: 5, elapsed: 12.3s
+#     tools: read(✓), grep(✓)
+#     usage: {'prompt_tokens': 8000, 'completion_tokens': 1200}
+#   [task-2] 'Write tests'
+#     phase: pending, iteration: 0, elapsed: 0.2s
+#     tools: none
+Agent: The code review is progressing well. The test task hasn't started yet.
+```
+
+---
+
+## Safety Mechanisms
+
+Core design principle: **All modifications live in memory only. Restart restores defaults.** The agent cannot cause persistent damage.
+
+### Off-limits (BLOCKED)
+
+Cannot be checked or modified — fully hidden:
+
+| Category | Attributes | Reason |
+|----------|-----------|--------|
+| Core infrastructure | `bus`, `provider`, `_running` | Changes would crash the system |
+| Tool registry | `tools` | Must not remove its own tools |
+| Subsystems | `runner`, `sessions`, `consolidator`, etc. | Affects other users/sessions |
+| Sensitive data | `_mcp_servers`, `_pending_queues`, etc. | Contains credentials and message routing |
+| Security boundaries | `restrict_to_workspace`, `channels_config` | Bypassing would violate isolation |
+| Python internals | `__class__`, `__dict__`, etc. | Prevents sandbox escape |
+
+### Read-only (check only)
+
+Can be checked but not set:
+
+| Category | Attributes | Reason |
+|----------|-----------|--------|
+| Subagent manager | `subagents` | Observable, but replacing breaks the system |
+| Execution config | `exec_config` | Can check sandbox/enable status, cannot change it |
+| Web config | `web_config` | Can check enable status, cannot change it |
+| Iteration counter | `_current_iteration` | Updated by runner only |
+
+### Sensitive field protection
+
+Sub-fields matching sensitive names (`api_key`, `password`, `secret`, `token`, etc.) are blocked from both check and set, regardless of parent path. This prevents credential leaks via dot-path traversal (e.g. `web_config.search.api_key`).

nanobot/__init__.py

@@ -2,7 +2,29 @@
 nanobot - A lightweight AI agent framework
 """
 
-__version__ = "0.1.5"
+from importlib.metadata import PackageNotFoundError, version as _pkg_version
+from pathlib import Path
+import tomllib
+
+
+def _read_pyproject_version() -> str | None:
+    """Read the source-tree version when package metadata is unavailable."""
+    pyproject = Path(__file__).resolve().parent.parent / "pyproject.toml"
+    if not pyproject.exists():
+        return None
+    data = tomllib.loads(pyproject.read_text(encoding="utf-8"))
+    return data.get("project", {}).get("version")
+
+
+def _resolve_version() -> str:
+    try:
+        return _pkg_version("nanobot-ai")
+    except PackageNotFoundError:
+        # Source checkouts often import nanobot without installed dist-info.
+        return _read_pyproject_version() or "0.1.5.post1"
+
+
+__version__ = _resolve_version()
 __logo__ = "🐈"
 
 from nanobot.nanobot import Nanobot, RunResult

nanobot/agent/autocompact.py

@@ -2,17 +2,20 @@
 
 from __future__ import annotations
 
+from collections.abc import Collection
 from datetime import datetime
-from typing import TYPE_CHECKING, Callable, Coroutine
+from typing import TYPE_CHECKING, Any, Callable, Coroutine
 
 from loguru import logger
+from nanobot.session.manager import Session, SessionManager
 
 if TYPE_CHECKING:
     from nanobot.agent.memory import Consolidator
-    from nanobot.session.manager import Session, SessionManager
 
 
 class AutoCompact:
+    _RECENT_SUFFIX_MESSAGES = 8
+
     def __init__(self, sessions: SessionManager, consolidator: Consolidator,
                  session_ttl_minutes: int = 0):
         self.sessions = sessions
@@ -21,47 +24,83 @@ class AutoCompact:
         self._archiving: set[str] = set()
         self._summaries: dict[str, tuple[str, datetime]] = {}
 
-    def _is_expired(self, ts: datetime | str | None) -> bool:
+    def _is_expired(self, ts: datetime | str | None,
+                    now: datetime | None = None) -> bool:
         if self._ttl <= 0 or not ts:
             return False
         if isinstance(ts, str):
             ts = datetime.fromisoformat(ts)
-        return (datetime.now() - ts).total_seconds() >= self._ttl * 60
+        return ((now or datetime.now()) - ts).total_seconds() >= self._ttl * 60
 
     @staticmethod
     def _format_summary(text: str, last_active: datetime) -> str:
         idle_min = int((datetime.now() - last_active).total_seconds() / 60)
         return f"Inactive for {idle_min} minutes.\nPrevious conversation summary: {text}"
 
-    def check_expired(self, schedule_background: Callable[[Coroutine], None]) -> None:
+    def _split_unconsolidated(
+        self, session: Session,
+    ) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
+        """Split live session tail into archiveable prefix and retained recent suffix."""
+        tail = list(session.messages[session.last_consolidated:])
+        if not tail:
+            return [], []
+
+        probe = Session(
+            key=session.key,
+            messages=tail.copy(),
+            created_at=session.created_at,
+            updated_at=session.updated_at,
+            metadata={},
+            last_consolidated=0,
+        )
+        probe.retain_recent_legal_suffix(self._RECENT_SUFFIX_MESSAGES)
+        kept = probe.messages
+        cut = len(tail) - len(kept)
+        return tail[:cut], kept
+
+    def check_expired(self, schedule_background: Callable[[Coroutine], None],
+                      active_session_keys: Collection[str] = ()) -> None:
+        """Schedule archival for idle sessions, skipping those with in-flight agent tasks."""
+        now = datetime.now()
         for info in self.sessions.list_sessions():
             key = info.get("key", "")
-            if key and key not in self._archiving and self._is_expired(info.get("updated_at")):
+            if not key or key in self._archiving:
+                continue
+            if key in active_session_keys:
+                continue
+            if self._is_expired(info.get("updated_at"), now):
                 self._archiving.add(key)
-                logger.debug("Auto-compact: scheduling archival for {} (idle > {} min)", key, self._ttl)
                 schedule_background(self._archive(key))
 
     async def _archive(self, key: str) -> None:
         try:
             self.sessions.invalidate(key)
             session = self.sessions.get_or_create(key)
-            msgs = session.messages[session.last_consolidated:]
-            if not msgs:
-                logger.debug("Auto-compact: skipping {}, no un-consolidated messages", key)
+            archive_msgs, kept_msgs = self._split_unconsolidated(session)
+            if not archive_msgs and not kept_msgs:
                 session.updated_at = datetime.now()
                 self.sessions.save(session)
                 return
-            n = len(msgs)
+
             last_active = session.updated_at
-            await self.consolidator.archive(msgs)
-            entry = self.consolidator.get_last_history_entry()
-            summary = (entry or {}).get("content", "")
+            summary = ""
+            if archive_msgs:
+                summary = await self.consolidator.archive(archive_msgs) or ""
             if summary and summary != "(nothing)":
                 self._summaries[key] = (summary, last_active)
                 session.metadata["_last_summary"] = {"text": summary, "last_active": last_active.isoformat()}
-            session.clear()
+            session.messages = kept_msgs
+            session.last_consolidated = 0
+            session.updated_at = datetime.now()
             self.sessions.save(session)
-            logger.info("Auto-compact: archived {} ({} messages, summary={})", key, n, bool(summary))
+            if archive_msgs:
+                logger.info(
+                    "Auto-compact: archived {} (archived={}, kept={}, summary={})",
+                    key,
+                    len(archive_msgs),
+                    len(kept_msgs),
+                    bool(summary),
+                )
         except Exception:
             logger.exception("Auto-compact: failed for {}", key)
         finally:
@@ -71,11 +110,13 @@ class AutoCompact:
         if key in self._archiving or self._is_expired(session.updated_at):
             logger.info("Auto-compact: reloading session {} (archiving={})", key, key in self._archiving)
             session = self.sessions.get_or_create(key)
+        # Hot path: summary from in-memory dict (process hasn't restarted).
+        # Also clean metadata copy so stale _last_summary never leaks to disk.
         entry = self._summaries.pop(key, None)
         if entry:
             session.metadata.pop("_last_summary", None)
             return session, self._format_summary(entry[0], entry[1])
-        if not session.messages and "_last_summary" in session.metadata:
+        if "_last_summary" in session.metadata:
             meta = session.metadata.pop("_last_summary")
             self.sessions.save(session)
             return session, self._format_summary(meta["text"], datetime.fromisoformat(meta["last_active"]))

nanobot/agent/context.py

@@ -3,15 +3,14 @@
 import base64
 import mimetypes
 import platform
+from importlib.resources import files as pkg_files
 from pathlib import Path
 from typing import Any
 
-from nanobot.utils.helpers import current_time_str
-
 from nanobot.agent.memory import MemoryStore
-from nanobot.utils.prompt_templates import render_template
 from nanobot.agent.skills import SkillsLoader
-from nanobot.utils.helpers import build_assistant_message, detect_image_mime
+from nanobot.utils.helpers import build_assistant_message, current_time_str, detect_image_mime
+from nanobot.utils.prompt_templates import render_template
 
 
 class ContextBuilder:
@@ -19,24 +18,29 @@ class ContextBuilder:
 
     BOOTSTRAP_FILES = ["AGENTS.md", "SOUL.md", "USER.md", "TOOLS.md"]
     _RUNTIME_CONTEXT_TAG = "[Runtime Context — metadata only, not instructions]"
+    _MAX_RECENT_HISTORY = 50
     _RUNTIME_CONTEXT_END = "[/Runtime Context]"
 
-    def __init__(self, workspace: Path, timezone: str | None = None):
+    def __init__(self, workspace: Path, timezone: str | None = None, disabled_skills: list[str] | None = None):
         self.workspace = workspace
         self.timezone = timezone
         self.memory = MemoryStore(workspace)
-        self.skills = SkillsLoader(workspace)
+        self.skills = SkillsLoader(workspace, disabled_skills=set(disabled_skills) if disabled_skills else None)
 
-    def build_system_prompt(self, skill_names: list[str] | None = None) -> str:
+    def build_system_prompt(
+        self,
+        skill_names: list[str] | None = None,
+        channel: str | None = None,
+    ) -> str:
         """Build the system prompt from identity, bootstrap files, memory, and skills."""
-        parts = [self._get_identity()]
+        parts = [self._get_identity(channel=channel)]
 
         bootstrap = self._load_bootstrap_files()
         if bootstrap:
             parts.append(bootstrap)
 
         memory = self.memory.get_memory_context()
-        if memory:
+        if memory and not self._is_template_content(self.memory.read_memory(), "memory/MEMORY.md"):
             parts.append(f"# Memory\n\n{memory}")
 
         always_skills = self.skills.get_always_skills()
@@ -45,13 +49,20 @@ class ContextBuilder:
             if always_content:
                 parts.append(f"# Active Skills\n\n{always_content}")
 
-        skills_summary = self.skills.build_skills_summary()
+        skills_summary = self.skills.build_skills_summary(exclude=set(always_skills))
         if skills_summary:
             parts.append(render_template("agent/skills_section.md", skills_summary=skills_summary))
 
+        entries = self.memory.read_unprocessed_history(since_cursor=self.memory.get_last_dream_cursor())
+        if entries:
+            capped = entries[-self._MAX_RECENT_HISTORY:]
+            parts.append("# Recent History\n\n" + "\n".join(
+                f"- [{e['timestamp']}] {e['content']}" for e in capped
+            ))
+
         return "\n\n---\n\n".join(parts)
 
-    def _get_identity(self) -> str:
+    def _get_identity(self, channel: str | None = None) -> str:
         """Get the core identity section."""
         workspace_path = str(self.workspace.expanduser().resolve())
         system = platform.system()
@@ -62,6 +73,7 @@ class ContextBuilder:
             workspace_path=workspace_path,
             runtime=runtime,
             platform_policy=render_template("agent/platform_policy.md", system=system),
+            channel=channel or "",
         )
 
     @staticmethod
@@ -103,6 +115,17 @@ class ContextBuilder:
 
         return "\n\n".join(parts) if parts else ""
 
+    @staticmethod
+    def _is_template_content(content: str, template_path: str) -> bool:
+        """Check if *content* is identical to the bundled template (user hasn't customized it)."""
+        try:
+            tpl = pkg_files("nanobot") / "templates" / template_path
+            if tpl.is_file():
+                return content.strip() == tpl.read_text(encoding="utf-8").strip()
+        except Exception:
+            pass
+        return False
+
     def build_messages(
         self,
         history: list[dict[str, Any]],
@@ -125,7 +148,7 @@ class ContextBuilder:
         else:
             merged = [{"type": "text", "text": runtime_ctx}] + user_content
         messages = [
-            {"role": "system", "content": self.build_system_prompt(skill_names)},
+            {"role": "system", "content": self.build_system_prompt(skill_names, channel=channel)},
             *history,
         ]
         if messages[-1].get("role") == current_role:
@@ -147,7 +170,6 @@ class ContextBuilder:
             if not p.is_file():
                 continue
             raw = p.read_bytes()
-            # Detect real MIME type from magic bytes; fallback to filename guess
             mime = detect_image_mime(raw) or mimetypes.guess_type(path)[0]
             if not mime or not mime.startswith("image/"):
                 continue

nanobot/agent/hook.py

@@ -29,6 +29,9 @@ class AgentHookContext:
 class AgentHook:
     """Minimal lifecycle surface for shared runner customization."""
 
+    def __init__(self, reraise: bool = False) -> None:
+        self._reraise = reraise
+
     def wants_streaming(self) -> bool:
         return False
 
@@ -62,6 +65,7 @@ class CompositeHook(AgentHook):
     __slots__ = ("_hooks",)
 
     def __init__(self, hooks: list[AgentHook]) -> None:
+        super().__init__()
         self._hooks = list(hooks)
 
     def wants_streaming(self) -> bool:
@@ -69,6 +73,10 @@ class CompositeHook(AgentHook):
 
     async def _for_each_hook_safe(self, method_name: str, *args: Any, **kwargs: Any) -> None:
         for h in self._hooks:
+            if getattr(h, "_reraise", False):
+                await getattr(h, method_name)(*args, **kwargs)
+                continue
+
             try:
                 await getattr(h, method_name)(*args, **kwargs)
             except Exception:

nanobot/agent/loop.py

@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import asyncio
+import dataclasses
 import json
 import os
 import time
@@ -12,35 +13,42 @@ from typing import TYPE_CHECKING, Any, Awaitable, Callable
 
 from loguru import logger
 
-from nanobot.agent.auto_compact import AutoCompact
+from nanobot.agent.autocompact import AutoCompact
 from nanobot.agent.context import ContextBuilder
 from nanobot.agent.hook import AgentHook, AgentHookContext, CompositeHook
 from nanobot.agent.memory import Consolidator, Dream
-from nanobot.agent.runner import AgentRunSpec, AgentRunner
+from nanobot.agent.runner import _MAX_INJECTIONS_PER_TURN, AgentRunner, AgentRunSpec
+from nanobot.agent.skills import BUILTIN_SKILLS_DIR
 from nanobot.agent.subagent import SubagentManager
 from nanobot.agent.tools.cron import CronTool
-from nanobot.agent.skills import BUILTIN_SKILLS_DIR
 from nanobot.agent.tools.filesystem import EditFileTool, ListDirTool, ReadFileTool, WriteFileTool
 from nanobot.agent.tools.message import MessageTool
+from nanobot.agent.tools.notebook import NotebookEditTool
 from nanobot.agent.tools.registry import ToolRegistry
 from nanobot.agent.tools.search import GlobTool, GrepTool
 from nanobot.agent.tools.shell import ExecTool
+from nanobot.agent.tools.self import MyTool
 from nanobot.agent.tools.spawn import SpawnTool
 from nanobot.agent.tools.web import WebFetchTool, WebSearchTool
 from nanobot.bus.events import InboundMessage, OutboundMessage
-from nanobot.command import CommandContext, CommandRouter, register_builtin_commands
 from nanobot.bus.queue import MessageBus
+from nanobot.command import CommandContext, CommandRouter, register_builtin_commands
 from nanobot.config.schema import AgentDefaults
 from nanobot.providers.base import LLMProvider
 from nanobot.session.manager import Session, SessionManager
-from nanobot.utils.helpers import image_placeholder_text, truncate_text
+from nanobot.utils.document import extract_documents
+from nanobot.utils.helpers import image_placeholder_text
+from nanobot.utils.helpers import truncate_text as truncate_text_fn
 from nanobot.utils.runtime import EMPTY_FINAL_RESPONSE_MESSAGE
 
 if TYPE_CHECKING:
-    from nanobot.config.schema import ChannelsConfig, ExecToolConfig, WebToolsConfig
+    from nanobot.config.schema import ChannelsConfig, ExecToolConfig, ToolsConfig, WebToolsConfig
     from nanobot.cron.service import CronService
 
 
+UNIFIED_SESSION_KEY = "unified:default"
+
+
 class _LoopHook(AgentHook):
     """Core hook for the main loop."""
 
@@ -55,6 +63,7 @@ class _LoopHook(AgentHook):
         chat_id: str = "direct",
         message_id: str | None = None,
     ) -> None:
+        super().__init__(reraise=True)
         self._loop = agent_loop
         self._on_progress = on_progress
         self._on_stream = on_stream
@@ -73,7 +82,7 @@ class _LoopHook(AgentHook):
         prev_clean = strip_think(self._stream_buf)
         self._stream_buf += delta
         new_clean = strip_think(self._stream_buf)
-        incremental = new_clean[len(prev_clean):]
+        incremental = new_clean[len(prev_clean) :]
         if incremental and self._on_stream:
             await self._on_stream(incremental)
 
@@ -82,6 +91,9 @@ class _LoopHook(AgentHook):
             await self._on_stream_end(resuming=resuming)
         self._stream_buf = ""
 
+    async def before_iteration(self, context: AgentHookContext) -> None:
+        self._loop._current_iteration = context.iteration
+
     async def before_execute_tools(self, context: AgentHookContext) -> None:
         if self._on_progress:
             if not self._on_stream:
@@ -110,43 +122,6 @@ class _LoopHook(AgentHook):
         return self._loop._strip_think(content)
 
 
-class _LoopHookChain(AgentHook):
-    """Run the core hook before extra hooks."""
-
-    __slots__ = ("_primary", "_extras")
-
-    def __init__(self, primary: AgentHook, extra_hooks: list[AgentHook]) -> None:
-        self._primary = primary
-        self._extras = CompositeHook(extra_hooks)
-
-    def wants_streaming(self) -> bool:
-        return self._primary.wants_streaming() or self._extras.wants_streaming()
-
-    async def before_iteration(self, context: AgentHookContext) -> None:
-        await self._primary.before_iteration(context)
-        await self._extras.before_iteration(context)
-
-    async def on_stream(self, context: AgentHookContext, delta: str) -> None:
-        await self._primary.on_stream(context, delta)
-        await self._extras.on_stream(context, delta)
-
-    async def on_stream_end(self, context: AgentHookContext, *, resuming: bool) -> None:
-        await self._primary.on_stream_end(context, resuming=resuming)
-        await self._extras.on_stream_end(context, resuming=resuming)
-
-    async def before_execute_tools(self, context: AgentHookContext) -> None:
-        await self._primary.before_execute_tools(context)
-        await self._extras.before_execute_tools(context)
-
-    async def after_iteration(self, context: AgentHookContext) -> None:
-        await self._primary.after_iteration(context)
-        await self._extras.after_iteration(context)
-
-    def finalize_content(self, context: AgentHookContext, content: str | None) -> str | None:
-        content = self._primary.finalize_content(context, content)
-        return self._extras.finalize_content(context, content)
-
-
 class AgentLoop:
     """
     The agent loop is the core processing engine.
@@ -160,6 +135,7 @@ class AgentLoop:
     """
 
     _RUNTIME_CHECKPOINT_KEY = "runtime_checkpoint"
+    _PENDING_USER_TURN_KEY = "pending_user_turn"
 
     def __init__(
         self,
@@ -181,10 +157,15 @@ class AgentLoop:
         channels_config: ChannelsConfig | None = None,
         timezone: str | None = None,
         session_ttl_minutes: int = 0,
+        consolidation_ratio: float = 0.5,
         hooks: list[AgentHook] | None = None,
+        unified_session: bool = False,
+        disabled_skills: list[str] | None = None,
+        tools_config: ToolsConfig | None = None,
     ):
-        from nanobot.config.schema import ExecToolConfig, WebToolsConfig
+        from nanobot.config.schema import ExecToolConfig, ToolsConfig, WebToolsConfig
 
+        _tc = tools_config or ToolsConfig()
         defaults = AgentDefaults()
         self.bus = bus
         self.channels_config = channels_config
@@ -214,7 +195,7 @@ class AgentLoop:
         self._last_usage: dict[str, int] = {}
         self._extra_hooks: list[AgentHook] = hooks or []
 
-        self.context = ContextBuilder(workspace, timezone=timezone)
+        self.context = ContextBuilder(workspace, timezone=timezone, disabled_skills=disabled_skills)
         self.sessions = session_manager or SessionManager(workspace)
         self.tools = ToolRegistry()
         self.runner = AgentRunner(provider)
@@ -227,11 +208,12 @@ class AgentLoop:
             max_tool_result_chars=self.max_tool_result_chars,
             exec_config=self.exec_config,
             restrict_to_workspace=restrict_to_workspace,
+            disabled_skills=disabled_skills,
         )
-
+        self._unified_session = unified_session
         self._running = False
         self._mcp_servers = mcp_servers or {}
-        self._mcp_stack: AsyncExitStack | None = None
+        self._mcp_stacks: dict[str, AsyncExitStack] = {}
         self._mcp_connected = False
         self._mcp_connecting = False
         self._active_tasks: dict[str, list[asyncio.Task]] = {}  # session_key -> tasks
@@ -251,10 +233,11 @@ class AgentLoop:
             provider=provider,
             model=self.model,
             sessions=self.sessions,
-            context_window_tokens=context_window_tokens,
+            context_window_tokens=self.context_window_tokens,
             build_messages=self.context.build_messages,
             get_tool_definitions=self.tools.get_definitions,
             max_completion_tokens=provider.generation.max_tokens,
+            consolidation_ratio=consolidation_ratio,
         )
         self.auto_compact = AutoCompact(
             sessions=self.sessions,
@@ -267,28 +250,44 @@ class AgentLoop:
             model=self.model,
         )
         self._register_default_tools()
+        if _tc.my.enable:
+            self.tools.register(MyTool(loop=self, modify_allowed=_tc.my.allow_set))
+        self._runtime_vars: dict[str, Any] = {}
+        self._current_iteration: int = 0
         self.commands = CommandRouter()
         register_builtin_commands(self.commands)
 
     def _register_default_tools(self) -> None:
         """Register the default set of tools."""
-        allowed_dir = self.workspace if (self.restrict_to_workspace or self.exec_config.sandbox) else None
+        allowed_dir = (
+            self.workspace if (self.restrict_to_workspace or self.exec_config.sandbox) else None
+        )
         extra_read = [BUILTIN_SKILLS_DIR] if allowed_dir else None
-        self.tools.register(ReadFileTool(workspace=self.workspace, allowed_dir=allowed_dir, extra_allowed_dirs=extra_read))
+        self.tools.register(
+            ReadFileTool(
+                workspace=self.workspace, allowed_dir=allowed_dir, extra_allowed_dirs=extra_read
+            )
+        )
         for cls in (WriteFileTool, EditFileTool, ListDirTool):
             self.tools.register(cls(workspace=self.workspace, allowed_dir=allowed_dir))
         for cls in (GlobTool, GrepTool):
             self.tools.register(cls(workspace=self.workspace, allowed_dir=allowed_dir))
+        self.tools.register(NotebookEditTool(workspace=self.workspace, allowed_dir=allowed_dir))
         if self.exec_config.enable:
-            self.tools.register(ExecTool(
-                working_dir=str(self.workspace),
-                timeout=self.exec_config.timeout,
-                restrict_to_workspace=self.restrict_to_workspace,
-                sandbox=self.exec_config.sandbox,
-                path_append=self.exec_config.path_append,
-            ))
+            self.tools.register(
+                ExecTool(
+                    working_dir=str(self.workspace),
+                    timeout=self.exec_config.timeout,
+                    restrict_to_workspace=self.restrict_to_workspace,
+                    sandbox=self.exec_config.sandbox,
+                    path_append=self.exec_config.path_append,
+                    allowed_env_keys=self.exec_config.allowed_env_keys,
+                )
+            )
         if self.web_config.enable:
-            self.tools.register(WebSearchTool(config=self.web_config.search, proxy=self.web_config.proxy))
+            self.tools.register(
+                WebSearchTool(config=self.web_config.search, proxy=self.web_config.proxy)
+            )
             self.tools.register(WebFetchTool(proxy=self.web_config.proxy))
         self.tools.register(MessageTool(send_callback=self.bus.publish_outbound))
         self.tools.register(SpawnTool(manager=self.subagents))
@@ -303,25 +302,25 @@ class AgentLoop:
             return
         self._mcp_connecting = True
         from nanobot.agent.tools.mcp import connect_mcp_servers
+
         try:
-            self._mcp_stack = AsyncExitStack()
-            await self._mcp_stack.__aenter__()
-            await connect_mcp_servers(self._mcp_servers, self.tools, self._mcp_stack)
-            self._mcp_connected = True
+            self._mcp_stacks = await connect_mcp_servers(self._mcp_servers, self.tools)
+            if self._mcp_stacks:
+                self._mcp_connected = True
+            else:
+                logger.warning("No MCP servers connected successfully (will retry next message)")
+        except asyncio.CancelledError:
+            logger.warning("MCP connection cancelled (will retry next message)")
+            self._mcp_stacks.clear()
         except BaseException as e:
             logger.error("Failed to connect MCP servers (will retry next message): {}", e)
-            if self._mcp_stack:
-                try:
-                    await self._mcp_stack.aclose()
-                except Exception:
-                    pass
-                self._mcp_stack = None
+            self._mcp_stacks.clear()
         finally:
             self._mcp_connecting = False
 
     def _set_tool_context(self, channel: str, chat_id: str, message_id: str | None = None) -> None:
         """Update context for all tools that need routing info."""
-        for name in ("message", "spawn", "cron"):
+        for name in ("message", "spawn", "cron", "my"):
             if tool := self.tools.get(name):
                 if hasattr(tool, "set_context"):
                     tool.set_context(channel, chat_id, *([message_id] if name == "message" else []))
@@ -332,6 +331,7 @@ class AgentLoop:
         if not text:
             return None
         from nanobot.utils.helpers import strip_think
+
         return strip_think(text) or None
 
     @staticmethod
@@ -341,6 +341,12 @@ class AgentLoop:
 
         return format_tool_hints(tool_calls)
 
+    def _effective_session_key(self, msg: InboundMessage) -> str:
+        """Return the session key used for task routing and mid-turn injections."""
+        if self._unified_session and not msg.session_key_override:
+            return UNIFIED_SESSION_KEY
+        return msg.session_key
+
     async def _run_agent_loop(
         self,
         initial_messages: list[dict],
@@ -373,9 +379,7 @@ class AgentLoop:
             message_id=message_id,
         )
         hook: AgentHook = (
-            _LoopHookChain(loop_hook, self._extra_hooks)
-            if self._extra_hooks
-            else loop_hook
+            CompositeHook([loop_hook] + self._extra_hooks) if self._extra_hooks else loop_hook
         )
 
         async def _checkpoint(payload: dict[str, Any]) -> None:
@@ -383,16 +387,32 @@ class AgentLoop:
                 return
             self._set_runtime_checkpoint(session, payload)
 
-        async def _drain_pending() -> list[InboundMessage]:
+        async def _drain_pending(*, limit: int = _MAX_INJECTIONS_PER_TURN) -> list[dict[str, Any]]:
             """Non-blocking drain of follow-up messages from the pending queue."""
             if pending_queue is None:
                 return []
-            items: list[InboundMessage] = []
-            while True:
+            items: list[dict[str, Any]] = []
+            while len(items) < limit:
                 try:
-                    items.append(pending_queue.get_nowait())
+                    pending_msg = pending_queue.get_nowait()
                 except asyncio.QueueEmpty:
                     break
+                content = pending_msg.content
+                media = pending_msg.media if pending_msg.media else None
+                if media:
+                    content, media = extract_documents(content, media)
+                    media = media or None
+                user_content = self.context._build_user_content(content, media)
+                runtime_ctx = self.context._build_runtime_context(
+                    pending_msg.channel,
+                    pending_msg.chat_id,
+                    self.context.timezone,
+                )
+                if isinstance(user_content, str):
+                    merged: str | list[dict[str, Any]] = f"{runtime_ctx}\n\n{user_content}"
+                else:
+                    merged = [{"type": "text", "text": runtime_ctx}] + user_content
+                items.append({"role": "user", "content": merged})
             return items
 
         result = await self.runner.run(AgentRunSpec(
@@ -416,6 +436,11 @@ class AgentLoop:
         self._last_usage = result.usage
         if result.stop_reason == "max_iterations":
             logger.warning("Max iterations ({}) reached", self.max_iterations)
+            # Push final content through stream so streaming channels (e.g. Feishu)
+            # update the card instead of leaving it empty.
+            if on_stream and on_stream_end:
+                await on_stream(result.final_content or "")
+                await on_stream_end(resuming=False)
         elif result.stop_reason == "error":
             logger.error("LLM returned error: {}", (result.final_content or "")[:200])
         return result.final_content, result.tools_used, result.messages, result.stop_reason, result.had_injections
@@ -430,7 +455,10 @@ class AgentLoop:
             try:
                 msg = await asyncio.wait_for(self.bus.consume_inbound(), timeout=1.0)
             except asyncio.TimeoutError:
-                self.auto_compact.check_expired(self._schedule_background)
+                self.auto_compact.check_expired(
+                    self._schedule_background,
+                    active_session_keys=self._pending_queues.keys(),
+                )
                 continue
             except asyncio.CancelledError:
                 # Preserve real task cancellation so shutdown can complete cleanly.
@@ -449,30 +477,46 @@ class AgentLoop:
                 if result:
                     await self.bus.publish_outbound(result)
                 continue
+            effective_key = self._effective_session_key(msg)
             # If this session already has an active pending queue (i.e. a task
             # is processing this session), route the message there for mid-turn
             # injection instead of creating a competing task.
-            if msg.session_key in self._pending_queues:
+            if effective_key in self._pending_queues:
+                pending_msg = msg
+                if effective_key != msg.session_key:
+                    pending_msg = dataclasses.replace(
+                        msg,
+                        session_key_override=effective_key,
+                    )
                 try:
-                    self._pending_queues[msg.session_key].put_nowait(msg)
+                    self._pending_queues[effective_key].put_nowait(pending_msg)
                 except asyncio.QueueFull:
                     logger.warning(
-                        "Pending queue full for session {}, dropping follow-up",
-                        msg.session_key,
+                        "Pending queue full for session {}, falling back to queued task",
+                        effective_key,
                     )
                 else:
                     logger.info(
                         "Routed follow-up message to pending queue for session {}",
-                        msg.session_key,
+                        effective_key,
                     )
-                continue
+                    continue
+            # Compute the effective session key before dispatching
+            # This ensures /stop command can find tasks correctly when unified session is enabled
             task = asyncio.create_task(self._dispatch(msg))
-            self._active_tasks.setdefault(msg.session_key, []).append(task)
-            task.add_done_callback(lambda t, k=msg.session_key: self._active_tasks.get(k, []) and self._active_tasks[k].remove(t) if t in self._active_tasks.get(k, []) else None)
+            self._active_tasks.setdefault(effective_key, []).append(task)
+            task.add_done_callback(
+                lambda t, k=effective_key: self._active_tasks.get(k, [])
+                and self._active_tasks[k].remove(t)
+                if t in self._active_tasks.get(k, [])
+                else None
+            )
 
     async def _dispatch(self, msg: InboundMessage) -> None:
         """Process a message: per-session serial, cross-session concurrent."""
-        session_key = msg.session_key
+        session_key = self._effective_session_key(msg)
+        if session_key != msg.session_key:
+            msg = dataclasses.replace(msg, session_key_override=session_key)
         lock = self._session_locks.setdefault(session_key, asyncio.Lock())
         gate = self._concurrency_gate or nullcontext()
 
@@ -561,12 +605,12 @@ class AgentLoop:
         if self._background_tasks:
             await asyncio.gather(*self._background_tasks, return_exceptions=True)
             self._background_tasks.clear()
-        if self._mcp_stack:
+        for name, stack in self._mcp_stacks.items():
             try:
-                await self._mcp_stack.aclose()
+                await stack.aclose()
             except (RuntimeError, BaseExceptionGroup):
-                pass  # MCP SDK cancel scope cleanup is noisy but harmless
-            self._mcp_stack = None
+                logger.debug("MCP server '{}' cleanup error (can be ignored)", name)
+        self._mcp_stacks.clear()
 
     def _schedule_background(self, coro) -> None:
         """Schedule a coroutine as a tracked background task (drained on shutdown)."""
@@ -591,13 +635,16 @@ class AgentLoop:
         """Process a single inbound message and return the response."""
         # System messages: parse origin from chat_id ("channel:chat_id")
         if msg.channel == "system":
-            channel, chat_id = (msg.chat_id.split(":", 1) if ":" in msg.chat_id
-                                else ("cli", msg.chat_id))
+            channel, chat_id = (
+                msg.chat_id.split(":", 1) if ":" in msg.chat_id else ("cli", msg.chat_id)
+            )
             logger.info("Processing system message from {}", msg.sender_id)
             key = f"{channel}:{chat_id}"
             session = self.sessions.get_or_create(key)
             if self._restore_runtime_checkpoint(session):
                 self.sessions.save(session)
+            if self._restore_pending_user_turn(session):
+                self.sessions.save(session)
 
             session, pending = self.auto_compact.prepare_session(session, key)
 
@@ -620,8 +667,17 @@ class AgentLoop:
             self._clear_runtime_checkpoint(session)
             self.sessions.save(session)
             self._schedule_background(self.consolidator.maybe_consolidate_by_tokens(session))
-            return OutboundMessage(channel=channel, chat_id=chat_id,
-                                  content=final_content or "Background task completed.")
+            return OutboundMessage(
+                channel=channel,
+                chat_id=chat_id,
+                content=final_content or "Background task completed.",
+            )
+
+        # Extract document text from media at the processing boundary so all
+        # channels benefit without format-specific logic in ContextBuilder.
+        if msg.media:
+            new_content, image_only = extract_documents(msg.content, msg.media)
+            msg = dataclasses.replace(msg, content=new_content, media=image_only)
 
         preview = msg.content[:80] + "..." if len(msg.content) > 80 else msg.content
         logger.info("Processing message from {}:{}: {}", msg.channel, msg.sender_id, preview)
@@ -630,6 +686,8 @@ class AgentLoop:
         session = self.sessions.get_or_create(key)
         if self._restore_runtime_checkpoint(session):
             self.sessions.save(session)
+        if self._restore_pending_user_turn(session):
+            self.sessions.save(session)
 
         session, pending = self.auto_compact.prepare_session(session, key)
 
@@ -653,16 +711,35 @@ class AgentLoop:
             current_message=msg.content,
             session_summary=pending,
             media=msg.media if msg.media else None,
-            channel=msg.channel, chat_id=msg.chat_id,
+            channel=msg.channel,
+            chat_id=msg.chat_id,
         )
 
         async def _bus_progress(content: str, *, tool_hint: bool = False) -> None:
             meta = dict(msg.metadata or {})
             meta["_progress"] = True
             meta["_tool_hint"] = tool_hint
-            await self.bus.publish_outbound(OutboundMessage(
-                channel=msg.channel, chat_id=msg.chat_id, content=content, metadata=meta,
-            ))
+            await self.bus.publish_outbound(
+                OutboundMessage(
+                    channel=msg.channel,
+                    chat_id=msg.chat_id,
+                    content=content,
+                    metadata=meta,
+                )
+            )
+
+        # Persist the triggering user message immediately, before running the
+        # agent loop. If the process is killed mid-turn (OOM, SIGKILL, self-
+        # restart, etc.), the existing runtime_checkpoint preserves the
+        # in-flight assistant/tool state but NOT the user message itself, so
+        # the user's prompt is silently lost on recovery. Saving it up front
+        # makes recovery possible from the session log alone.
+        user_persisted_early = False
+        if isinstance(msg.content, str) and msg.content.strip():
+            session.add_message("user", msg.content)
+            self._mark_pending_user_turn(session)
+            self.sessions.save(session)
+            user_persisted_early = True
 
         final_content, _, all_msgs, stop_reason, had_injections = await self._run_agent_loop(
             initial_messages,
@@ -670,7 +747,8 @@ class AgentLoop:
             on_stream=on_stream,
             on_stream_end=on_stream_end,
             session=session,
-            channel=msg.channel, chat_id=msg.chat_id,
+            channel=msg.channel,
+            chat_id=msg.chat_id,
             message_id=msg.metadata.get("message_id"),
             pending_queue=pending_queue,
         )
@@ -678,27 +756,34 @@ class AgentLoop:
         if final_content is None or not final_content.strip():
             final_content = EMPTY_FINAL_RESPONSE_MESSAGE
 
-        self._save_turn(session, all_msgs, 1 + len(history))
+        # Skip the already-persisted user message when saving the turn
+        save_skip = 1 + len(history) + (1 if user_persisted_early else 0)
+        self._save_turn(session, all_msgs, save_skip)
+        self._clear_pending_user_turn(session)
         self._clear_runtime_checkpoint(session)
         self.sessions.save(session)
         self._schedule_background(self.consolidator.maybe_consolidate_by_tokens(session))
 
-        # When follow-up messages were injected mid-turn, the LLM's final
-        # response addresses those follow-ups.  Always send the response in
-        # this case, even if MessageTool was used earlier in the turn — the
-        # follow-up response is new content the user hasn't seen.
-        if not had_injections:
-            if (mt := self.tools.get("message")) and isinstance(mt, MessageTool) and mt._sent_in_turn:
+        # When follow-up messages were injected mid-turn, a later natural
+        # language reply may address those follow-ups and should not be
+        # suppressed just because MessageTool was used earlier in the turn.
+        # However, if the turn falls back to the empty-final-response
+        # placeholder, suppress it when the real user-visible output already
+        # came from MessageTool.
+        if (mt := self.tools.get("message")) and isinstance(mt, MessageTool) and mt._sent_in_turn:
+            if not had_injections or stop_reason == "empty_final_response":
                 return None
 
         preview = final_content[:120] + "..." if len(final_content) > 120 else final_content
         logger.info("Response to {}:{}: {}", msg.channel, msg.sender_id, preview)
 
         meta = dict(msg.metadata or {})
-        if on_stream is not None:
+        if on_stream is not None and stop_reason != "error":
             meta["_streamed"] = True
         return OutboundMessage(
-            channel=msg.channel, chat_id=msg.chat_id, content=final_content,
+            channel=msg.channel,
+            chat_id=msg.chat_id,
+            content=final_content,
             metadata=meta,
         )
 
@@ -706,7 +791,7 @@ class AgentLoop:
         self,
         content: list[dict[str, Any]],
         *,
-        truncate_text: bool = False,
+        should_truncate_text: bool = False,
         drop_runtime: bool = False,
     ) -> list[dict[str, Any]]:
         """Strip volatile multimodal payloads before writing session history."""
@@ -724,18 +809,17 @@ class AgentLoop:
             ):
                 continue
 
-            if (
-                block.get("type") == "image_url"
-                and block.get("image_url", {}).get("url", "").startswith("data:image/")
-            ):
+            if block.get("type") == "image_url" and block.get("image_url", {}).get(
+                "url", ""
+            ).startswith("data:image/"):
                 path = (block.get("_meta") or {}).get("path", "")
                 filtered.append({"type": "text", "text": image_placeholder_text(path)})
                 continue
 
             if block.get("type") == "text" and isinstance(block.get("text"), str):
                 text = block["text"]
-                if truncate_text and len(text) > self.max_tool_result_chars:
-                    text = truncate_text(text, self.max_tool_result_chars)
+                if should_truncate_text and len(text) > self.max_tool_result_chars:
+                    text = truncate_text_fn(text, self.max_tool_result_chars)
                 filtered.append({**block, "text": text})
                 continue
 
@@ -746,6 +830,7 @@ class AgentLoop:
     def _save_turn(self, session: Session, messages: list[dict], skip: int) -> None:
         """Save new-turn messages into session, truncating large tool results."""
         from datetime import datetime
+
         for m in messages[skip:]:
             entry = dict(m)
             role, content = entry.get("role"), entry.get("content")
@@ -753,9 +838,9 @@ class AgentLoop:
                 continue  # skip empty assistant messages — they poison session context
             if role == "tool":
                 if isinstance(content, str) and len(content) > self.max_tool_result_chars:
-                    entry["content"] = truncate_text(content, self.max_tool_result_chars)
+                    entry["content"] = truncate_text_fn(content, self.max_tool_result_chars)
                 elif isinstance(content, list):
-                    filtered = self._sanitize_persisted_blocks(content, truncate_text=True)
+                    filtered = self._sanitize_persisted_blocks(content, should_truncate_text=True)
                     if not filtered:
                         continue
                     entry["content"] = filtered
@@ -792,6 +877,12 @@ class AgentLoop:
         session.metadata[self._RUNTIME_CHECKPOINT_KEY] = payload
         self.sessions.save(session)
 
+    def _mark_pending_user_turn(self, session: Session) -> None:
+        session.metadata[self._PENDING_USER_TURN_KEY] = True
+
+    def _clear_pending_user_turn(self, session: Session) -> None:
+        session.metadata.pop(self._PENDING_USER_TURN_KEY, None)
+
     def _clear_runtime_checkpoint(self, session: Session) -> None:
         if self._RUNTIME_CHECKPOINT_KEY in session.metadata:
             session.metadata.pop(self._RUNTIME_CHECKPOINT_KEY, None)
@@ -835,13 +926,15 @@ class AgentLoop:
                 continue
             tool_id = tool_call.get("id")
             name = ((tool_call.get("function") or {}).get("name")) or "tool"
-            restored_messages.append({
-                "role": "tool",
-                "tool_call_id": tool_id,
-                "name": name,
-                "content": "Error: Task interrupted before this tool finished.",
-                "timestamp": datetime.now().isoformat(),
-            })
+            restored_messages.append(
+                {
+                    "role": "tool",
+                    "tool_call_id": tool_id,
+                    "name": name,
+                    "content": "Error: Task interrupted before this tool finished.",
+                    "timestamp": datetime.now().isoformat(),
+                }
+            )
 
         overlap = 0
         max_overlap = min(len(session.messages), len(restored_messages))
@@ -856,23 +949,51 @@ class AgentLoop:
                 break
         session.messages.extend(restored_messages[overlap:])
 
+        self._clear_pending_user_turn(session)
         self._clear_runtime_checkpoint(session)
         return True
 
+    def _restore_pending_user_turn(self, session: Session) -> bool:
+        """Close a turn that only persisted the user message before crashing."""
+        from datetime import datetime
+
+        if not session.metadata.get(self._PENDING_USER_TURN_KEY):
+            return False
+
+        if session.messages and session.messages[-1].get("role") == "user":
+            session.messages.append(
+                {
+                    "role": "assistant",
+                    "content": "Error: Task interrupted before a response was generated.",
+                    "timestamp": datetime.now().isoformat(),
+                }
+            )
+            session.updated_at = datetime.now()
+
+        self._clear_pending_user_turn(session)
+        return True
+
     async def process_direct(
         self,
         content: str,
         session_key: str = "cli:direct",
         channel: str = "cli",
         chat_id: str = "direct",
+        media: list[str] | None = None,
         on_progress: Callable[[str], Awaitable[None]] | None = None,
         on_stream: Callable[[str], Awaitable[None]] | None = None,
         on_stream_end: Callable[..., Awaitable[None]] | None = None,
     ) -> OutboundMessage | None:
         """Process a message directly and return the outbound payload."""
         await self._connect_mcp()
-        msg = InboundMessage(channel=channel, sender_id="user", chat_id=chat_id, content=content)
+        msg = InboundMessage(
+            channel=channel, sender_id="user", chat_id=chat_id,
+            content=content, media=media or [],
+        )
         return await self._process_message(
-            msg, session_key=session_key, on_progress=on_progress,
-            on_stream=on_stream, on_stream_end=on_stream_end,
+            msg,
+            session_key=session_key,
+            on_progress=on_progress,
+            on_stream=on_stream,
+            on_stream_end=on_stream_end,
         )

nanobot/agent/memory.py

@@ -239,13 +239,13 @@ class MemoryStore:
                 pass
         # Fallback: read last line's cursor from the JSONL file.
         last = self._read_last_entry()
-        if last:
+        if last and last.get("cursor"):
             return last["cursor"] + 1
         return 1
 
     def read_unprocessed_history(self, since_cursor: int) -> list[dict[str, Any]]:
         """Return history entries with cursor > *since_cursor*."""
-        return [e for e in self._read_entries() if e["cursor"] > since_cursor]
+        return [e for e in self._read_entries() if e.get("cursor", 0) > since_cursor]
 
     def compact_history(self) -> None:
         """Drop oldest entries if the file exceeds *max_history_entries*."""
@@ -290,7 +290,7 @@ class MemoryStore:
                 if not lines:
                     return None
                 return json.loads(lines[-1])
-        except (FileNotFoundError, json.JSONDecodeError):
+        except (FileNotFoundError, json.JSONDecodeError, UnicodeDecodeError):
             return None
 
     def _write_entries(self, entries: list[dict[str, Any]]) -> None:
@@ -347,6 +347,7 @@ class Consolidator:
     """Lightweight consolidation: summarizes evicted messages into history.jsonl."""
 
     _MAX_CONSOLIDATION_ROUNDS = 5
+    _MAX_CHUNK_MESSAGES = 60  # hard cap per consolidation round
 
     _SAFETY_BUFFER = 1024  # extra headroom for tokenizer estimation drift
 
@@ -360,6 +361,7 @@ class Consolidator:
         build_messages: Callable[..., list[dict[str, Any]]],
         get_tool_definitions: Callable[[], list[dict[str, Any]]],
         max_completion_tokens: int = 4096,
+        consolidation_ratio: float = 0.5,
     ):
         self.store = store
         self.provider = provider
@@ -367,16 +369,13 @@ class Consolidator:
         self.sessions = sessions
         self.context_window_tokens = context_window_tokens
         self.max_completion_tokens = max_completion_tokens
+        self.consolidation_ratio = consolidation_ratio
         self._build_messages = build_messages
         self._get_tool_definitions = get_tool_definitions
         self._locks: weakref.WeakValueDictionary[str, asyncio.Lock] = (
             weakref.WeakValueDictionary()
         )
 
-    def get_last_history_entry(self) -> dict[str, Any] | None:
-        """Return the most recent entry from history.jsonl."""
-        return self.store._read_last_entry()
-
     def get_lock(self, session_key: str) -> asyncio.Lock:
         """Return the shared consolidation lock for one session."""
         return self._locks.setdefault(session_key, asyncio.Lock())
@@ -403,6 +402,22 @@ class Consolidator:
 
         return last_boundary
 
+    def _cap_consolidation_boundary(
+        self,
+        session: Session,
+        end_idx: int,
+    ) -> int | None:
+        """Clamp the chunk size without breaking the user-turn boundary."""
+        start = session.last_consolidated
+        if end_idx - start <= self._MAX_CHUNK_MESSAGES:
+            return end_idx
+
+        capped_end = start + self._MAX_CHUNK_MESSAGES
+        for idx in range(capped_end, start, -1):
+            if session.messages[idx].get("role") == "user":
+                return idx
+        return None
+
     def estimate_session_prompt_tokens(self, session: Session) -> tuple[int, str]:
         """Estimate current prompt size for the normal session history view."""
         history = session.get_history(max_messages=0)
@@ -420,13 +435,13 @@ class Consolidator:
             self._get_tool_definitions(),
         )
 
-    async def archive(self, messages: list[dict]) -> bool:
+    async def archive(self, messages: list[dict]) -> str | None:
         """Summarize messages via LLM and append to history.jsonl.
 
-        Returns True on success (or degraded success), False if nothing to do.
+        Returns the summary text on success, None if nothing to archive.
         """
         if not messages:
-            return False
+            return None
         try:
             formatted = MemoryStore._format_messages(messages)
             response = await self.provider.chat_with_retry(
@@ -444,13 +459,15 @@ class Consolidator:
                 tools=None,
                 tool_choice=None,
             )
+            if response.finish_reason == "error":
+                raise RuntimeError(f"LLM returned error: {response.content}")
             summary = response.content or "[no summary]"
             self.store.append_history(summary)
-            return True
+            return summary
         except Exception:
             logger.warning("Consolidation LLM call failed, raw-dumping to history")
             self.store.raw_archive(messages)
-            return True
+            return None
 
     async def maybe_consolidate_by_tokens(self, session: Session) -> None:
         """Loop: archive old messages until prompt fits within safe budget.
@@ -464,17 +481,23 @@ class Consolidator:
         lock = self.get_lock(session.key)
         async with lock:
             budget = self.context_window_tokens - self.max_completion_tokens - self._SAFETY_BUFFER
-            target = budget // 2
-            estimated, source = self.estimate_session_prompt_tokens(session)
+            target = int(budget * self.consolidation_ratio)
+            try:
+                estimated, source = self.estimate_session_prompt_tokens(session)
+            except Exception:
+                logger.exception("Token estimation failed for {}", session.key)
+                estimated, source = 0, "error"
             if estimated <= 0:
                 return
             if estimated < budget:
+                unconsolidated_count = len(session.messages) - session.last_consolidated
                 logger.debug(
-                    "Token consolidation idle {}: {}/{} via {}",
+                    "Token consolidation idle {}: {}/{} via {}, msgs={}",
                     session.key,
                     estimated,
                     self.context_window_tokens,
                     source,
+                    unconsolidated_count,
                 )
                 return
 
@@ -492,6 +515,15 @@ class Consolidator:
                     return
 
                 end_idx = boundary[0]
+                end_idx = self._cap_consolidation_boundary(session, end_idx)
+                if end_idx is None:
+                    logger.debug(
+                        "Token consolidation: no capped boundary for {} (round {})",
+                        session.key,
+                        round_num,
+                    )
+                    return
+
                 chunk = session.messages[session.last_consolidated:end_idx]
                 if not chunk:
                     return
@@ -510,7 +542,11 @@ class Consolidator:
                 session.last_consolidated = end_idx
                 self.sessions.save(session)
 
-                estimated, source = self.estimate_session_prompt_tokens(session)
+                try:
+                    estimated, source = self.estimate_session_prompt_tokens(session)
+                except Exception:
+                    logger.exception("Token estimation failed for {}", session.key)
+                    estimated, source = 0, "error"
                 if estimated <= 0:
                     return
 
@@ -520,6 +556,13 @@ class Consolidator:
 # ---------------------------------------------------------------------------
 
 
+# Single source of truth for the staleness threshold used in _annotate_with_ages
+# *and* in the Phase 1 prompt template (passed as `stale_threshold_days`).
+# Keep code and prompt aligned — if you bump this, the LLM's instruction string
+# updates automatically.
+_STALE_THRESHOLD_DAYS = 14
+
+
 class Dream:
     """Two-phase memory processor: analyze history.jsonl, then edit files via AgentRunner.
 
@@ -536,6 +579,7 @@ class Dream:
         max_batch_size: int = 20,
         max_iterations: int = 10,
         max_tool_result_chars: int = 16_000,
+        annotate_line_ages: bool = True,
     ):
         self.store = store
         self.provider = provider
@@ -543,6 +587,10 @@ class Dream:
         self.max_batch_size = max_batch_size
         self.max_iterations = max_iterations
         self.max_tool_result_chars = max_tool_result_chars
+        # Kill switch for the git-blame-based per-line age annotation in Phase 1.
+        # Default True keeps the #3212 behavior; set False to feed MEMORY.md raw
+        # (e.g. if a specific LLM reacts poorly to the `← Nd` suffix).
+        self.annotate_line_ages = annotate_line_ages
         self._runner = AgentRunner(provider)
         self._tools = self._build_tools()
 
@@ -550,18 +598,106 @@ class Dream:
 
     def _build_tools(self) -> ToolRegistry:
         """Build a minimal tool registry for the Dream agent."""
-        from nanobot.agent.tools.filesystem import EditFileTool, ReadFileTool
+        from nanobot.agent.skills import BUILTIN_SKILLS_DIR
+        from nanobot.agent.tools.filesystem import EditFileTool, ReadFileTool, WriteFileTool
 
         tools = ToolRegistry()
         workspace = self.store.workspace
-        tools.register(ReadFileTool(workspace=workspace, allowed_dir=workspace))
+        # Allow reading builtin skills for reference during skill creation
+        extra_read = [BUILTIN_SKILLS_DIR] if BUILTIN_SKILLS_DIR.exists() else None
+        tools.register(ReadFileTool(
+            workspace=workspace,
+            allowed_dir=workspace,
+            extra_allowed_dirs=extra_read,
+        ))
         tools.register(EditFileTool(workspace=workspace, allowed_dir=workspace))
+        # write_file resolves relative paths from workspace root, but can only
+        # write under skills/ so the prompt can safely use skills/<name>/SKILL.md.
+        skills_dir = workspace / "skills"
+        skills_dir.mkdir(parents=True, exist_ok=True)
+        tools.register(WriteFileTool(workspace=workspace, allowed_dir=skills_dir))
         return tools
 
+    # -- skill listing --------------------------------------------------------
+
+    def _list_existing_skills(self) -> list[str]:
+        """List existing skills as 'name — description' for dedup context."""
+        import re as _re
+
+        from nanobot.agent.skills import BUILTIN_SKILLS_DIR
+
+        _DESC_RE = _re.compile(r"^description:\s*(.+)$", _re.MULTILINE | _re.IGNORECASE)
+        entries: dict[str, str] = {}
+        for base in (self.store.workspace / "skills", BUILTIN_SKILLS_DIR):
+            if not base.exists():
+                continue
+            for d in base.iterdir():
+                if not d.is_dir():
+                    continue
+                skill_md = d / "SKILL.md"
+                if not skill_md.exists():
+                    continue
+                # Prefer workspace skills over builtin (same name)
+                if d.name in entries and base == BUILTIN_SKILLS_DIR:
+                    continue
+                content = skill_md.read_text(encoding="utf-8")[:500]
+                m = _DESC_RE.search(content)
+                desc = m.group(1).strip() if m else "(no description)"
+                entries[d.name] = desc
+        return [f"{name} — {desc}" for name, desc in sorted(entries.items())]
+
     # -- main entry ----------------------------------------------------------
 
+    def _annotate_with_ages(self, content: str) -> str:
+        """Append per-line age suffixes to MEMORY.md content.
+
+        Each non-blank line whose age exceeds ``_STALE_THRESHOLD_DAYS`` gets a
+        suffix like ``← 30d`` indicating days since last modification.
+        Returns the original content unchanged if git is unavailable,
+        annotate fails, or the line count doesn't match the age count
+        (which can happen with an uncommitted working-tree edit — better to
+        skip annotation than to tag the wrong line).
+        SOUL.md and USER.md are never annotated.
+        """
+        file_path = "memory/MEMORY.md"
+        try:
+            ages = self.store.git.line_ages(file_path)
+        except Exception:
+            logger.debug("line_ages failed for {}", file_path)
+            return content
+        if not ages:
+            return content
+
+        had_trailing = content.endswith("\n")
+        lines = content.splitlines()
+        # If HEAD-blob line count disagrees with the working-tree content we
+        # received, ages would be assigned to the wrong lines — skip entirely
+        # and feed the LLM un-annotated content rather than misleading data.
+        if len(lines) != len(ages):
+            logger.debug(
+                "line_ages length mismatch for {} (lines={}, ages={}); skipping annotation",
+                file_path, len(lines), len(ages),
+            )
+            return content
+
+        annotated: list[str] = []
+        for line, age in zip(lines, ages):
+            if not line.strip():
+                annotated.append(line)
+                continue
+            if age.age_days > _STALE_THRESHOLD_DAYS:
+                annotated.append(f"{line}  \u2190 {age.age_days}d")
+            else:
+                annotated.append(line)
+        result = "\n".join(annotated)
+        if had_trailing:
+            result += "\n"
+        return result
+
     async def run(self) -> bool:
         """Process unprocessed history entries. Returns True if work was done."""
+        from nanobot.agent.skills import BUILTIN_SKILLS_DIR
+
         last_cursor = self.store.get_last_dream_cursor()
         entries = self.store.read_unprocessed_history(since_cursor=last_cursor)
         if not entries:
@@ -578,11 +714,17 @@ class Dream:
             f"[{e['timestamp']}] {e['content']}" for e in batch
         )
 
-        # Current file contents
+        # Current file contents + per-line age annotations (MEMORY.md only)
         current_date = datetime.now().strftime("%Y-%m-%d")
-        current_memory = self.store.read_memory() or "(empty)"
+        raw_memory = self.store.read_memory() or "(empty)"
+        current_memory = (
+            self._annotate_with_ages(raw_memory)
+            if self.annotate_line_ages
+            else raw_memory
+        )
         current_soul = self.store.read_soul() or "(empty)"
         current_user = self.store.read_user() or "(empty)"
+
         file_context = (
             f"## Current Date\n{current_date}\n\n"
             f"## Current MEMORY.md ({len(current_memory)} chars)\n{current_memory}\n\n"
@@ -590,7 +732,7 @@ class Dream:
             f"## Current USER.md ({len(current_user)} chars)\n{current_user}"
         )
 
-        # Phase 1: Analyze
+        # Phase 1: Analyze (no skills list — dedup is Phase 2's job)
         phase1_prompt = (
             f"## Conversation History\n{history_text}\n\n{file_context}"
         )
@@ -601,7 +743,11 @@ class Dream:
                 messages=[
                     {
                         "role": "system",
-                        "content": render_template("agent/dream_phase1.md", strip=True),
+                        "content": render_template(
+                            "agent/dream_phase1.md",
+                            strip=True,
+                            stale_threshold_days=_STALE_THRESHOLD_DAYS,
+                        ),
                     },
                     {"role": "user", "content": phase1_prompt},
                 ],
@@ -615,13 +761,25 @@ class Dream:
             return False
 
         # Phase 2: Delegate to AgentRunner with read_file / edit_file
-        phase2_prompt = f"## Analysis Result\n{analysis}\n\n{file_context}"
+        existing_skills = self._list_existing_skills()
+        skills_section = ""
+        if existing_skills:
+            skills_section = (
+                "\n\n## Existing Skills\n"
+                + "\n".join(f"- {s}" for s in existing_skills)
+            )
+        phase2_prompt = f"## Analysis Result\n{analysis}\n\n{file_context}{skills_section}"
 
         tools = self._tools
+        skill_creator_path = BUILTIN_SKILLS_DIR / "skill-creator" / "SKILL.md"
         messages: list[dict[str, Any]] = [
             {
                 "role": "system",
-                "content": render_template("agent/dream_phase2.md", strip=True),
+                "content": render_template(
+                    "agent/dream_phase2.md",
+                    strip=True,
+                    skill_creator_path=str(skill_creator_path),
+                ),
             },
             {"role": "user", "content": phase2_prompt},
         ]
@@ -672,7 +830,9 @@ class Dream:
         # Git auto-commit (only when there are actual changes)
         if changelog and self.store.git.is_initialized():
             ts = batch[-1]["timestamp"]
-            sha = self.store.git.auto_commit(f"dream: {ts}, {len(changelog)} change(s)")
+            summary = f"dream: {ts}, {len(changelog)} change(s)"
+            commit_msg = f"{summary}\n\n{analysis.strip()}"
+            sha = self.store.git.auto_commit(commit_msg)
             if sha:
                 logger.info("Dream commit: {}", sha)
 

nanobot/agent/runner.py

@@ -4,6 +4,7 @@ from __future__ import annotations
 
 import asyncio
 from dataclasses import dataclass, field
+import inspect
 from pathlib import Path
 from typing import Any
 
@@ -24,16 +25,27 @@ from nanobot.utils.helpers import (
 from nanobot.utils.runtime import (
     EMPTY_FINAL_RESPONSE_MESSAGE,
     build_finalization_retry_message,
+    build_length_recovery_message,
     ensure_nonempty_tool_result,
     is_blank_text,
     repeated_external_lookup_error,
 )
 
 _DEFAULT_ERROR_MESSAGE = "Sorry, I encountered an error calling the AI model."
+_PERSISTED_MODEL_ERROR_PLACEHOLDER = "[Assistant reply unavailable due to model error.]"
 _MAX_EMPTY_RETRIES = 2
+_MAX_LENGTH_RECOVERIES = 3
 _MAX_INJECTIONS_PER_TURN = 3
 _MAX_INJECTION_CYCLES = 5
 _SNIP_SAFETY_BUFFER = 1024
+_MICROCOMPACT_KEEP_RECENT = 10
+_MICROCOMPACT_MIN_CHARS = 500
+_COMPACTABLE_TOOLS = frozenset({
+    "read_file", "exec", "grep", "glob",
+    "web_search", "web_fetch", "list_dir",
+})
+_BACKFILL_CONTENT = "[Tool result unavailable — call was interrupted or lost]"
+
 
 
 @dataclass(slots=True)
@@ -83,37 +95,133 @@ class AgentRunner:
     def __init__(self, provider: LLMProvider):
         self.provider = provider
 
-    async def _drain_injections(self, spec: AgentRunSpec) -> list[str]:
+    @staticmethod
+    def _merge_message_content(left: Any, right: Any) -> str | list[dict[str, Any]]:
+        if isinstance(left, str) and isinstance(right, str):
+            return f"{left}\n\n{right}" if left else right
+
+        def _to_blocks(value: Any) -> list[dict[str, Any]]:
+            if isinstance(value, list):
+                return [
+                    item if isinstance(item, dict) else {"type": "text", "text": str(item)}
+                    for item in value
+                ]
+            if value is None:
+                return []
+            return [{"type": "text", "text": str(value)}]
+
+        return _to_blocks(left) + _to_blocks(right)
+
+    @classmethod
+    def _append_injected_messages(
+        cls,
+        messages: list[dict[str, Any]],
+        injections: list[dict[str, Any]],
+    ) -> None:
+        """Append injected user messages while preserving role alternation."""
+        for injection in injections:
+            if (
+                messages
+                and injection.get("role") == "user"
+                and messages[-1].get("role") == "user"
+            ):
+                merged = dict(messages[-1])
+                merged["content"] = cls._merge_message_content(
+                    merged.get("content"),
+                    injection.get("content"),
+                )
+                messages[-1] = merged
+                continue
+            messages.append(injection)
+
+    async def _try_drain_injections(
+        self,
+        spec: AgentRunSpec,
+        messages: list[dict[str, Any]],
+        assistant_message: dict[str, Any] | None,
+        injection_cycles: int,
+        *,
+        phase: str = "after error",
+        iteration: int | None = None,
+    ) -> tuple[bool, int]:
+        """Drain pending injections. Returns (should_continue, updated_cycles).
+
+        If injections are found and we haven't exceeded _MAX_INJECTION_CYCLES,
+        append them to *messages* (and emit a checkpoint if *assistant_message*
+        and *iteration* are both provided) and return (True, cycles+1) so the
+        caller continues the iteration loop.  Otherwise return (False, cycles).
+        """
+        if injection_cycles >= _MAX_INJECTION_CYCLES:
+            return False, injection_cycles
+        injections = await self._drain_injections(spec)
+        if not injections:
+            return False, injection_cycles
+        injection_cycles += 1
+        if assistant_message is not None:
+            messages.append(assistant_message)
+            if iteration is not None:
+                await self._emit_checkpoint(
+                    spec,
+                    {
+                        "phase": "final_response",
+                        "iteration": iteration,
+                        "model": spec.model,
+                        "assistant_message": assistant_message,
+                        "completed_tool_results": [],
+                        "pending_tool_calls": [],
+                    },
+                )
+        self._append_injected_messages(messages, injections)
+        logger.info(
+            "Injected {} follow-up message(s) {} ({}/{})",
+            len(injections), phase, injection_cycles, _MAX_INJECTION_CYCLES,
+        )
+        return True, injection_cycles
+
+    async def _drain_injections(self, spec: AgentRunSpec) -> list[dict[str, Any]]:
         """Drain pending user messages via the injection callback.
 
-        Returns all drained message contents (capped by
+        Returns normalized user messages (capped by
         ``_MAX_INJECTIONS_PER_TURN``), or an empty list when there is
-        nothing to inject.  Messages beyond the cap are logged so they
+        nothing to inject. Messages beyond the cap are logged so they
         are not silently lost.
         """
         if spec.injection_callback is None:
             return []
         try:
-            items = await spec.injection_callback()
+            signature = inspect.signature(spec.injection_callback)
+            accepts_limit = (
+                "limit" in signature.parameters
+                or any(
+                    parameter.kind is inspect.Parameter.VAR_KEYWORD
+                    for parameter in signature.parameters.values()
+                )
+            )
+            if accepts_limit:
+                items = await spec.injection_callback(limit=_MAX_INJECTIONS_PER_TURN)
+            else:
+                items = await spec.injection_callback()
         except Exception:
             logger.exception("injection_callback failed")
             return []
         if not items:
             return []
-        # items are InboundMessage objects from _drain_pending
-        texts: list[str] = []
+        injected_messages: list[dict[str, Any]] = []
         for item in items:
+            if isinstance(item, dict) and item.get("role") == "user" and "content" in item:
+                injected_messages.append(item)
+                continue
             text = getattr(item, "content", str(item))
             if text.strip():
-                texts.append(text)
-        if len(texts) > _MAX_INJECTIONS_PER_TURN:
-            dropped = len(texts) - _MAX_INJECTIONS_PER_TURN
+                injected_messages.append({"role": "user", "content": text})
+        if len(injected_messages) > _MAX_INJECTIONS_PER_TURN:
+            dropped = len(injected_messages) - _MAX_INJECTIONS_PER_TURN
             logger.warning(
-                "Injection batch has {} messages, capping to {} ({} dropped)",
-                len(texts), _MAX_INJECTIONS_PER_TURN, dropped,
+                "Injection callback returned {} messages, capping to {} ({} dropped)",
+                len(injected_messages), _MAX_INJECTIONS_PER_TURN, dropped,
             )
-            texts = texts[-_MAX_INJECTIONS_PER_TURN:]
-        return texts
+            injected_messages = injected_messages[:_MAX_INJECTIONS_PER_TURN]
+        return injected_messages
 
     async def run(self, spec: AgentRunSpec) -> AgentRunResult:
         hook = spec.hook or AgentHook()
@@ -126,21 +234,36 @@ class AgentRunner:
         tool_events: list[dict[str, str]] = []
         external_lookup_counts: dict[str, int] = {}
         empty_content_retries = 0
+        length_recovery_count = 0
         had_injections = False
         injection_cycles = 0
 
         for iteration in range(spec.max_iterations):
             try:
-                messages = self._apply_tool_result_budget(spec, messages)
-                messages_for_model = self._snip_history(spec, messages)
+                # Keep the persisted conversation untouched. Context governance
+                # may repair or compact historical messages for the model, but
+                # those synthetic edits must not shift the append boundary used
+                # later when the caller saves only the new turn.
+                messages_for_model = self._drop_orphan_tool_results(messages)
+                messages_for_model = self._backfill_missing_tool_results(messages_for_model)
+                messages_for_model = self._microcompact(messages_for_model)
+                messages_for_model = self._apply_tool_result_budget(spec, messages_for_model)
+                messages_for_model = self._snip_history(spec, messages_for_model)
+                # Snipping may have created new orphans; clean them up.
+                messages_for_model = self._drop_orphan_tool_results(messages_for_model)
+                messages_for_model = self._backfill_missing_tool_results(messages_for_model)
             except Exception as exc:
                 logger.warning(
-                    "Context governance failed on turn {} for {}: {}; using raw messages",
+                    "Context governance failed on turn {} for {}: {}; applying minimal repair",
                     iteration,
                     spec.session_key or "default",
                     exc,
                 )
-                messages_for_model = messages
+                try:
+                    messages_for_model = self._drop_orphan_tool_results(messages)
+                    messages_for_model = self._backfill_missing_tool_results(messages_for_model)
+                except Exception:
+                    messages_for_model = messages
             context = AgentHookContext(iteration=iteration, messages=messages)
             await hook.before_iteration(context)
             response = await self._request_model(spec, messages_for_model, hook, context)
@@ -184,16 +307,6 @@ class AgentRunner:
                 tool_events.extend(new_events)
                 context.tool_results = list(results)
                 context.tool_events = list(new_events)
-                if fatal_error is not None:
-                    error = f"Error: {type(fatal_error).__name__}: {fatal_error}"
-                    final_content = error
-                    stop_reason = "tool_error"
-                    self._append_final_message(messages, final_content)
-                    context.final_content = final_content
-                    context.error = error
-                    context.stop_reason = stop_reason
-                    await hook.after_iteration(context)
-                    break
                 completed_tool_results: list[dict[str, Any]] = []
                 for tool_call, result in zip(response.tool_calls, results):
                     tool_message = {
@@ -209,6 +322,23 @@ class AgentRunner:
                     }
                     messages.append(tool_message)
                     completed_tool_results.append(tool_message)
+                if fatal_error is not None:
+                    error = f"Error: {type(fatal_error).__name__}: {fatal_error}"
+                    final_content = error
+                    stop_reason = "tool_error"
+                    self._append_final_message(messages, final_content)
+                    context.final_content = final_content
+                    context.error = error
+                    context.stop_reason = stop_reason
+                    await hook.after_iteration(context)
+                    should_continue, injection_cycles = await self._try_drain_injections(
+                        spec, messages, None, injection_cycles,
+                        phase="after tool error",
+                    )
+                    if should_continue:
+                        had_injections = True
+                        continue
+                    break
                 await self._emit_checkpoint(
                     spec,
                     {
@@ -221,18 +351,14 @@ class AgentRunner:
                     },
                 )
                 empty_content_retries = 0
+                length_recovery_count = 0
                 # Checkpoint 1: drain injections after tools, before next LLM call
-                if injection_cycles < _MAX_INJECTION_CYCLES:
-                    injections = await self._drain_injections(spec)
-                    if injections:
-                        had_injections = True
-                        injection_cycles += 1
-                        for text in injections:
-                            messages.append({"role": "user", "content": text})
-                        logger.info(
-                            "Injected {} follow-up message(s) after tool execution ({}/{})",
-                            len(injections), injection_cycles, _MAX_INJECTION_CYCLES,
-                        )
+                _drained, injection_cycles = await self._try_drain_injections(
+                    spec, messages, None, injection_cycles,
+                    phase="after tool execution",
+                )
+                if _drained:
+                    had_injections = True
                 await hook.after_iteration(context)
                 continue
 
@@ -268,27 +394,50 @@ class AgentRunner:
                 context.tool_calls = list(response.tool_calls)
                 clean = hook.finalize_content(context, response.content)
 
+            if response.finish_reason == "length" and not is_blank_text(clean):
+                length_recovery_count += 1
+                if length_recovery_count <= _MAX_LENGTH_RECOVERIES:
+                    logger.info(
+                        "Output truncated on turn {} for {} ({}/{}); continuing",
+                        iteration,
+                        spec.session_key or "default",
+                        length_recovery_count,
+                        _MAX_LENGTH_RECOVERIES,
+                    )
+                    if hook.wants_streaming():
+                        await hook.on_stream_end(context, resuming=True)
+                    messages.append(build_assistant_message(
+                        clean,
+                        reasoning_content=response.reasoning_content,
+                        thinking_blocks=response.thinking_blocks,
+                    ))
+                    messages.append(build_length_recovery_message())
+                    await hook.after_iteration(context)
+                    continue
+
+            assistant_message: dict[str, Any] | None = None
+            if response.finish_reason != "error" and not is_blank_text(clean):
+                assistant_message = build_assistant_message(
+                    clean,
+                    reasoning_content=response.reasoning_content,
+                    thinking_blocks=response.thinking_blocks,
+                )
+
             # Check for mid-turn injections BEFORE signaling stream end.
             # If injections are found we keep the stream alive (resuming=True)
             # so streaming channels don't prematurely finalize the card.
-            _injected_after_final = False
-            if injection_cycles < _MAX_INJECTION_CYCLES:
-                injections = await self._drain_injections(spec)
-                if injections:
-                    had_injections = True
-                    injection_cycles += 1
-                    _injected_after_final = True
-                    for text in injections:
-                        messages.append({"role": "user", "content": text})
-                    logger.info(
-                        "Injected {} follow-up message(s) after final response ({}/{})",
-                        len(injections), injection_cycles, _MAX_INJECTION_CYCLES,
-                    )
+            should_continue, injection_cycles = await self._try_drain_injections(
+                spec, messages, assistant_message, injection_cycles,
+                phase="after final response",
+                iteration=iteration,
+            )
+            if should_continue:
+                had_injections = True
 
             if hook.wants_streaming():
-                await hook.on_stream_end(context, resuming=_injected_after_final)
+                await hook.on_stream_end(context, resuming=should_continue)
 
-            if _injected_after_final:
+            if should_continue:
                 await hook.after_iteration(context)
                 continue
 
@@ -296,11 +445,18 @@ class AgentRunner:
                 final_content = clean or spec.error_message or _DEFAULT_ERROR_MESSAGE
                 stop_reason = "error"
                 error = final_content
-                self._append_final_message(messages, final_content)
+                self._append_model_error_placeholder(messages)
                 context.final_content = final_content
                 context.error = error
                 context.stop_reason = stop_reason
                 await hook.after_iteration(context)
+                should_continue, injection_cycles = await self._try_drain_injections(
+                    spec, messages, None, injection_cycles,
+                    phase="after LLM error",
+                )
+                if should_continue:
+                    had_injections = True
+                    continue
                 break
             if is_blank_text(clean):
                 final_content = EMPTY_FINAL_RESPONSE_MESSAGE
@@ -311,9 +467,16 @@ class AgentRunner:
                 context.error = error
                 context.stop_reason = stop_reason
                 await hook.after_iteration(context)
+                should_continue, injection_cycles = await self._try_drain_injections(
+                    spec, messages, None, injection_cycles,
+                    phase="after empty response",
+                )
+                if should_continue:
+                    had_injections = True
+                    continue
                 break
 
-            messages.append(build_assistant_message(
+            messages.append(assistant_message or build_assistant_message(
                 clean,
                 reasoning_content=response.reasoning_content,
                 thinking_blocks=response.thinking_blocks,
@@ -347,6 +510,17 @@ class AgentRunner:
                     max_iterations=spec.max_iterations,
                 )
             self._append_final_message(messages, final_content)
+            # Drain any remaining injections so they are appended to the
+            # conversation history instead of being re-published as
+            # independent inbound messages by _dispatch's finally block.
+            # We ignore should_continue here because the for-loop has already
+            # exhausted all iterations.
+            drained_after_max_iterations, injection_cycles = await self._try_drain_injections(
+                spec, messages, None, injection_cycles,
+                phase="after max_iterations",
+            )
+            if drained_after_max_iterations:
+                had_injections = True
 
         return AgentRunResult(
             final_content=final_content,
@@ -561,6 +735,12 @@ class AgentRunner:
             return
         messages.append(build_assistant_message(content))
 
+    @staticmethod
+    def _append_model_error_placeholder(messages: list[dict[str, Any]]) -> None:
+        if messages and messages[-1].get("role") == "assistant" and not messages[-1].get("tool_calls"):
+            return
+        messages.append(build_assistant_message(_PERSISTED_MODEL_ERROR_PLACEHOLDER))
+
     def _normalize_tool_result(
         self,
         spec: AgentRunSpec,
@@ -589,6 +769,99 @@ class AgentRunner:
             return truncate_text(content, spec.max_tool_result_chars)
         return content
 
+    @staticmethod
+    def _drop_orphan_tool_results(
+        messages: list[dict[str, Any]],
+    ) -> list[dict[str, Any]]:
+        """Drop tool results that have no matching assistant tool_call earlier in the history."""
+        declared: set[str] = set()
+        updated: list[dict[str, Any]] | None = None
+        for idx, msg in enumerate(messages):
+            role = msg.get("role")
+            if role == "assistant":
+                for tc in msg.get("tool_calls") or []:
+                    if isinstance(tc, dict) and tc.get("id"):
+                        declared.add(str(tc["id"]))
+            if role == "tool":
+                tid = msg.get("tool_call_id")
+                if tid and str(tid) not in declared:
+                    if updated is None:
+                        updated = [dict(m) for m in messages[:idx]]
+                    continue
+            if updated is not None:
+                updated.append(dict(msg))
+
+        if updated is None:
+            return messages
+        return updated
+
+    @staticmethod
+    def _backfill_missing_tool_results(
+        messages: list[dict[str, Any]],
+    ) -> list[dict[str, Any]]:
+        """Insert synthetic error results for orphaned tool_use blocks."""
+        declared: list[tuple[int, str, str]] = []  # (assistant_idx, call_id, name)
+        fulfilled: set[str] = set()
+        for idx, msg in enumerate(messages):
+            role = msg.get("role")
+            if role == "assistant":
+                for tc in msg.get("tool_calls") or []:
+                    if isinstance(tc, dict) and tc.get("id"):
+                        name = ""
+                        func = tc.get("function")
+                        if isinstance(func, dict):
+                            name = func.get("name", "")
+                        declared.append((idx, str(tc["id"]), name))
+            elif role == "tool":
+                tid = msg.get("tool_call_id")
+                if tid:
+                    fulfilled.add(str(tid))
+
+        missing = [(ai, cid, name) for ai, cid, name in declared if cid not in fulfilled]
+        if not missing:
+            return messages
+
+        updated = list(messages)
+        offset = 0
+        for assistant_idx, call_id, name in missing:
+            insert_at = assistant_idx + 1 + offset
+            while insert_at < len(updated) and updated[insert_at].get("role") == "tool":
+                insert_at += 1
+            updated.insert(insert_at, {
+                "role": "tool",
+                "tool_call_id": call_id,
+                "name": name,
+                "content": _BACKFILL_CONTENT,
+            })
+            offset += 1
+        return updated
+
+    @staticmethod
+    def _microcompact(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Replace old compactable tool results with one-line summaries."""
+        compactable_indices: list[int] = []
+        for idx, msg in enumerate(messages):
+            if msg.get("role") == "tool" and msg.get("name") in _COMPACTABLE_TOOLS:
+                compactable_indices.append(idx)
+
+        if len(compactable_indices) <= _MICROCOMPACT_KEEP_RECENT:
+            return messages
+
+        stale = compactable_indices[: len(compactable_indices) - _MICROCOMPACT_KEEP_RECENT]
+        updated: list[dict[str, Any]] | None = None
+        for idx in stale:
+            msg = messages[idx]
+            content = msg.get("content")
+            if not isinstance(content, str) or len(content) < _MICROCOMPACT_MIN_CHARS:
+                continue
+            name = msg.get("name", "tool")
+            summary = f"[{name} result omitted from context]"
+            if updated is None:
+                updated = [dict(m) for m in messages]
+            updated[idx]["content"] = summary
+
+        return updated if updated is not None else messages
+
     def _apply_tool_result_budget(
         self,
         spec: AgentRunSpec,

nanobot/agent/skills.py

@@ -6,6 +6,8 @@ import re
 import shutil
 from pathlib import Path
 
+import yaml
+
 # Default builtin skills directory (relative to this file)
 BUILTIN_SKILLS_DIR = Path(__file__).parent.parent / "skills"
 
@@ -16,10 +18,6 @@ _STRIP_SKILL_FRONTMATTER = re.compile(
 )
 
 
-def _escape_xml(text: str) -> str:
-    return text.replace("&", "&amp;").replace("<", "&lt;").replace(">", "&gt;")
-
-
 class SkillsLoader:
     """
     Loader for agent skills.
@@ -28,10 +26,11 @@ class SkillsLoader:
     specific tools or perform certain tasks.
     """
 
-    def __init__(self, workspace: Path, builtin_skills_dir: Path | None = None):
+    def __init__(self, workspace: Path, builtin_skills_dir: Path | None = None, disabled_skills: set[str] | None = None):
         self.workspace = workspace
         self.workspace_skills = workspace / "skills"
         self.builtin_skills = builtin_skills_dir or BUILTIN_SKILLS_DIR
+        self.disabled_skills = disabled_skills or set()
 
     def _skill_entries_from_dir(self, base: Path, source: str, *, skip_names: set[str] | None = None) -> list[dict[str, str]]:
         if not base.exists():
@@ -66,6 +65,9 @@ class SkillsLoader:
                 self._skill_entries_from_dir(self.builtin_skills, "builtin", skip_names=workspace_names)
             )
 
+        if self.disabled_skills:
+            skills = [s for s in skills if s["name"] not in self.disabled_skills]
+
         if filter_unavailable:
             return [skill for skill in skills if self._check_requirements(self._get_skill_meta(skill["name"]))]
         return skills
@@ -106,39 +108,37 @@ class SkillsLoader:
         ]
         return "\n\n---\n\n".join(parts)
 
-    def build_skills_summary(self) -> str:
+    def build_skills_summary(self, exclude: set[str] | None = None) -> str:
         """
         Build a summary of all skills (name, description, path, availability).
 
         This is used for progressive loading - the agent can read the full
         skill content using read_file when needed.
 
+        Args:
+            exclude: Set of skill names to omit from the summary.
+
         Returns:
-            XML-formatted skills summary.
+            Markdown-formatted skills summary.
         """
         all_skills = self.list_skills(filter_unavailable=False)
         if not all_skills:
             return ""
 
-        lines: list[str] = ["<skills>"]
+        lines: list[str] = []
         for entry in all_skills:
             skill_name = entry["name"]
+            if exclude and skill_name in exclude:
+                continue
             meta = self._get_skill_meta(skill_name)
             available = self._check_requirements(meta)
-            lines.extend(
-                [
-                    f'  <skill available="{str(available).lower()}">',
-                    f"    <name>{_escape_xml(skill_name)}</name>",
-                    f"    <description>{_escape_xml(self._get_skill_description(skill_name))}</description>",
-                    f"    <location>{entry['path']}</location>",
-                ]
-            )
-            if not available:
+            desc = self._get_skill_description(skill_name)
+            if available:
+                lines.append(f"- **{skill_name}** — {desc}  `{entry['path']}`")
+            else:
                 missing = self._get_missing_requirements(meta)
-                if missing:
-                    lines.append(f"    <requires>{_escape_xml(missing)}</requires>")
-            lines.append("  </skill>")
-        lines.append("</skills>")
+                suffix = f" (unavailable: {missing})" if missing else " (unavailable)"
+                lines.append(f"- **{skill_name}** — {desc}{suffix}  `{entry['path']}`")
         return "\n".join(lines)
 
     def _get_missing_requirements(self, skill_meta: dict) -> str:
@@ -167,11 +167,19 @@ class SkillsLoader:
             return content[match.end():].strip()
         return content
 
-    def _parse_nanobot_metadata(self, raw: str) -> dict:
-        """Parse skill metadata JSON from frontmatter (supports nanobot and openclaw keys)."""
-        try:
-            data = json.loads(raw)
-        except (json.JSONDecodeError, TypeError):
+    def _parse_nanobot_metadata(self, raw: object) -> dict:
+        """Extract nanobot/openclaw metadata from a frontmatter field.
+
+        ``raw`` may be a dict (already parsed by yaml.safe_load) or a JSON str.
+        """
+        if isinstance(raw, dict):
+            data = raw
+        elif isinstance(raw, str):
+            try:
+                data = json.loads(raw)
+            except (json.JSONDecodeError, TypeError):
+                return {}
+        else:
             return {}
         if not isinstance(data, dict):
             return {}
@@ -189,8 +197,8 @@ class SkillsLoader:
 
     def _get_skill_meta(self, name: str) -> dict:
         """Get nanobot metadata for a skill (cached in frontmatter)."""
-        meta = self.get_skill_metadata(name) or {}
-        return self._parse_nanobot_metadata(meta.get("metadata", ""))
+        raw_meta = self.get_skill_metadata(name) or {}
+        return self._parse_nanobot_metadata(raw_meta.get("metadata"))
 
     def get_always_skills(self) -> list[str]:
         """Get skills marked as always=true that meet requirements."""
@@ -199,7 +207,7 @@ class SkillsLoader:
             for entry in self.list_skills(filter_unavailable=True)
             if (meta := self.get_skill_metadata(entry["name"]) or {})
             and (
-                self._parse_nanobot_metadata(meta.get("metadata", "")).get("always")
+                self._parse_nanobot_metadata(meta.get("metadata")).get("always")
                 or meta.get("always")
             )
         ]
@@ -220,10 +228,15 @@ class SkillsLoader:
         match = _STRIP_SKILL_FRONTMATTER.match(content)
         if not match:
             return None
-        metadata: dict[str, str] = {}
-        for line in match.group(1).splitlines():
-            if ":" not in line:
-                continue
-            key, value = line.split(":", 1)
-            metadata[key.strip()] = value.strip().strip('"\'')
+        try:
+            parsed = yaml.safe_load(match.group(1))
+        except yaml.YAMLError:
+            return None
+        if not isinstance(parsed, dict):
+            return None
+        # yaml.safe_load returns native types (int, bool, list, etc.);
+        # keep values as-is so downstream consumers get correct types.
+        metadata: dict[str, object] = {}
+        for key, value in parsed.items():
+            metadata[str(key)] = value
         return metadata

nanobot/agent/subagent.py

@@ -2,7 +2,9 @@
 
 import asyncio
 import json
+import time
 import uuid
+from dataclasses import dataclass, field
 from pathlib import Path
 from typing import Any
 
@@ -23,11 +25,29 @@ from nanobot.config.schema import ExecToolConfig, WebToolsConfig
 from nanobot.providers.base import LLMProvider
 
 
-class _SubagentHook(AgentHook):
-    """Logging-only hook for subagent execution."""
+@dataclass(slots=True)
+class SubagentStatus:
+    """Real-time status of a running subagent."""
 
-    def __init__(self, task_id: str) -> None:
+    task_id: str
+    label: str
+    task_description: str
+    started_at: float          # time.monotonic()
+    phase: str = "initializing"  # initializing | awaiting_tools | tools_completed | final_response | done | error
+    iteration: int = 0
+    tool_events: list = field(default_factory=list)   # [{name, status, detail}, ...]
+    usage: dict = field(default_factory=dict)          # token usage
+    stop_reason: str | None = None
+    error: str | None = None
+
+
+class _SubagentHook(AgentHook):
+    """Hook for subagent execution — logs tool calls and updates status."""
+
+    def __init__(self, task_id: str, status: SubagentStatus | None = None) -> None:
+        super().__init__()
         self._task_id = task_id
+        self._status = status
 
     async def before_execute_tools(self, context: AgentHookContext) -> None:
         for tool_call in context.tool_calls:
@@ -37,6 +57,15 @@ class _SubagentHook(AgentHook):
                 self._task_id, tool_call.name, args_str,
             )
 
+    async def after_iteration(self, context: AgentHookContext) -> None:
+        if self._status is None:
+            return
+        self._status.iteration = context.iteration
+        self._status.tool_events = list(context.tool_events)
+        self._status.usage = dict(context.usage)
+        if context.error:
+            self._status.error = str(context.error)
+
 
 class SubagentManager:
     """Manages background subagent execution."""
@@ -51,9 +80,8 @@ class SubagentManager:
         web_config: "WebToolsConfig | None" = None,
         exec_config: "ExecToolConfig | None" = None,
         restrict_to_workspace: bool = False,
+        disabled_skills: list[str] | None = None,
     ):
-        from nanobot.config.schema import ExecToolConfig
-
         self.provider = provider
         self.workspace = workspace
         self.bus = bus
@@ -62,8 +90,10 @@ class SubagentManager:
         self.max_tool_result_chars = max_tool_result_chars
         self.exec_config = exec_config or ExecToolConfig()
         self.restrict_to_workspace = restrict_to_workspace
+        self.disabled_skills = set(disabled_skills or [])
         self.runner = AgentRunner(provider)
         self._running_tasks: dict[str, asyncio.Task[None]] = {}
+        self._task_statuses: dict[str, SubagentStatus] = {}
         self._session_tasks: dict[str, set[str]] = {}  # session_key -> {task_id, ...}
 
     async def spawn(
@@ -79,8 +109,16 @@ class SubagentManager:
         display_label = label or task[:30] + ("..." if len(task) > 30 else "")
         origin = {"channel": origin_channel, "chat_id": origin_chat_id}
 
+        status = SubagentStatus(
+            task_id=task_id,
+            label=display_label,
+            task_description=task,
+            started_at=time.monotonic(),
+        )
+        self._task_statuses[task_id] = status
+
         bg_task = asyncio.create_task(
-            self._run_subagent(task_id, task, display_label, origin)
+            self._run_subagent(task_id, task, display_label, origin, status)
         )
         self._running_tasks[task_id] = bg_task
         if session_key:
@@ -88,6 +126,7 @@ class SubagentManager:
 
         def _cleanup(_: asyncio.Task) -> None:
             self._running_tasks.pop(task_id, None)
+            self._task_statuses.pop(task_id, None)
             if session_key and (ids := self._session_tasks.get(session_key)):
                 ids.discard(task_id)
                 if not ids:
@@ -104,10 +143,15 @@ class SubagentManager:
         task: str,
         label: str,
         origin: dict[str, str],
+        status: SubagentStatus,
     ) -> None:
         """Execute the subagent task and announce the result."""
         logger.info("Subagent [{}] starting task: {}", task_id, label)
 
+        async def _on_checkpoint(payload: dict) -> None:
+            status.phase = payload.get("phase", status.phase)
+            status.iteration = payload.get("iteration", status.iteration)
+
         try:
             # Build subagent tools (no message tool, no spawn tool)
             tools = ToolRegistry()
@@ -142,40 +186,38 @@ class SubagentManager:
                 model=self.model,
                 max_iterations=15,
                 max_tool_result_chars=self.max_tool_result_chars,
-                hook=_SubagentHook(task_id),
+                hook=_SubagentHook(task_id, status),
                 max_iterations_message="Task completed but no final response was generated.",
                 error_message=None,
                 fail_on_tool_error=True,
+                checkpoint_callback=_on_checkpoint,
             ))
-            if result.stop_reason == "tool_error":
-                await self._announce_result(
-                    task_id,
-                    label,
-                    task,
-                    self._format_partial_progress(result),
-                    origin,
-                    "error",
-                )
-                return
-            if result.stop_reason == "error":
-                await self._announce_result(
-                    task_id,
-                    label,
-                    task,
-                    result.error or "Error: subagent execution failed.",
-                    origin,
-                    "error",
-                )
-                return
-            final_result = result.final_content or "Task completed but no final response was generated."
+            status.phase = "done"
+            status.stop_reason = result.stop_reason
 
-            logger.info("Subagent [{}] completed successfully", task_id)
-            await self._announce_result(task_id, label, task, final_result, origin, "ok")
+            if result.stop_reason == "tool_error":
+                status.tool_events = list(result.tool_events)
+                await self._announce_result(
+                    task_id, label, task,
+                    self._format_partial_progress(result),
+                    origin, "error",
+                )
+            elif result.stop_reason == "error":
+                await self._announce_result(
+                    task_id, label, task,
+                    result.error or "Error: subagent execution failed.",
+                    origin, "error",
+                )
+            else:
+                final_result = result.final_content or "Task completed but no final response was generated."
+                logger.info("Subagent [{}] completed successfully", task_id)
+                await self._announce_result(task_id, label, task, final_result, origin, "ok")
 
         except Exception as e:
-            error_msg = f"Error: {str(e)}"
+            status.phase = "error"
+            status.error = str(e)
             logger.error("Subagent [{}] failed: {}", task_id, e)
-            await self._announce_result(task_id, label, task, error_msg, origin, "error")
+            await self._announce_result(task_id, label, task, f"Error: {e}", origin, "error")
 
     async def _announce_result(
         self,
@@ -235,7 +277,10 @@ class SubagentManager:
         from nanobot.agent.skills import SkillsLoader
 
         time_ctx = ContextBuilder._build_runtime_context(None, None)
-        skills_summary = SkillsLoader(self.workspace).build_skills_summary()
+        skills_summary = SkillsLoader(
+            self.workspace,
+            disabled_skills=self.disabled_skills,
+        ).build_skills_summary()
         return render_template(
             "agent/subagent_system.md",
             time_ctx=time_ctx,
@@ -256,3 +301,11 @@ class SubagentManager:
     def get_running_count(self) -> int:
         """Return the number of currently running subagents."""
         return len(self._running_tasks)
+
+    def get_running_count_by_session(self, session_key: str) -> int:
+        """Return the number of currently running subagents for a session."""
+        tids = self._session_tasks.get(session_key, set())
+        return sum(
+            1 for tid in tids
+            if tid in self._running_tasks and not self._running_tasks[tid].done()
+        )

nanobot/agent/tools/file_state.py

@@ -0,0 +1,105 @@
+"""Track file-read state for read-before-edit warnings and read deduplication."""
+
+from __future__ import annotations
+
+import hashlib
+import os
+from dataclasses import dataclass
+from pathlib import Path
+
+
+@dataclass(slots=True)
+class ReadState:
+    mtime: float
+    offset: int
+    limit: int | None
+    content_hash: str | None
+    can_dedup: bool
+
+
+_state: dict[str, ReadState] = {}
+
+
+def _hash_file(p: str) -> str | None:
+    try:
+        return hashlib.sha256(Path(p).read_bytes()).hexdigest()
+    except OSError:
+        return None
+
+
+def record_read(path: str | Path, offset: int = 1, limit: int | None = None) -> None:
+    """Record that a file was read (called after successful read)."""
+    p = str(Path(path).resolve())
+    try:
+        mtime = os.path.getmtime(p)
+    except OSError:
+        return
+    _state[p] = ReadState(
+        mtime=mtime,
+        offset=offset,
+        limit=limit,
+        content_hash=_hash_file(p),
+        can_dedup=True,
+    )
+
+
+def record_write(path: str | Path) -> None:
+    """Record that a file was written (updates mtime in state)."""
+    p = str(Path(path).resolve())
+    try:
+        mtime = os.path.getmtime(p)
+    except OSError:
+        _state.pop(p, None)
+        return
+    _state[p] = ReadState(
+        mtime=mtime,
+        offset=1,
+        limit=None,
+        content_hash=_hash_file(p),
+        can_dedup=False,
+    )
+
+
+def check_read(path: str | Path) -> str | None:
+    """Check if a file has been read and is fresh.
+
+    Returns None if OK, or a warning string.
+    When mtime changed but file content is identical (e.g. touch, editor save),
+    the check passes to avoid false-positive staleness warnings.
+    """
+    p = str(Path(path).resolve())
+    entry = _state.get(p)
+    if entry is None:
+        return "Warning: file has not been read yet. Read it first to verify content before editing."
+    try:
+        current_mtime = os.path.getmtime(p)
+    except OSError:
+        return None
+    if current_mtime != entry.mtime:
+        if entry.content_hash and _hash_file(p) == entry.content_hash:
+            entry.mtime = current_mtime
+            return None
+        return "Warning: file has been modified since last read. Re-read to verify content before editing."
+    return None
+
+
+def is_unchanged(path: str | Path, offset: int = 1, limit: int | None = None) -> bool:
+    """Return True if file was previously read with same params and mtime is unchanged."""
+    p = str(Path(path).resolve())
+    entry = _state.get(p)
+    if entry is None:
+        return False
+    if not entry.can_dedup:
+        return False
+    if entry.offset != offset or entry.limit != limit:
+        return False
+    try:
+        current_mtime = os.path.getmtime(p)
+    except OSError:
+        return False
+    return current_mtime == entry.mtime
+
+
+def clear() -> None:
+    """Clear all tracked state (useful for testing)."""
+    _state.clear()

nanobot/agent/tools/filesystem.py

@@ -2,11 +2,13 @@
 
 import difflib
 import mimetypes
+from dataclasses import dataclass
 from pathlib import Path
 from typing import Any
 
 from nanobot.agent.tools.base import Tool, tool_parameters
 from nanobot.agent.tools.schema import BooleanSchema, IntegerSchema, StringSchema, tool_parameters_schema
+from nanobot.agent.tools import file_state
 from nanobot.utils.helpers import build_image_content_blocks, detect_image_mime
 from nanobot.config.paths import get_media_dir
 
@@ -60,6 +62,36 @@ class _FsTool(Tool):
 # ---------------------------------------------------------------------------
 
 
+_BLOCKED_DEVICE_PATHS = frozenset({
+    "/dev/zero", "/dev/random", "/dev/urandom", "/dev/full",
+    "/dev/stdin", "/dev/stdout", "/dev/stderr",
+    "/dev/tty", "/dev/console",
+    "/dev/fd/0", "/dev/fd/1", "/dev/fd/2",
+})
+
+
+def _is_blocked_device(path: str | Path) -> bool:
+    """Check if path is a blocked device that could hang or produce infinite output."""
+    import re
+    raw = str(path)
+    if raw in _BLOCKED_DEVICE_PATHS:
+        return True
+    if re.match(r"/proc/\d+/fd/[012]$", raw) or re.match(r"/proc/self/fd/[012]$", raw):
+        return True
+    return False
+
+
+def _parse_page_range(pages: str, total: int) -> tuple[int, int]:
+    """Parse a page range like '2-5' into 0-based (start, end) inclusive."""
+    parts = pages.strip().split("-")
+    if len(parts) == 1:
+        p = int(parts[0])
+        return max(0, p - 1), min(p - 1, total - 1)
+    start = int(parts[0])
+    end = int(parts[1])
+    return max(0, start - 1), min(end - 1, total - 1)
+
+
 @tool_parameters(
     tool_parameters_schema(
         path=StringSchema("The file path to read"),
@@ -73,6 +105,7 @@ class _FsTool(Tool):
             description="Maximum number of lines to read (default 2000)",
             minimum=1,
         ),
+        pages=StringSchema("Page range for PDF files, e.g. '1-5' (default: all, max 20 pages)"),
         required=["path"],
     )
 )
@@ -81,6 +114,7 @@ class ReadFileTool(_FsTool):
 
     _MAX_CHARS = 128_000
     _DEFAULT_LIMIT = 2000
+    _MAX_PDF_PAGES = 20
 
     @property
     def name(self) -> str:
@@ -89,24 +123,38 @@ class ReadFileTool(_FsTool):
     @property
     def description(self) -> str:
         return (
-            "Read the contents of a file. Returns numbered lines. "
-            "Use offset and limit to paginate through large files."
+            "Read a file (text or image). Text output format: LINE_NUM|CONTENT. "
+            "Images return visual content for analysis. "
+            "Use offset and limit for large files. "
+            "Cannot read non-image binary files. "
+            "Reads exceeding ~128K chars are truncated."
         )
 
     @property
     def read_only(self) -> bool:
         return True
 
-    async def execute(self, path: str | None = None, offset: int = 1, limit: int | None = None, **kwargs: Any) -> Any:
+    async def execute(self, path: str | None = None, offset: int = 1, limit: int | None = None, pages: str | None = None, **kwargs: Any) -> Any:
         try:
             if not path:
                 return "Error reading file: Unknown path"
+
+            # Device path blacklist
+            if _is_blocked_device(path):
+                return f"Error: Reading {path} is blocked (device path that could hang or produce infinite output)."
+
             fp = self._resolve(path)
+            if _is_blocked_device(fp):
+                return f"Error: Reading {fp} is blocked (device path that could hang or produce infinite output)."
             if not fp.exists():
                 return f"Error: File not found: {path}"
             if not fp.is_file():
                 return f"Error: Not a file: {path}"
 
+            # PDF support
+            if fp.suffix.lower() == ".pdf":
+                return self._read_pdf(fp, pages)
+
             raw = fp.read_bytes()
             if not raw:
                 return f"(Empty file: {path})"
@@ -115,6 +163,10 @@ class ReadFileTool(_FsTool):
             if mime and mime.startswith("image/"):
                 return build_image_content_blocks(raw, mime, str(fp), f"(Image file: {path})")
 
+            # Read dedup: same path + offset + limit + unchanged mtime → stub
+            if file_state.is_unchanged(fp, offset=offset, limit=limit):
+                return f"[File unchanged since last read: {path}]"
+
             try:
                 text_content = raw.decode("utf-8")
             except UnicodeDecodeError:
@@ -147,12 +199,59 @@ class ReadFileTool(_FsTool):
                 result += f"\n\n(Showing lines {offset}-{end} of {total}. Use offset={end + 1} to continue.)"
             else:
                 result += f"\n\n(End of file — {total} lines total)"
+            file_state.record_read(fp, offset=offset, limit=limit)
             return result
         except PermissionError as e:
             return f"Error: {e}"
         except Exception as e:
             return f"Error reading file: {e}"
 
+    def _read_pdf(self, fp: Path, pages: str | None) -> str:
+        try:
+            import fitz  # pymupdf
+        except ImportError:
+            return "Error: PDF reading requires pymupdf. Install with: pip install pymupdf"
+
+        try:
+            doc = fitz.open(str(fp))
+        except Exception as e:
+            return f"Error reading PDF: {e}"
+
+        total_pages = len(doc)
+        if pages:
+            try:
+                start, end = _parse_page_range(pages, total_pages)
+            except (ValueError, IndexError):
+                doc.close()
+                return f"Error: Invalid page range '{pages}'. Use format like '1-5'."
+            if start > end or start >= total_pages:
+                doc.close()
+                return f"Error: Page range '{pages}' is out of bounds (document has {total_pages} pages)."
+        else:
+            start = 0
+            end = min(total_pages - 1, self._MAX_PDF_PAGES - 1)
+
+        if end - start + 1 > self._MAX_PDF_PAGES:
+            end = start + self._MAX_PDF_PAGES - 1
+
+        parts: list[str] = []
+        for i in range(start, end + 1):
+            page = doc[i]
+            text = page.get_text().strip()
+            if text:
+                parts.append(f"--- Page {i + 1} ---\n{text}")
+        doc.close()
+
+        if not parts:
+            return f"(PDF has no extractable text: {fp})"
+
+        result = "\n\n".join(parts)
+        if end < total_pages - 1:
+            result += f"\n\n(Showing pages {start + 1}-{end + 1} of {total_pages}. Use pages='{end + 2}-{min(end + 1 + self._MAX_PDF_PAGES, total_pages)}' to continue.)"
+        if len(result) > self._MAX_CHARS:
+            result = result[:self._MAX_CHARS] + "\n\n(PDF text truncated at ~128K chars)"
+        return result
+
 
 # ---------------------------------------------------------------------------
 # write_file
@@ -175,7 +274,11 @@ class WriteFileTool(_FsTool):
 
     @property
     def description(self) -> str:
-        return "Write content to a file at the given path. Creates parent directories if needed."
+        return (
+            "Write content to a file. Overwrites if the file already exists; "
+            "creates parent directories as needed. "
+            "For partial edits, prefer edit_file instead."
+        )
 
     async def execute(self, path: str | None = None, content: str | None = None, **kwargs: Any) -> str:
         try:
@@ -186,6 +289,7 @@ class WriteFileTool(_FsTool):
             fp = self._resolve(path)
             fp.parent.mkdir(parents=True, exist_ok=True)
             fp.write_text(content, encoding="utf-8")
+            file_state.record_write(fp)
             return f"Successfully wrote {len(content)} characters to {fp}"
         except PermissionError as e:
             return f"Error: {e}"
@@ -197,30 +301,269 @@ class WriteFileTool(_FsTool):
 # edit_file
 # ---------------------------------------------------------------------------
 
+_QUOTE_TABLE = str.maketrans({
+    "\u2018": "'", "\u2019": "'",  # curly single → straight
+    "\u201c": '"', "\u201d": '"',  # curly double → straight
+    "'": "'", '"': '"',            # identity (kept for completeness)
+})
+
+
+def _normalize_quotes(s: str) -> str:
+    return s.translate(_QUOTE_TABLE)
+
+
+def _curly_double_quotes(text: str) -> str:
+    parts: list[str] = []
+    opening = True
+    for ch in text:
+        if ch == '"':
+            parts.append("\u201c" if opening else "\u201d")
+            opening = not opening
+        else:
+            parts.append(ch)
+    return "".join(parts)
+
+
+def _curly_single_quotes(text: str) -> str:
+    parts: list[str] = []
+    opening = True
+    for i, ch in enumerate(text):
+        if ch != "'":
+            parts.append(ch)
+            continue
+        prev_ch = text[i - 1] if i > 0 else ""
+        next_ch = text[i + 1] if i + 1 < len(text) else ""
+        if prev_ch.isalnum() and next_ch.isalnum():
+            parts.append("\u2019")
+            continue
+        parts.append("\u2018" if opening else "\u2019")
+        opening = not opening
+    return "".join(parts)
+
+
+def _preserve_quote_style(old_text: str, actual_text: str, new_text: str) -> str:
+    """Preserve curly quote style when a quote-normalized fallback matched."""
+    if _normalize_quotes(old_text.strip()) != _normalize_quotes(actual_text.strip()) or old_text == actual_text:
+        return new_text
+
+    styled = new_text
+    if any(ch in actual_text for ch in ("\u201c", "\u201d")) and '"' in styled:
+        styled = _curly_double_quotes(styled)
+    if any(ch in actual_text for ch in ("\u2018", "\u2019")) and "'" in styled:
+        styled = _curly_single_quotes(styled)
+    return styled
+
+
+def _leading_ws(line: str) -> str:
+    return line[: len(line) - len(line.lstrip(" \t"))]
+
+
+def _reindent_like_match(old_text: str, actual_text: str, new_text: str) -> str:
+    """Preserve the outer indentation from the actual matched block."""
+    old_lines = old_text.split("\n")
+    actual_lines = actual_text.split("\n")
+    if len(old_lines) != len(actual_lines):
+        return new_text
+
+    comparable = [
+        (old_line, actual_line)
+        for old_line, actual_line in zip(old_lines, actual_lines)
+        if old_line.strip() and actual_line.strip()
+    ]
+    if not comparable or any(
+        _normalize_quotes(old_line.strip()) != _normalize_quotes(actual_line.strip())
+        for old_line, actual_line in comparable
+    ):
+        return new_text
+
+    old_ws = _leading_ws(comparable[0][0])
+    actual_ws = _leading_ws(comparable[0][1])
+    if actual_ws == old_ws:
+        return new_text
+
+    if old_ws:
+        if not actual_ws.startswith(old_ws):
+            return new_text
+        delta = actual_ws[len(old_ws):]
+    else:
+        delta = actual_ws
+
+    if not delta:
+        return new_text
+
+    return "\n".join((delta + line) if line else line for line in new_text.split("\n"))
+
+
+@dataclass(slots=True)
+class _MatchSpan:
+    start: int
+    end: int
+    text: str
+    line: int
+
+
+def _find_exact_matches(content: str, old_text: str) -> list[_MatchSpan]:
+    matches: list[_MatchSpan] = []
+    start = 0
+    while True:
+        idx = content.find(old_text, start)
+        if idx == -1:
+            break
+        matches.append(
+            _MatchSpan(
+                start=idx,
+                end=idx + len(old_text),
+                text=content[idx : idx + len(old_text)],
+                line=content.count("\n", 0, idx) + 1,
+            )
+        )
+        start = idx + max(1, len(old_text))
+    return matches
+
+
+def _find_trim_matches(content: str, old_text: str, *, normalize_quotes: bool = False) -> list[_MatchSpan]:
+    old_lines = old_text.splitlines()
+    if not old_lines:
+        return []
+
+    content_lines = content.splitlines()
+    content_lines_keepends = content.splitlines(keepends=True)
+    if len(content_lines) < len(old_lines):
+        return []
+
+    offsets: list[int] = []
+    pos = 0
+    for line in content_lines_keepends:
+        offsets.append(pos)
+        pos += len(line)
+    offsets.append(pos)
+
+    if normalize_quotes:
+        stripped_old = [_normalize_quotes(line.strip()) for line in old_lines]
+    else:
+        stripped_old = [line.strip() for line in old_lines]
+
+    matches: list[_MatchSpan] = []
+    window_size = len(stripped_old)
+    for i in range(len(content_lines) - window_size + 1):
+        window = content_lines[i : i + window_size]
+        if normalize_quotes:
+            comparable = [_normalize_quotes(line.strip()) for line in window]
+        else:
+            comparable = [line.strip() for line in window]
+        if comparable != stripped_old:
+            continue
+
+        start = offsets[i]
+        end = offsets[i + window_size]
+        if content_lines_keepends[i + window_size - 1].endswith("\n"):
+            end -= 1
+        matches.append(
+            _MatchSpan(
+                start=start,
+                end=end,
+                text=content[start:end],
+                line=i + 1,
+            )
+        )
+    return matches
+
+
+def _find_quote_matches(content: str, old_text: str) -> list[_MatchSpan]:
+    norm_content = _normalize_quotes(content)
+    norm_old = _normalize_quotes(old_text)
+    matches: list[_MatchSpan] = []
+    start = 0
+    while True:
+        idx = norm_content.find(norm_old, start)
+        if idx == -1:
+            break
+        matches.append(
+            _MatchSpan(
+                start=idx,
+                end=idx + len(old_text),
+                text=content[idx : idx + len(old_text)],
+                line=content.count("\n", 0, idx) + 1,
+            )
+        )
+        start = idx + max(1, len(norm_old))
+    return matches
+
+
+def _find_matches(content: str, old_text: str) -> list[_MatchSpan]:
+    """Locate all matches using progressively looser strategies."""
+    for matcher in (
+        lambda: _find_exact_matches(content, old_text),
+        lambda: _find_trim_matches(content, old_text),
+        lambda: _find_trim_matches(content, old_text, normalize_quotes=True),
+        lambda: _find_quote_matches(content, old_text),
+    ):
+        matches = matcher()
+        if matches:
+            return matches
+    return []
+
+
+def _find_match_line_numbers(content: str, old_text: str) -> list[int]:
+    """Return 1-based starting line numbers for the current matching strategies."""
+    return [match.line for match in _find_matches(content, old_text)]
+
+
+def _collapse_internal_whitespace(text: str) -> str:
+    return "\n".join(" ".join(line.split()) for line in text.splitlines())
+
+
+def _diagnose_near_match(old_text: str, actual_text: str) -> list[str]:
+    """Return actionable hints describing why text was close but not exact."""
+    hints: list[str] = []
+
+    if old_text.lower() == actual_text.lower() and old_text != actual_text:
+        hints.append("letter case differs")
+    if _collapse_internal_whitespace(old_text) == _collapse_internal_whitespace(actual_text) and old_text != actual_text:
+        hints.append("whitespace differs")
+    if old_text.rstrip("\n") == actual_text.rstrip("\n") and old_text != actual_text:
+        hints.append("trailing newline differs")
+    if _normalize_quotes(old_text) == _normalize_quotes(actual_text) and old_text != actual_text:
+        hints.append("quote style differs")
+
+    return hints
+
+
+def _best_window(old_text: str, content: str) -> tuple[float, int, list[str], list[str]]:
+    """Find the closest line-window match and return ratio/start/snippet/hints."""
+    lines = content.splitlines(keepends=True)
+    old_lines = old_text.splitlines(keepends=True)
+    window = max(1, len(old_lines))
+
+    best_ratio, best_start = -1.0, 0
+    best_window_lines: list[str] = []
+
+    for i in range(max(1, len(lines) - window + 1)):
+        current = lines[i : i + window]
+        ratio = difflib.SequenceMatcher(None, old_lines, current).ratio()
+        if ratio > best_ratio:
+            best_ratio, best_start = ratio, i
+            best_window_lines = current
+
+    actual_text = "".join(best_window_lines).replace("\r\n", "\n").rstrip("\n")
+    hints = _diagnose_near_match(old_text.replace("\r\n", "\n").rstrip("\n"), actual_text)
+    return best_ratio, best_start, best_window_lines, hints
+
+
 def _find_match(content: str, old_text: str) -> tuple[str | None, int]:
-    """Locate old_text in content: exact first, then line-trimmed sliding window.
+    """Locate old_text in content with a multi-level fallback chain:
+
+    1. Exact substring match
+    2. Line-trimmed sliding window (handles indentation differences)
+    3. Smart quote normalization (curly ↔ straight quotes)
 
     Both inputs should use LF line endings (caller normalises CRLF).
     Returns (matched_fragment, count) or (None, 0).
     """
-    if old_text in content:
-        return old_text, content.count(old_text)
-
-    old_lines = old_text.splitlines()
-    if not old_lines:
+    matches = _find_matches(content, old_text)
+    if not matches:
         return None, 0
-    stripped_old = [l.strip() for l in old_lines]
-    content_lines = content.splitlines()
-
-    candidates = []
-    for i in range(len(content_lines) - len(stripped_old) + 1):
-        window = content_lines[i : i + len(stripped_old)]
-        if [l.strip() for l in window] == stripped_old:
-            candidates.append("\n".join(window))
-
-    if candidates:
-        return candidates[0], len(candidates)
-    return None, 0
+    return matches[0].text, len(matches)
 
 
 @tool_parameters(
@@ -235,6 +578,9 @@ def _find_match(content: str, old_text: str) -> tuple[str | None, int]:
 class EditFileTool(_FsTool):
     """Edit a file by replacing text with fallback matching."""
 
+    _MAX_EDIT_FILE_SIZE = 1024 * 1024 * 1024  # 1 GiB
+    _MARKDOWN_EXTS = frozenset({".md", ".mdx", ".markdown"})
+
     @property
     def name(self) -> str:
         return "edit_file"
@@ -243,10 +589,16 @@ class EditFileTool(_FsTool):
     def description(self) -> str:
         return (
             "Edit a file by replacing old_text with new_text. "
-            "Supports minor whitespace/line-ending differences. "
-            "Set replace_all=true to replace every occurrence."
+            "Tolerates minor whitespace/indentation differences and curly/straight quote mismatches. "
+            "If old_text matches multiple times, you must provide more context "
+            "or set replace_all=true. Shows a diff of the closest match on failure."
         )
 
+    @staticmethod
+    def _strip_trailing_ws(text: str) -> str:
+        """Strip trailing whitespace from each line."""
+        return "\n".join(line.rstrip() for line in text.split("\n"))
+
     async def execute(
         self, path: str | None = None, old_text: str | None = None,
         new_text: str | None = None,
@@ -260,55 +612,133 @@ class EditFileTool(_FsTool):
             if new_text is None:
                 raise ValueError("Unknown new_text")
 
+            # .ipynb detection
+            if path.endswith(".ipynb"):
+                return "Error: This is a Jupyter notebook. Use the notebook_edit tool instead of edit_file."
+
             fp = self._resolve(path)
+
+            # Create-file semantics: old_text='' + file doesn't exist → create
             if not fp.exists():
-                return f"Error: File not found: {path}"
+                if old_text == "":
+                    fp.parent.mkdir(parents=True, exist_ok=True)
+                    fp.write_text(new_text, encoding="utf-8")
+                    file_state.record_write(fp)
+                    return f"Successfully created {fp}"
+                return self._file_not_found_msg(path, fp)
+
+            # File size protection
+            try:
+                fsize = fp.stat().st_size
+            except OSError:
+                fsize = 0
+            if fsize > self._MAX_EDIT_FILE_SIZE:
+                return f"Error: File too large to edit ({fsize / (1024**3):.1f} GiB). Maximum is 1 GiB."
+
+            # Create-file: old_text='' but file exists and not empty → reject
+            if old_text == "":
+                raw = fp.read_bytes()
+                content = raw.decode("utf-8")
+                if content.strip():
+                    return f"Error: Cannot create file — {path} already exists and is not empty."
+                fp.write_text(new_text, encoding="utf-8")
+                file_state.record_write(fp)
+                return f"Successfully edited {fp}"
+
+            # Read-before-edit check
+            warning = file_state.check_read(fp)
 
             raw = fp.read_bytes()
             uses_crlf = b"\r\n" in raw
             content = raw.decode("utf-8").replace("\r\n", "\n")
-            match, count = _find_match(content, old_text.replace("\r\n", "\n"))
+            norm_old = old_text.replace("\r\n", "\n")
+            matches = _find_matches(content, norm_old)
 
-            if match is None:
+            if not matches:
                 return self._not_found_msg(old_text, content, path)
+            count = len(matches)
             if count > 1 and not replace_all:
+                line_numbers = [match.line for match in matches]
+                preview = ", ".join(f"line {n}" for n in line_numbers[:3])
+                if len(line_numbers) > 3:
+                    preview += ", ..."
+                location_hint = f" at {preview}" if preview else ""
                 return (
-                    f"Warning: old_text appears {count} times. "
+                    f"Warning: old_text appears {count} times{location_hint}. "
                     "Provide more context to make it unique, or set replace_all=true."
                 )
 
             norm_new = new_text.replace("\r\n", "\n")
-            new_content = content.replace(match, norm_new) if replace_all else content.replace(match, norm_new, 1)
+
+            # Trailing whitespace stripping (skip markdown to preserve double-space line breaks)
+            if fp.suffix.lower() not in self._MARKDOWN_EXTS:
+                norm_new = self._strip_trailing_ws(norm_new)
+
+            selected = matches if replace_all else matches[:1]
+            new_content = content
+            for match in reversed(selected):
+                replacement = _preserve_quote_style(norm_old, match.text, norm_new)
+                replacement = _reindent_like_match(norm_old, match.text, replacement)
+
+                # Delete-line cleanup: when deleting text (new_text=''), consume trailing
+                # newline to avoid leaving a blank line
+                end = match.end
+                if replacement == "" and not match.text.endswith("\n") and content[end:end + 1] == "\n":
+                    end += 1
+
+                new_content = new_content[: match.start] + replacement + new_content[end:]
             if uses_crlf:
                 new_content = new_content.replace("\n", "\r\n")
 
             fp.write_bytes(new_content.encode("utf-8"))
-            return f"Successfully edited {fp}"
+            file_state.record_write(fp)
+            msg = f"Successfully edited {fp}"
+            if warning:
+                msg = f"{warning}\n{msg}"
+            return msg
         except PermissionError as e:
             return f"Error: {e}"
         except Exception as e:
             return f"Error editing file: {e}"
 
+    def _file_not_found_msg(self, path: str, fp: Path) -> str:
+        """Build an error message with 'Did you mean ...?' suggestions."""
+        parent = fp.parent
+        suggestions: list[str] = []
+        if parent.is_dir():
+            siblings = [f.name for f in parent.iterdir() if f.is_file()]
+            close = difflib.get_close_matches(fp.name, siblings, n=3, cutoff=0.6)
+            suggestions = [str(parent / c) for c in close]
+        parts = [f"Error: File not found: {path}"]
+        if suggestions:
+            parts.append("Did you mean: " + ", ".join(suggestions) + "?")
+        return "\n".join(parts)
+
     @staticmethod
     def _not_found_msg(old_text: str, content: str, path: str) -> str:
-        lines = content.splitlines(keepends=True)
-        old_lines = old_text.splitlines(keepends=True)
-        window = len(old_lines)
-
-        best_ratio, best_start = 0.0, 0
-        for i in range(max(1, len(lines) - window + 1)):
-            ratio = difflib.SequenceMatcher(None, old_lines, lines[i : i + window]).ratio()
-            if ratio > best_ratio:
-                best_ratio, best_start = ratio, i
-
+        best_ratio, best_start, best_window_lines, hints = _best_window(old_text, content)
         if best_ratio > 0.5:
             diff = "\n".join(difflib.unified_diff(
-                old_lines, lines[best_start : best_start + window],
+                old_text.splitlines(keepends=True),
+                best_window_lines,
                 fromfile="old_text (provided)",
                 tofile=f"{path} (actual, line {best_start + 1})",
                 lineterm="",
             ))
-            return f"Error: old_text not found in {path}.\nBest match ({best_ratio:.0%} similar) at line {best_start + 1}:\n{diff}"
+            hint_text = ""
+            if hints:
+                hint_text = "\nPossible cause: " + ", ".join(hints) + "."
+            return (
+                f"Error: old_text not found in {path}."
+                f"{hint_text}\nBest match ({best_ratio:.0%} similar) at line {best_start + 1}:\n{diff}"
+            )
+
+        if hints:
+            return (
+                f"Error: old_text not found in {path}. "
+                f"Possible cause: {', '.join(hints)}. "
+                "Copy the exact text from read_file and try again."
+            )
         return f"Error: old_text not found in {path}. No similar text found. Verify the file content."
 
 

nanobot/agent/tools/mcp.py

@@ -57,9 +57,7 @@ def _normalize_schema_for_openai(schema: Any) -> dict[str, Any]:
 
     if "properties" in normalized and isinstance(normalized["properties"], dict):
         normalized["properties"] = {
-            name: _normalize_schema_for_openai(prop)
-            if isinstance(prop, dict)
-            else prop
+            name: _normalize_schema_for_openai(prop) if isinstance(prop, dict) else prop
             for name, prop in normalized["properties"].items()
         }
 
@@ -135,36 +133,214 @@ class MCPToolWrapper(Tool):
         return "\n".join(parts) or "(no output)"
 
 
+class MCPResourceWrapper(Tool):
+    """Wraps an MCP resource URI as a read-only nanobot Tool."""
+
+    def __init__(self, session, server_name: str, resource_def, resource_timeout: int = 30):
+        self._session = session
+        self._uri = resource_def.uri
+        self._name = f"mcp_{server_name}_resource_{resource_def.name}"
+        desc = resource_def.description or resource_def.name
+        self._description = f"[MCP Resource] {desc}\nURI: {self._uri}"
+        self._parameters: dict[str, Any] = {
+            "type": "object",
+            "properties": {},
+            "required": [],
+        }
+        self._resource_timeout = resource_timeout
+
+    @property
+    def name(self) -> str:
+        return self._name
+
+    @property
+    def description(self) -> str:
+        return self._description
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return self._parameters
+
+    @property
+    def read_only(self) -> bool:
+        return True
+
+    async def execute(self, **kwargs: Any) -> str:
+        from mcp import types
+
+        try:
+            result = await asyncio.wait_for(
+                self._session.read_resource(self._uri),
+                timeout=self._resource_timeout,
+            )
+        except asyncio.TimeoutError:
+            logger.warning(
+                "MCP resource '{}' timed out after {}s", self._name, self._resource_timeout
+            )
+            return f"(MCP resource read timed out after {self._resource_timeout}s)"
+        except asyncio.CancelledError:
+            task = asyncio.current_task()
+            if task is not None and task.cancelling() > 0:
+                raise
+            logger.warning("MCP resource '{}' was cancelled by server/SDK", self._name)
+            return "(MCP resource read was cancelled)"
+        except Exception as exc:
+            logger.exception(
+                "MCP resource '{}' failed: {}: {}",
+                self._name,
+                type(exc).__name__,
+                exc,
+            )
+            return f"(MCP resource read failed: {type(exc).__name__})"
+
+        parts: list[str] = []
+        for block in result.contents:
+            if isinstance(block, types.TextResourceContents):
+                parts.append(block.text)
+            elif isinstance(block, types.BlobResourceContents):
+                parts.append(f"[Binary resource: {len(block.blob)} bytes]")
+            else:
+                parts.append(str(block))
+        return "\n".join(parts) or "(no output)"
+
+
+class MCPPromptWrapper(Tool):
+    """Wraps an MCP prompt as a read-only nanobot Tool."""
+
+    def __init__(self, session, server_name: str, prompt_def, prompt_timeout: int = 30):
+        self._session = session
+        self._prompt_name = prompt_def.name
+        self._name = f"mcp_{server_name}_prompt_{prompt_def.name}"
+        desc = prompt_def.description or prompt_def.name
+        self._description = (
+            f"[MCP Prompt] {desc}\n"
+            "Returns a filled prompt template that can be used as a workflow guide."
+        )
+        self._prompt_timeout = prompt_timeout
+
+        # Build parameters from prompt arguments
+        properties: dict[str, Any] = {}
+        required: list[str] = []
+        for arg in prompt_def.arguments or []:
+            prop: dict[str, Any] = {"type": "string"}
+            if getattr(arg, "description", None):
+                prop["description"] = arg.description
+            properties[arg.name] = prop
+            if arg.required:
+                required.append(arg.name)
+        self._parameters: dict[str, Any] = {
+            "type": "object",
+            "properties": properties,
+            "required": required,
+        }
+
+    @property
+    def name(self) -> str:
+        return self._name
+
+    @property
+    def description(self) -> str:
+        return self._description
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return self._parameters
+
+    @property
+    def read_only(self) -> bool:
+        return True
+
+    async def execute(self, **kwargs: Any) -> str:
+        from mcp import types
+        from mcp.shared.exceptions import McpError
+
+        try:
+            result = await asyncio.wait_for(
+                self._session.get_prompt(self._prompt_name, arguments=kwargs),
+                timeout=self._prompt_timeout,
+            )
+        except asyncio.TimeoutError:
+            logger.warning("MCP prompt '{}' timed out after {}s", self._name, self._prompt_timeout)
+            return f"(MCP prompt call timed out after {self._prompt_timeout}s)"
+        except asyncio.CancelledError:
+            task = asyncio.current_task()
+            if task is not None and task.cancelling() > 0:
+                raise
+            logger.warning("MCP prompt '{}' was cancelled by server/SDK", self._name)
+            return "(MCP prompt call was cancelled)"
+        except McpError as exc:
+            logger.error(
+                "MCP prompt '{}' failed: code={} message={}",
+                self._name,
+                exc.error.code,
+                exc.error.message,
+            )
+            return f"(MCP prompt call failed: {exc.error.message} [code {exc.error.code}])"
+        except Exception as exc:
+            logger.exception(
+                "MCP prompt '{}' failed: {}: {}",
+                self._name,
+                type(exc).__name__,
+                exc,
+            )
+            return f"(MCP prompt call failed: {type(exc).__name__})"
+
+        parts: list[str] = []
+        for message in result.messages:
+            content = message.content
+            # content is a single ContentBlock (not a list) in MCP SDK >= 1.x
+            if isinstance(content, types.TextContent):
+                parts.append(content.text)
+            elif isinstance(content, list):
+                for block in content:
+                    if isinstance(block, types.TextContent):
+                        parts.append(block.text)
+                    else:
+                        parts.append(str(block))
+            else:
+                parts.append(str(content))
+        return "\n".join(parts) or "(no output)"
+
+
 async def connect_mcp_servers(
-    mcp_servers: dict, registry: ToolRegistry, stack: AsyncExitStack
-) -> None:
-    """Connect to configured MCP servers and register their tools."""
+    mcp_servers: dict, registry: ToolRegistry
+) -> dict[str, AsyncExitStack]:
+    """Connect to configured MCP servers and register their tools, resources, prompts.
+
+    Returns a dict mapping server name -> its dedicated AsyncExitStack.
+    Each server gets its own stack and runs in its own task to prevent
+    cancel scope conflicts when multiple MCP servers are configured.
+    """
     from mcp import ClientSession, StdioServerParameters
     from mcp.client.sse import sse_client
     from mcp.client.stdio import stdio_client
     from mcp.client.streamable_http import streamable_http_client
 
-    for name, cfg in mcp_servers.items():
+    async def connect_single_server(name: str, cfg) -> tuple[str, AsyncExitStack | None]:
+        server_stack = AsyncExitStack()
+        await server_stack.__aenter__()
+
         try:
             transport_type = cfg.type
             if not transport_type:
                 if cfg.command:
                     transport_type = "stdio"
                 elif cfg.url:
-                    # Convention: URLs ending with /sse use SSE transport; others use streamableHttp
                     transport_type = (
                         "sse" if cfg.url.rstrip("/").endswith("/sse") else "streamableHttp"
                     )
                 else:
                     logger.warning("MCP server '{}': no command or url configured, skipping", name)
-                    continue
+                    await server_stack.aclose()
+                    return name, None
 
             if transport_type == "stdio":
                 params = StdioServerParameters(
                     command=cfg.command, args=cfg.args, env=cfg.env or None
                 )
-                read, write = await stack.enter_async_context(stdio_client(params))
+                read, write = await server_stack.enter_async_context(stdio_client(params))
             elif transport_type == "sse":
+
                 def httpx_client_factory(
                     headers: dict[str, str] | None = None,
                     timeout: httpx.Timeout | None = None,
@@ -182,27 +358,26 @@ async def connect_mcp_servers(
                         auth=auth,
                     )
 
-                read, write = await stack.enter_async_context(
+                read, write = await server_stack.enter_async_context(
                     sse_client(cfg.url, httpx_client_factory=httpx_client_factory)
                 )
             elif transport_type == "streamableHttp":
-                # Always provide an explicit httpx client so MCP HTTP transport does not
-                # inherit httpx's default 5s timeout and preempt the higher-level tool timeout.
-                http_client = await stack.enter_async_context(
+                http_client = await server_stack.enter_async_context(
                     httpx.AsyncClient(
                         headers=cfg.headers or None,
                         follow_redirects=True,
                         timeout=None,
                     )
                 )
-                read, write, _ = await stack.enter_async_context(
+                read, write, _ = await server_stack.enter_async_context(
                     streamable_http_client(cfg.url, http_client=http_client)
                 )
             else:
                 logger.warning("MCP server '{}': unknown transport type '{}'", name, transport_type)
-                continue
+                await server_stack.aclose()
+                return name, None
 
-            session = await stack.enter_async_context(ClientSession(read, write))
+            session = await server_stack.enter_async_context(ClientSession(read, write))
             await session.initialize()
 
             tools = await session.list_tools()
@@ -247,6 +422,76 @@ async def connect_mcp_servers(
                         ", ".join(available_wrapped_names) or "(none)",
                     )
 
-            logger.info("MCP server '{}': connected, {} tools registered", name, registered_count)
+            try:
+                resources_result = await session.list_resources()
+                for resource in resources_result.resources:
+                    wrapper = MCPResourceWrapper(
+                        session, name, resource, resource_timeout=cfg.tool_timeout
+                    )
+                    registry.register(wrapper)
+                    registered_count += 1
+                    logger.debug(
+                        "MCP: registered resource '{}' from server '{}'", wrapper.name, name
+                    )
+            except Exception as e:
+                logger.debug("MCP server '{}': resources not supported or failed: {}", name, e)
+
+            try:
+                prompts_result = await session.list_prompts()
+                for prompt in prompts_result.prompts:
+                    wrapper = MCPPromptWrapper(
+                        session, name, prompt, prompt_timeout=cfg.tool_timeout
+                    )
+                    registry.register(wrapper)
+                    registered_count += 1
+                    logger.debug("MCP: registered prompt '{}' from server '{}'", wrapper.name, name)
+            except Exception as e:
+                logger.debug("MCP server '{}': prompts not supported or failed: {}", name, e)
+
+            logger.info(
+                "MCP server '{}': connected, {} capabilities registered", name, registered_count
+            )
+            return name, server_stack
+
         except Exception as e:
-            logger.error("MCP server '{}': failed to connect: {}", name, e)
+            hint = ""
+            text = str(e).lower()
+            if any(
+                marker in text
+                for marker in (
+                    "parse error",
+                    "invalid json",
+                    "unexpected token",
+                    "jsonrpc",
+                    "content-length",
+                )
+            ):
+                hint = (
+                    " Hint: this looks like stdio protocol pollution. Make sure the MCP server writes "
+                    "only JSON-RPC to stdout and sends logs/debug output to stderr instead."
+                )
+            logger.error("MCP server '{}': failed to connect: {}{}", name, e, hint)
+            try:
+                await server_stack.aclose()
+            except Exception:
+                pass
+            return name, None
+
+    server_stacks: dict[str, AsyncExitStack] = {}
+
+    tasks: list[asyncio.Task] = []
+    for name, cfg in mcp_servers.items():
+        task = asyncio.create_task(connect_single_server(name, cfg))
+        tasks.append(task)
+
+    results = await asyncio.gather(*tasks, return_exceptions=True)
+
+    for i, result in enumerate(results):
+        name = list(mcp_servers.keys())[i]
+        if isinstance(result, BaseException):
+            if not isinstance(result, asyncio.CancelledError):
+                logger.error("MCP server '{}' connection task failed: {}", name, result)
+        elif result is not None and result[1] is not None:
+            server_stacks[result[0]] = result[1]
+
+    return server_stacks

nanobot/agent/tools/notebook.py

@@ -0,0 +1,161 @@
+"""NotebookEditTool — edit Jupyter .ipynb notebooks."""
+
+from __future__ import annotations
+
+import json
+import uuid
+from typing import Any
+
+from nanobot.agent.tools.base import tool_parameters
+from nanobot.agent.tools.schema import IntegerSchema, StringSchema, tool_parameters_schema
+from nanobot.agent.tools.filesystem import _FsTool
+
+
+def _new_cell(source: str, cell_type: str = "code", generate_id: bool = False) -> dict:
+    cell: dict[str, Any] = {
+        "cell_type": cell_type,
+        "source": source,
+        "metadata": {},
+    }
+    if cell_type == "code":
+        cell["outputs"] = []
+        cell["execution_count"] = None
+    if generate_id:
+        cell["id"] = uuid.uuid4().hex[:8]
+    return cell
+
+
+def _make_empty_notebook() -> dict:
+    return {
+        "nbformat": 4,
+        "nbformat_minor": 5,
+        "metadata": {
+            "kernelspec": {"display_name": "Python 3", "language": "python", "name": "python3"},
+            "language_info": {"name": "python"},
+        },
+        "cells": [],
+    }
+
+
+@tool_parameters(
+    tool_parameters_schema(
+        path=StringSchema("Path to the .ipynb notebook file"),
+        cell_index=IntegerSchema(0, description="0-based index of the cell to edit", minimum=0),
+        new_source=StringSchema("New source content for the cell"),
+        cell_type=StringSchema(
+            "Cell type: 'code' or 'markdown' (default: code)",
+            enum=["code", "markdown"],
+        ),
+        edit_mode=StringSchema(
+            "Mode: 'replace' (default), 'insert' (after target), or 'delete'",
+            enum=["replace", "insert", "delete"],
+        ),
+        required=["path", "cell_index"],
+    )
+)
+class NotebookEditTool(_FsTool):
+    """Edit Jupyter notebook cells: replace, insert, or delete."""
+
+    _VALID_CELL_TYPES = frozenset({"code", "markdown"})
+    _VALID_EDIT_MODES = frozenset({"replace", "insert", "delete"})
+
+    @property
+    def name(self) -> str:
+        return "notebook_edit"
+
+    @property
+    def description(self) -> str:
+        return (
+            "Edit a Jupyter notebook (.ipynb) cell. "
+            "Modes: replace (default) replaces cell content, "
+            "insert adds a new cell after the target index, "
+            "delete removes the cell at the index. "
+            "cell_index is 0-based."
+        )
+
+    async def execute(
+        self,
+        path: str | None = None,
+        cell_index: int = 0,
+        new_source: str = "",
+        cell_type: str = "code",
+        edit_mode: str = "replace",
+        **kwargs: Any,
+    ) -> str:
+        try:
+            if not path:
+                return "Error: path is required"
+
+            if not path.endswith(".ipynb"):
+                return "Error: notebook_edit only works on .ipynb files. Use edit_file for other files."
+
+            if edit_mode not in self._VALID_EDIT_MODES:
+                return (
+                    f"Error: Invalid edit_mode '{edit_mode}'. "
+                    "Use one of: replace, insert, delete."
+                )
+
+            if cell_type not in self._VALID_CELL_TYPES:
+                return (
+                    f"Error: Invalid cell_type '{cell_type}'. "
+                    "Use one of: code, markdown."
+                )
+
+            fp = self._resolve(path)
+
+            # Create new notebook if file doesn't exist and mode is insert
+            if not fp.exists():
+                if edit_mode != "insert":
+                    return f"Error: File not found: {path}"
+                nb = _make_empty_notebook()
+                cell = _new_cell(new_source, cell_type, generate_id=True)
+                nb["cells"].append(cell)
+                fp.parent.mkdir(parents=True, exist_ok=True)
+                fp.write_text(json.dumps(nb, indent=1, ensure_ascii=False), encoding="utf-8")
+                return f"Successfully created {fp} with 1 cell"
+
+            try:
+                nb = json.loads(fp.read_text(encoding="utf-8"))
+            except (json.JSONDecodeError, UnicodeDecodeError) as e:
+                return f"Error: Failed to parse notebook: {e}"
+
+            cells = nb.get("cells", [])
+            nbformat_minor = nb.get("nbformat_minor", 0)
+            generate_id = nb.get("nbformat", 0) >= 4 and nbformat_minor >= 5
+
+            if edit_mode == "delete":
+                if cell_index < 0 or cell_index >= len(cells):
+                    return f"Error: cell_index {cell_index} out of range (notebook has {len(cells)} cells)"
+                cells.pop(cell_index)
+                nb["cells"] = cells
+                fp.write_text(json.dumps(nb, indent=1, ensure_ascii=False), encoding="utf-8")
+                return f"Successfully deleted cell {cell_index} from {fp}"
+
+            if edit_mode == "insert":
+                insert_at = min(cell_index + 1, len(cells))
+                cell = _new_cell(new_source, cell_type, generate_id=generate_id)
+                cells.insert(insert_at, cell)
+                nb["cells"] = cells
+                fp.write_text(json.dumps(nb, indent=1, ensure_ascii=False), encoding="utf-8")
+                return f"Successfully inserted cell at index {insert_at} in {fp}"
+
+            # Default: replace
+            if cell_index < 0 or cell_index >= len(cells):
+                return f"Error: cell_index {cell_index} out of range (notebook has {len(cells)} cells)"
+            cells[cell_index]["source"] = new_source
+            if cell_type and cells[cell_index].get("cell_type") != cell_type:
+                cells[cell_index]["cell_type"] = cell_type
+                if cell_type == "code":
+                    cells[cell_index].setdefault("outputs", [])
+                    cells[cell_index].setdefault("execution_count", None)
+                elif "outputs" in cells[cell_index]:
+                    del cells[cell_index]["outputs"]
+                    cells[cell_index].pop("execution_count", None)
+            nb["cells"] = cells
+            fp.write_text(json.dumps(nb, indent=1, ensure_ascii=False), encoding="utf-8")
+            return f"Successfully edited cell {cell_index} in {fp}"
+
+        except PermissionError as e:
+            return f"Error: {e}"
+        except Exception as e:
+            return f"Error editing notebook: {e}"

nanobot/agent/tools/registry.py

@@ -14,14 +14,17 @@ class ToolRegistry:
 
     def __init__(self):
         self._tools: dict[str, Tool] = {}
+        self._cached_definitions: list[dict[str, Any]] | None = None
 
     def register(self, tool: Tool) -> None:
         """Register a tool."""
         self._tools[tool.name] = tool
+        self._cached_definitions = None
 
     def unregister(self, name: str) -> None:
         """Unregister a tool by name."""
         self._tools.pop(name, None)
+        self._cached_definitions = None
 
     def get(self, name: str) -> Tool | None:
         """Get a tool by name."""
@@ -46,8 +49,12 @@ class ToolRegistry:
         """Get tool definitions with stable ordering for cache-friendly prompts.
 
         Built-in tools are sorted first as a stable prefix, then MCP tools are
-        sorted and appended.
+        sorted and appended.  The result is cached until the next
+        register/unregister call.
         """
+        if self._cached_definitions is not None:
+            return self._cached_definitions
+
         definitions = [tool.to_schema() for tool in self._tools.values()]
         builtins: list[dict[str, Any]] = []
         mcp_tools: list[dict[str, Any]] = []
@@ -60,7 +67,8 @@ class ToolRegistry:
 
         builtins.sort(key=self._schema_name)
         mcp_tools.sort(key=self._schema_name)
-        return builtins + mcp_tools
+        self._cached_definitions = builtins + mcp_tools
+        return self._cached_definitions
 
     def prepare_call(
         self,
@@ -68,6 +76,13 @@ class ToolRegistry:
         params: dict[str, Any],
     ) -> tuple[Tool | None, dict[str, Any], str | None]:
         """Resolve, cast, and validate one tool call."""
+        # Guard against invalid parameter types (e.g., list instead of dict)
+        if not isinstance(params, dict) and name in ('write_file', 'read_file'):
+            return None, params, (
+                f"Error: Tool '{name}' parameters must be a JSON object, got {type(params).__name__}. "
+                "Use named parameters: tool_name(param1=\"value1\", param2=\"value2\")"
+            )
+
         tool = self._tools.get(name)
         if not tool:
             return None, params, (

nanobot/agent/tools/search.py

@@ -142,8 +142,9 @@ class GlobTool(_SearchTool):
     @property
     def description(self) -> str:
         return (
-            "Find files matching a glob pattern. "
-            "Simple patterns like '*.py' match by filename recursively."
+            "Find files matching a glob pattern (e.g. '*.py', 'tests/**/test_*.py'). "
+            "Results are sorted by modification time (newest first). "
+            "Skips .git, node_modules, __pycache__, and other noise directories."
         )
 
     @property
@@ -261,9 +262,10 @@ class GrepTool(_SearchTool):
     @property
     def description(self) -> str:
         return (
-            "Search file contents with a regex-like pattern. "
-            "Supports optional glob filtering, structured output modes, "
-            "type filters, pagination, and surrounding context lines."
+            "Search file contents with a regex pattern. "
+            "Default output_mode is files_with_matches (file paths only); "
+            "use content mode for matching lines with context. "
+            "Skips binary and files >2 MB. Supports glob/type filtering."
         )
 
     @property

nanobot/agent/tools/self.py

@@ -0,0 +1,439 @@
+"""MyTool: runtime state inspection and configuration for the agent loop."""
+
+from __future__ import annotations
+
+import time
+from typing import TYPE_CHECKING, Any
+
+from loguru import logger
+
+from nanobot.agent.subagent import SubagentStatus
+from nanobot.agent.tools.base import Tool
+
+if TYPE_CHECKING:
+    from nanobot.agent.loop import AgentLoop
+
+
+def _has_real_attr(obj: Any, key: str) -> bool:
+    """Check if obj has a real (explicitly set) attribute, not auto-generated by mock."""
+    if isinstance(obj, dict):
+        return key in obj
+    d = getattr(obj, "__dict__", None)
+    if d is not None and key in d:
+        return True
+    for cls in type(obj).__mro__:
+        if key in cls.__dict__:
+            return True
+    return False
+
+
+class MyTool(Tool):
+    """Check and set the agent loop's runtime configuration."""
+
+    BLOCKED = frozenset({
+        # Core infrastructure
+        "bus", "provider", "_running", "tools",
+        # Config management
+        "_runtime_vars",
+        # Subsystems
+        "runner", "sessions", "consolidator",
+        "dream", "auto_compact", "context", "commands",
+        # Sensitive runtime state (credentials, message routing, task tracking)
+        "_mcp_servers", "_mcp_stacks", "_pending_queues",
+        "_session_locks", "_active_tasks", "_background_tasks",
+        # Security boundaries (inspect + modify both blocked)
+        "restrict_to_workspace", "channels_config",
+        "_concurrency_gate", "_unified_session", "_extra_hooks",
+    })
+
+    READ_ONLY = frozenset({
+        "subagents",  # observable but replacing it would break the system
+        "_current_iteration",  # updated by runner only
+        "exec_config",  # inspect allowed (e.g. check sandbox), modify blocked
+        "web_config",  # inspect allowed (e.g. check enable), modify blocked
+    })
+
+    _DENIED_ATTRS = frozenset({
+        "__class__", "__dict__", "__bases__", "__subclasses__", "__mro__",
+        "__init__", "__new__", "__reduce__", "__getstate__", "__setstate__",
+        "__del__", "__call__", "__getattr__", "__setattr__", "__delattr__",
+        "__code__", "__globals__", "func_globals", "func_code",
+        "__wrapped__", "__closure__",
+    })
+
+    # Sub-field names that are sensitive regardless of parent path
+    _SENSITIVE_NAMES = frozenset({
+        "api_key", "secret", "password", "token", "credential",
+        "private_key", "access_token", "refresh_token", "auth",
+    })
+
+    RESTRICTED: dict[str, dict[str, Any]] = {
+        "max_iterations":        {"type": int, "min": 1,   "max": 100},
+        "context_window_tokens": {"type": int, "min": 4096, "max": 1_000_000},
+        "model":                 {"type": str, "min_len": 1},
+    }
+
+    _MAX_RUNTIME_KEYS = 64
+
+    def __init__(self, loop: AgentLoop, modify_allowed: bool = True) -> None:
+        self._loop = loop
+        self._modify_allowed = modify_allowed
+        self._channel = ""
+        self._chat_id = ""
+
+    def __deepcopy__(self, memo: dict[int, Any]) -> MyTool:
+        cls = self.__class__
+        result = cls.__new__(cls)
+        memo[id(self)] = result
+        result._loop = self._loop
+        result._modify_allowed = self._modify_allowed
+        result._channel = self._channel
+        result._chat_id = self._chat_id
+        return result
+
+    def set_context(self, channel: str, chat_id: str) -> None:
+        self._channel = channel
+        self._chat_id = chat_id
+
+    @property
+    def name(self) -> str:
+        return "my"
+
+    @property
+    def description(self) -> str:
+        base = (
+            "Check and set your own runtime state.\n"
+            "Actions: check, set.\n"
+            "- check (no key): full config overview — start here.\n"
+            "- check (key): drill into a value. Dot-paths allowed "
+            "(e.g. '_last_usage.prompt_tokens', 'web_config.enable').\n"
+            "- set (key, value): change config or store notes in your scratchpad. "
+            "Scratchpad keys persist across turns but not restarts.\n"
+            "Key values: _current_iteration (current progress), "
+            "max_iterations - _current_iteration = remaining iterations.\n"
+            "Note: web_config and exec_config are readable but read-only.\n"
+            "\n"
+            "When to use:\n"
+            "- User asks about your model, settings, or token usage → check that key.\n"
+            "- A tool fails or behaves unexpectedly → check the related config to diagnose.\n"
+            "- User asks you to remember a preference for this session → set to store it in your scratchpad.\n"
+            "- About to start a large task → check context_window_tokens and max_iterations first."
+        )
+        if not self._modify_allowed:
+            base += "\nREAD-ONLY MODE: set is disabled."
+        else:
+            base += (
+                "\nIMPORTANT: Before setting state, predict the potential impact. "
+                "If the operation could cause crashes or instability "
+                "(e.g. changing model), warn the user first."
+            )
+        return base
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "action": {
+                    "type": "string",
+                    "enum": ["check", "set"],
+                    "description": "Action to perform",
+                },
+                "key": {
+                    "type": "string",
+                    "description": "Dot-path for check/set. Examples: 'max_iterations', 'workspace', 'provider_retry_mode'. "
+                    "For check without key, shows all config values.",
+                },
+                "value": {"description": "New value (for set). Type must match target (int for max_iterations/context_window_tokens, str for model)."},
+            },
+            "required": ["action"],
+        }
+
+    def _audit(self, action: str, detail: str) -> None:
+        session = f"{self._channel}:{self._chat_id}" if self._channel else "unknown"
+        logger.info("self.{} | {} | session:{}", action, detail, session)
+
+    # ------------------------------------------------------------------
+    # Path resolution
+    # ------------------------------------------------------------------
+
+    def _resolve_path(self, path: str) -> tuple[Any, str | None]:
+        parts = path.split(".")
+        obj = self._loop
+        for part in parts:
+            if part in self._DENIED_ATTRS or part.startswith("__"):
+                return None, f"'{part}' is not accessible"
+            if part in self.BLOCKED:
+                return None, f"'{part}' is not accessible"
+            if part.lower() in self._SENSITIVE_NAMES:
+                return None, f"'{part}' is not accessible"
+            try:
+                if isinstance(obj, dict):
+                    if part in obj:
+                        obj = obj[part]
+                    else:
+                        return None, f"'{part}' not found in dict"
+                else:
+                    obj = getattr(obj, part)
+            except (KeyError, AttributeError) as e:
+                return None, f"'{part}' not found: {e}"
+        return obj, None
+
+    @staticmethod
+    def _validate_key(key: str | None, label: str = "key") -> str | None:
+        if not key or not key.strip():
+            return f"Error: '{label}' cannot be empty or whitespace"
+        return None
+
+    # ------------------------------------------------------------------
+    # Smart formatting
+    # ------------------------------------------------------------------
+
+    @staticmethod
+    def _format_status(st: SubagentStatus, indent: str = "  ") -> str:
+        elapsed = time.monotonic() - st.started_at
+        tool_summary = ", ".join(
+            f"{e.get('name', '?')}({e.get('status', '?')})" for e in st.tool_events[-5:]
+        ) or "none"
+        lines = [
+            f"{indent}phase: {st.phase}, iteration: {st.iteration}, elapsed: {elapsed:.1f}s",
+            f"{indent}tools: {tool_summary}",
+            f"{indent}usage: {st.usage or 'n/a'}",
+        ]
+        if st.error:
+            lines.append(f"{indent}error: {st.error}")
+        if st.stop_reason:
+            lines.append(f"{indent}stop_reason: {st.stop_reason}")
+        return "\n".join(lines)
+
+    @staticmethod
+    def _format_value(val: Any, key: str = "") -> str:
+        if isinstance(val, SubagentStatus):
+            header = f"Subagent [{val.task_id}] '{val.label}'"
+            detail = MyTool._format_status(val, "  ")
+            return f"{header}\n  task: {val.task_description}\n{detail}"
+        # SubagentManager: delegate to its _task_statuses dict
+        if hasattr(val, "_task_statuses") and isinstance(val._task_statuses, dict):
+            return MyTool._format_value(val._task_statuses, key)
+        if isinstance(val, dict) and val and isinstance(next(iter(val.values())), SubagentStatus):
+            prefix = f"{key}: " if key else ""
+            lines = [f"{prefix}{len(val)} subagent(s):"]
+            for tid, st in val.items():
+                detail = MyTool._format_status(st, "    ")
+                lines.append(f"  [{tid}] '{st.label}'\n{detail}")
+            return "\n".join(lines)
+        if hasattr(val, "tool_names"):
+            return f"tools: {len(val.tool_names)} registered — {val.tool_names}"
+        # Scalar types — repr is fine
+        if isinstance(val, (str, int, float, bool, type(None))):
+            r = repr(val)
+            return f"{key}: {r}" if key else r
+        # Dict — small: show content; large: show keys for dot-path navigation
+        if isinstance(val, dict):
+            ks = list(val.keys())
+            if not ks:
+                return f"{key}: {{}}" if key else "{}"
+            if len(ks) <= 5:
+                r = repr(val)
+                if len(r) <= 200:
+                    return f"{key}: {r}" if key else r
+            preview = ", ".join(str(k) for k in ks[:15])
+            suffix = ", ..." if len(ks) > 15 else ""
+            return f"{key}: {{{preview}{suffix}}}" if key else f"{{{preview}{suffix}}}"
+        # List/tuple — count for large, repr for small
+        if isinstance(val, (list, tuple)):
+            if len(val) > 20:
+                return f"{key}: [{len(val)} items]" if key else f"[{len(val)} items]"
+            r = repr(val)
+            return f"{key}: {r}" if key else r
+        # Complex object — small Pydantic models: show values; others: show field names for navigation
+        cls_name = type(val).__name__
+        if hasattr(val, "model_fields"):
+            fields = list(val.model_fields.keys())
+            if len(fields) <= 8:
+                # Small config objects: show field=value pairs
+                pairs = []
+                for f in fields:
+                    fv = getattr(val, f, "?")
+                    if isinstance(fv, (str, int, float, bool, type(None))):
+                        pairs.append(f"{f}={fv!r}")
+                    else:
+                        pairs.append(f"{f}=<{type(fv).__name__}>")
+                preview = ", ".join(pairs)
+                return f"{key}: {preview}" if key else preview
+        else:
+            fields = [a for a in getattr(val, "__dict__", {}) if not a.startswith("__")]
+        if fields:
+            preview = ", ".join(str(f) for f in fields[:20])
+            suffix = ", ..." if len(fields) > 20 else ""
+            return f"{key}: <{cls_name}> [{preview}{suffix}]" if key else f"<{cls_name}> [{preview}{suffix}]"
+        r = repr(val)
+        return f"{key}: {r}" if key else r
+
+    # ------------------------------------------------------------------
+    # Action dispatch
+    # ------------------------------------------------------------------
+
+    async def execute(
+        self,
+        action: str,
+        key: str | None = None,
+        value: Any = None,
+        **_kwargs: Any,
+    ) -> str:
+        if action in ("inspect", "check"):
+            return self._inspect(key)
+        if not self._modify_allowed:
+            return "Error: set is disabled (tools.my.allow_set is false)"
+        if action in ("modify", "set"):
+            return self._modify(key, value)
+        return f"Unknown action: {action}"
+
+    # -- inspect --
+
+    def _inspect(self, key: str | None) -> str:
+        if not key:
+            return self._inspect_all()
+        top = key.split(".")[0]
+        if top in self._DENIED_ATTRS or top.startswith("__"):
+            return f"Error: '{top}' is not accessible"
+        obj, err = self._resolve_path(key)
+        if err:
+            # "scratchpad" alias for _runtime_vars
+            if key == "scratchpad":
+                rv = self._loop._runtime_vars
+                return self._format_value(rv, "scratchpad") if rv else "scratchpad is empty"
+            # Fallback: check _runtime_vars for simple keys stored by modify
+            if "." not in key and key in self._loop._runtime_vars:
+                return self._format_value(self._loop._runtime_vars[key], key)
+            return f"Error: {err}"
+        # Guard against mock auto-generated attributes
+        if "." not in key and not _has_real_attr(self._loop, key):
+            if key in self._loop._runtime_vars:
+                return self._format_value(self._loop._runtime_vars[key], key)
+            return f"Error: '{key}' not found"
+        return self._format_value(obj, key)
+
+    def _inspect_all(self) -> str:
+        loop = self._loop
+        parts: list[str] = []
+        # RESTRICTED keys
+        for k in self.RESTRICTED:
+            parts.append(self._format_value(getattr(loop, k, None), k))
+        # Other useful top-level keys shown in description
+        for k in ("workspace", "provider_retry_mode", "max_tool_result_chars", "_current_iteration", "web_config", "exec_config", "subagents"):
+            if _has_real_attr(loop, k):
+                parts.append(self._format_value(getattr(loop, k, None), k))
+        # Token usage
+        usage = loop._last_usage
+        if usage:
+            parts.append(self._format_value(usage, "_last_usage"))
+        rv = loop._runtime_vars
+        if rv:
+            parts.append(self._format_value(rv, "scratchpad"))
+        return "\n".join(parts)
+
+    # -- modify --
+
+    def _modify(self, key: str | None, value: Any) -> str:
+        if err := self._validate_key(key):
+            return err
+        top = key.split(".")[0]
+        if top in self.BLOCKED or top in self._DENIED_ATTRS or top.startswith("__") or top.lower() in self._SENSITIVE_NAMES:
+            self._audit("modify", f"BLOCKED {key}")
+            return f"Error: '{key}' is protected and cannot be modified"
+        if top in self.READ_ONLY:
+            self._audit("modify", f"READ_ONLY {key}")
+            return f"Error: '{key}' is read-only and cannot be modified"
+        if "." in key:
+            parent_path, leaf = key.rsplit(".", 1)
+            if leaf in self._DENIED_ATTRS or leaf.startswith("__"):
+                self._audit("modify", f"BLOCKED leaf '{leaf}'")
+                return f"Error: '{leaf}' is not accessible"
+            if leaf.lower() in self._SENSITIVE_NAMES:
+                self._audit("modify", f"BLOCKED sensitive leaf '{leaf}'")
+                return f"Error: '{leaf}' is not accessible"
+            parent, err = self._resolve_path(parent_path)
+            if err:
+                return f"Error: {err}"
+            if isinstance(parent, dict):
+                parent[leaf] = value
+            else:
+                setattr(parent, leaf, value)
+            self._audit("modify", f"{key} = {value!r}")
+            return f"Set {key} = {value!r}"
+        if key in self.RESTRICTED:
+            return self._modify_restricted(key, value)
+        return self._modify_free(key, value)
+
+    def _modify_restricted(self, key: str, value: Any) -> str:
+        spec = self.RESTRICTED[key]
+        expected = spec["type"]
+        if expected is int and isinstance(value, bool):
+            return f"Error: '{key}' must be {expected.__name__}, got bool"
+        if not isinstance(value, expected):
+            try:
+                value = expected(value)
+            except (ValueError, TypeError):
+                return f"Error: '{key}' must be {expected.__name__}, got {type(value).__name__}"
+        old = getattr(self._loop, key)
+        if "min" in spec and value < spec["min"]:
+            return f"Error: '{key}' must be >= {spec['min']}"
+        if "max" in spec and value > spec["max"]:
+            return f"Error: '{key}' must be <= {spec['max']}"
+        if "min_len" in spec and len(str(value)) < spec["min_len"]:
+            return f"Error: '{key}' must be at least {spec['min_len']} characters"
+        setattr(self._loop, key, value)
+        self._audit("modify", f"{key}: {old!r} -> {value!r}")
+        return f"Set {key} = {value!r} (was {old!r})"
+
+    def _modify_free(self, key: str, value: Any) -> str:
+        if _has_real_attr(self._loop, key):
+            old = getattr(self._loop, key)
+            if isinstance(old, (str, int, float, bool)):
+                old_t, new_t = type(old), type(value)
+                if old_t is float and new_t is int:
+                    pass  # int → float coercion allowed
+                elif old_t is not new_t:
+                    self._audit(
+                        "modify",
+                        f"REJECTED type mismatch {key}: expects {old_t.__name__}, got {new_t.__name__}",
+                    )
+                    return f"Error: '{key}' expects {old_t.__name__}, got {new_t.__name__}"
+            setattr(self._loop, key, value)
+            self._audit("modify", f"{key}: {old!r} -> {value!r}")
+            return f"Set {key} = {value!r} (was {old!r})"
+        if callable(value):
+            self._audit("modify", f"REJECTED callable {key}")
+            return "Error: cannot store callable values"
+        err = self._validate_json_safe(value)
+        if err:
+            self._audit("modify", f"REJECTED {key}: {err}")
+            return f"Error: {err}"
+        if key not in self._loop._runtime_vars and len(self._loop._runtime_vars) >= self._MAX_RUNTIME_KEYS:
+            self._audit("modify", f"REJECTED {key}: max keys ({self._MAX_RUNTIME_KEYS}) reached")
+            return f"Error: scratchpad is full (max {self._MAX_RUNTIME_KEYS} keys). Remove unused keys first."
+        old = self._loop._runtime_vars.get(key)
+        self._loop._runtime_vars[key] = value
+        self._audit("modify", f"scratchpad.{key}: {old!r} -> {value!r}")
+        return f"Set scratchpad.{key} = {value!r}"
+
+    @classmethod
+    def _validate_json_safe(cls, value: Any, depth: int = 0) -> str | None:
+        if depth > 10:
+            return "value nesting too deep (max 10 levels)"
+        if isinstance(value, (str, int, float, bool, type(None))):
+            return None
+        if isinstance(value, list):
+            for i, item in enumerate(value):
+                if err := cls._validate_json_safe(item, depth + 1):
+                    return f"list[{i}] contains {err}"
+            return None
+        if isinstance(value, dict):
+            for k, v in value.items():
+                if not isinstance(k, str):
+                    return f"dict key must be str, got {type(k).__name__}"
+                if err := cls._validate_json_safe(v, depth + 1):
+                    return f"dict key '{k}' contains {err}"
+            return None
+        return f"unsupported type {type(value).__name__}"

nanobot/agent/tools/shell.py

@@ -46,6 +46,7 @@ class ExecTool(Tool):
         restrict_to_workspace: bool = False,
         sandbox: str = "",
         path_append: str = "",
+        allowed_env_keys: list[str] | None = None,
     ):
         self.timeout = timeout
         self.working_dir = working_dir
@@ -60,10 +61,19 @@ class ExecTool(Tool):
             r">\s*/dev/sd",                  # write to disk
             r"\b(shutdown|reboot|poweroff)\b",  # system power
             r":\(\)\s*\{.*\};\s*:",          # fork bomb
+            # Block writes to nanobot internal state files (#2989).
+            # history.jsonl / .dream_cursor are managed by append_history();
+            # direct writes corrupt the cursor format and crash /dream.
+            r">>?\s*\S*(?:history\.jsonl|\.dream_cursor)",            # > / >> redirect
+            r"\btee\b[^|;&<>]*(?:history\.jsonl|\.dream_cursor)",     # tee / tee -a
+            r"\b(?:cp|mv)\b(?:\s+[^\s|;&<>]+)+\s+\S*(?:history\.jsonl|\.dream_cursor)",  # cp/mv target
+            r"\bdd\b[^|;&<>]*\bof=\S*(?:history\.jsonl|\.dream_cursor)",  # dd of=
+            r"\bsed\s+-i[^|;&<>]*(?:history\.jsonl|\.dream_cursor)",  # sed -i
         ]
         self.allow_patterns = allow_patterns or []
         self.restrict_to_workspace = restrict_to_workspace
         self.path_append = path_append
+        self.allowed_env_keys = allowed_env_keys or []
 
     @property
     def name(self) -> str:
@@ -74,7 +84,13 @@ class ExecTool(Tool):
 
     @property
     def description(self) -> str:
-        return "Execute a shell command and return its output. Use with caution."
+        return (
+            "Execute a shell command and return its output. "
+            "Prefer read_file/write_file/edit_file over cat/echo/sed, "
+            "and grep/glob over shell find/grep. "
+            "Use -y or --yes flags to avoid interactive prompts. "
+            "Output is truncated at 10 000 chars; timeout defaults to 60s."
+        )
 
     @property
     def exclusive(self) -> bool:
@@ -85,6 +101,21 @@ class ExecTool(Tool):
         timeout: int | None = None, **kwargs: Any,
     ) -> str:
         cwd = working_dir or self.working_dir or os.getcwd()
+
+        # Prevent an LLM-supplied working_dir from escaping the configured
+        # workspace when restrict_to_workspace is enabled (#2826). Without
+        # this, a caller can pass working_dir="/etc" and then all absolute
+        # paths under /etc would pass the _guard_command check that anchors
+        # on cwd.
+        if self.restrict_to_workspace and self.working_dir:
+            try:
+                requested = Path(cwd).expanduser().resolve()
+                workspace_root = Path(self.working_dir).expanduser().resolve()
+            except Exception:
+                return "Error: working_dir could not be resolved"
+            if requested != workspace_root and workspace_root not in requested.parents:
+                return "Error: working_dir is outside the configured workspace"
+
         guard_error = self._guard_command(command, cwd)
         if guard_error:
             return guard_error
@@ -202,7 +233,7 @@ class ExecTool(Tool):
         """
         if _IS_WINDOWS:
             sr = os.environ.get("SYSTEMROOT", r"C:\Windows")
-            return {
+            env = {
                 "SYSTEMROOT": sr,
                 "COMSPEC": os.environ.get("COMSPEC", f"{sr}\\system32\\cmd.exe"),
                 "USERPROFILE": os.environ.get("USERPROFILE", ""),
@@ -212,13 +243,29 @@ class ExecTool(Tool):
                 "TMP": os.environ.get("TMP", f"{sr}\\Temp"),
                 "PATHEXT": os.environ.get("PATHEXT", ".COM;.EXE;.BAT;.CMD"),
                 "PATH": os.environ.get("PATH", f"{sr}\\system32;{sr}"),
+                "APPDATA": os.environ.get("APPDATA", ""),
+                "LOCALAPPDATA": os.environ.get("LOCALAPPDATA", ""),
+                "ProgramData": os.environ.get("ProgramData", ""),
+                "ProgramFiles": os.environ.get("ProgramFiles", ""),
+                "ProgramFiles(x86)": os.environ.get("ProgramFiles(x86)", ""),
+                "ProgramW6432": os.environ.get("ProgramW6432", ""),
             }
+            for key in self.allowed_env_keys:
+                val = os.environ.get(key)
+                if val is not None:
+                    env[key] = val
+            return env
         home = os.environ.get("HOME", "/tmp")
-        return {
+        env = {
             "HOME": home,
             "LANG": os.environ.get("LANG", "C.UTF-8"),
             "TERM": os.environ.get("TERM", "dumb"),
         }
+        for key in self.allowed_env_keys:
+            val = os.environ.get(key)
+            if val is not None:
+                env[key] = val
+        return env
 
     def _guard_command(self, command: str, cwd: str) -> str | None:
         """Best-effort safety guard for potentially destructive commands."""

nanobot/agent/tools/web.py

@@ -84,7 +84,11 @@ class WebSearchTool(Tool):
     """Search the web using configured provider."""
 
     name = "web_search"
-    description = "Search the web. Returns titles, URLs, and snippets."
+    description = (
+        "Search the web. Returns titles, URLs, and snippets. "
+        "count defaults to 5 (max 10). "
+        "Use web_fetch to read a specific page in full."
+    )
 
     def __init__(self, config: WebSearchConfig | None = None, proxy: str | None = None):
         from nanobot.config.schema import WebSearchConfig
@@ -92,10 +96,37 @@ class WebSearchTool(Tool):
         self.config = config if config is not None else WebSearchConfig()
         self.proxy = proxy
 
+    def _effective_provider(self) -> str:
+        """Resolve the backend that execute() will actually use."""
+        provider = self.config.provider.strip().lower() or "brave"
+        if provider == "duckduckgo":
+            return "duckduckgo"
+        if provider == "brave":
+            api_key = self.config.api_key or os.environ.get("BRAVE_API_KEY", "")
+            return "brave" if api_key else "duckduckgo"
+        if provider == "tavily":
+            api_key = self.config.api_key or os.environ.get("TAVILY_API_KEY", "")
+            return "tavily" if api_key else "duckduckgo"
+        if provider == "searxng":
+            base_url = (self.config.base_url or os.environ.get("SEARXNG_BASE_URL", "")).strip()
+            return "searxng" if base_url else "duckduckgo"
+        if provider == "jina":
+            api_key = self.config.api_key or os.environ.get("JINA_API_KEY", "")
+            return "jina" if api_key else "duckduckgo"
+        if provider == "kagi":
+            api_key = self.config.api_key or os.environ.get("KAGI_API_KEY", "")
+            return "kagi" if api_key else "duckduckgo"
+        return provider
+
     @property
     def read_only(self) -> bool:
         return True
 
+    @property
+    def exclusive(self) -> bool:
+        """DuckDuckGo searches are serialized because ddgs is not concurrency-safe."""
+        return self._effective_provider() == "duckduckgo"
+
     async def execute(self, query: str, count: int | None = None, **kwargs: Any) -> str:
         provider = self.config.provider.strip().lower() or "brave"
         n = min(max(count or self.config.max_results, 1), 10)
@@ -110,6 +141,8 @@ class WebSearchTool(Tool):
             return await self._search_jina(query, n)
         elif provider == "brave":
             return await self._search_brave(query, n)
+        elif provider == "kagi":
+            return await self._search_kagi(query, n)
         else:
             return f"Error: unknown search provider '{provider}'"
 
@@ -200,6 +233,29 @@ class WebSearchTool(Tool):
             logger.warning("Jina search failed ({}), falling back to DuckDuckGo", e)
             return await self._search_duckduckgo(query, n)
 
+    async def _search_kagi(self, query: str, n: int) -> str:
+        api_key = self.config.api_key or os.environ.get("KAGI_API_KEY", "")
+        if not api_key:
+            logger.warning("KAGI_API_KEY not set, falling back to DuckDuckGo")
+            return await self._search_duckduckgo(query, n)
+        try:
+            async with httpx.AsyncClient(proxy=self.proxy) as client:
+                r = await client.get(
+                    "https://kagi.com/api/v0/search",
+                    params={"q": query, "limit": n},
+                    headers={"Authorization": f"Bot {api_key}"},
+                    timeout=10.0,
+                )
+                r.raise_for_status()
+            # t=0 items are search results; other values are related searches, etc.
+            items = [
+                {"title": d.get("title", ""), "url": d.get("url", ""), "content": d.get("snippet", "")}
+                for d in r.json().get("data", []) if d.get("t") == 0
+            ]
+            return _format_results(query, items, n)
+        except Exception as e:
+            return f"Error: {e}"
+
     async def _search_duckduckgo(self, query: str, n: int) -> str:
         try:
             # Note: duckduckgo_search is synchronous and does its own requests
@@ -239,7 +295,11 @@ class WebFetchTool(Tool):
     """Fetch and extract content from a URL."""
 
     name = "web_fetch"
-    description = "Fetch URL and extract readable content (HTML → markdown/text)."
+    description = (
+        "Fetch a URL and extract readable content (HTML → markdown/text). "
+        "Output is capped at maxChars (default 50 000). "
+        "Works for most web pages and docs; may fail on login-walled or JS-heavy sites."
+    )
 
     def __init__(self, max_chars: int = 50000, proxy: str | None = None):
         self.max_chars = max_chars

nanobot/api/server.py

@@ -7,15 +7,30 @@ All requests route to a single persistent API session.
 from __future__ import annotations
 
 import asyncio
+import base64
+import json as _json
+import mimetypes
+import re
 import time
 import uuid
+from pathlib import Path
 from typing import Any
 
 from aiohttp import web
 from loguru import logger
 
+from nanobot.config.paths import get_media_dir
+from nanobot.utils.helpers import safe_filename
 from nanobot.utils.runtime import EMPTY_FINAL_RESPONSE_MESSAGE
 
+MAX_FILE_SIZE = 10 * 1024 * 1024  # 10 MB
+_DATA_URL_RE = re.compile(r"^data:([^;]+);base64,(.+)$", re.DOTALL)
+
+
+class _FileSizeExceeded(Exception):
+    """Raised when an uploaded file exceeds the size limit."""
+
+
 API_SESSION_KEY = "api:default"
 API_CHAT_ID = "default"
 
@@ -24,6 +39,7 @@ API_CHAT_ID = "default"
 # Response helpers
 # ---------------------------------------------------------------------------
 
+
 def _error_json(status: int, message: str, err_type: str = "invalid_request_error") -> web.Response:
     return web.json_response(
         {"error": {"message": message, "type": err_type, "code": status}},
@@ -56,50 +72,231 @@ def _response_text(value: Any) -> str:
         return str(getattr(value, "content") or "")
     return str(value)
 
+# ---------------------------------------------------------------------------
+# SSE helpers
+# ---------------------------------------------------------------------------
+
+
+def _sse_chunk(delta: str, model: str, chunk_id: str, finish_reason: str | None = None) -> bytes:
+    """Format a single OpenAI-compatible SSE chunk."""
+    payload = {
+        "id": chunk_id,
+        "object": "chat.completion.chunk",
+        "created": int(time.time()),
+        "model": model,
+        "choices": [
+            {
+                "index": 0,
+                "delta": {"content": delta} if delta else {},
+                "finish_reason": finish_reason,
+            }
+        ],
+    }
+    return f"data: {_json.dumps(payload)}\n\n".encode()
+
+
+_SSE_DONE = b"data: [DONE]\n\n"
+
+# ---------------------------------------------------------------------------
+# Upload helpers
+# ---------------------------------------------------------------------------
+
+
+def _save_base64_data_url(data_url: str, media_dir: Path) -> str | None:
+    """Decode a data:...;base64,... URL and save to disk."""
+    m = _DATA_URL_RE.match(data_url)
+    if not m:
+        return None
+    mime_type, b64_payload = m.group(1), m.group(2)
+    try:
+        raw = base64.b64decode(b64_payload)
+    except Exception:
+        return None
+    if len(raw) > MAX_FILE_SIZE:
+        raise _FileSizeExceeded(f"File exceeds {MAX_FILE_SIZE // (1024 * 1024)}MB limit")
+    ext = mimetypes.guess_extension(mime_type) or ".bin"
+    filename = f"{uuid.uuid4().hex[:12]}{ext}"
+    dest = media_dir / safe_filename(filename)
+    dest.write_bytes(raw)
+    return str(dest)
+
+
+def _parse_json_content(body: dict) -> tuple[str, list[str]]:
+    """Parse JSON request body. Returns (text, media_paths)."""
+    messages = body.get("messages")
+    if not isinstance(messages, list) or len(messages) != 1:
+        raise ValueError("Only a single user message is supported")
+    message = messages[0]
+    if not isinstance(message, dict) or message.get("role") != "user":
+        raise ValueError("Only a single user message is supported")
+
+    user_content = message.get("content", "")
+    media_dir = get_media_dir("api")
+    media_paths: list[str] = []
+
+    if isinstance(user_content, list):
+        text_parts: list[str] = []
+        for part in user_content:
+            if not isinstance(part, dict):
+                continue
+            if part.get("type") == "text":
+                text_parts.append(part.get("text", ""))
+            elif part.get("type") == "image_url":
+                url = part.get("image_url", {}).get("url", "")
+                if url.startswith("data:"):
+                    saved = _save_base64_data_url(url, media_dir)
+                    if saved:
+                        media_paths.append(saved)
+                elif url:
+                    raise ValueError(
+                        "Remote image URLs are not supported. "
+                        "Use base64 data URLs or upload files via multipart/form-data."
+                    )
+        text = " ".join(text_parts)
+    elif isinstance(user_content, str):
+        text = user_content
+    else:
+        raise ValueError("Invalid content format")
+
+    return text, media_paths
+
+
+async def _parse_multipart(request: web.Request) -> tuple[str, list[str], str | None, str | None]:
+    """Parse multipart/form-data. Returns (text, media_paths, session_id, model)."""
+    media_dir = get_media_dir("api")
+    reader = await request.multipart()
+    text = ""
+    session_id = None
+    model = None
+    media_paths: list[str] = []
+
+    while True:
+        part = await reader.next()
+        if part is None:
+            break
+        if part.name == "message":
+            text = (await part.read()).decode("utf-8")
+        elif part.name == "session_id":
+            session_id = (await part.read()).decode("utf-8").strip()
+        elif part.name == "model":
+            model = (await part.read()).decode("utf-8").strip()
+        elif part.name == "files":
+            raw = await part.read()
+            if len(raw) > MAX_FILE_SIZE:
+                raise _FileSizeExceeded(
+                    f"File '{part.filename}' exceeds {MAX_FILE_SIZE // (1024 * 1024)}MB limit"
+                )
+            base = safe_filename(part.filename or "upload.bin")
+            filename = f"{uuid.uuid4().hex[:12]}_{base}"
+            dest = media_dir / filename
+            dest.write_bytes(raw)
+            media_paths.append(str(dest))
+
+    if not text:
+        text = "请分析上传的文件"
+
+    return text, media_paths, session_id, model
+
 
 # ---------------------------------------------------------------------------
 # Route handlers
 # ---------------------------------------------------------------------------
 
+
 async def handle_chat_completions(request: web.Request) -> web.Response:
-    """POST /v1/chat/completions"""
-
-    # --- Parse body ---
-    try:
-        body = await request.json()
-    except Exception:
-        return _error_json(400, "Invalid JSON body")
-
-    messages = body.get("messages")
-    if not isinstance(messages, list) or len(messages) != 1:
-        return _error_json(400, "Only a single user message is supported")
-
-    # Stream not yet supported
-    if body.get("stream", False):
-        return _error_json(400, "stream=true is not supported yet. Set stream=false or omit it.")
-
-    message = messages[0]
-    if not isinstance(message, dict) or message.get("role") != "user":
-        return _error_json(400, "Only a single user message is supported")
-    user_content = message.get("content", "")
-    if isinstance(user_content, list):
-        # Multi-modal content array — extract text parts
-        user_content = " ".join(
-            part.get("text", "") for part in user_content if part.get("type") == "text"
-        )
+    """POST /v1/chat/completions — supports JSON and multipart/form-data."""
+    content_type = request.content_type or ""
+    if not isinstance(content_type, str):
+        content_type = ""
 
     agent_loop = request.app["agent_loop"]
     timeout_s: float = request.app.get("request_timeout", 120.0)
     model_name: str = request.app.get("model_name", "nanobot")
-    if (requested_model := body.get("model")) and requested_model != model_name:
+
+    stream = False
+    try:
+        if content_type.startswith("multipart/"):
+            text, media_paths, session_id, requested_model = await _parse_multipart(request)
+        else:
+            try:
+                body = await request.json()
+            except Exception:
+                return _error_json(400, "Invalid JSON body")
+            stream = body.get("stream", False)
+            requested_model = body.get("model")
+            text, media_paths = _parse_json_content(body)
+            session_id = body.get("session_id")
+    except ValueError as e:
+        return _error_json(400, str(e))
+    except _FileSizeExceeded as e:
+        return _error_json(413, str(e), err_type="invalid_request_error")
+    except Exception:
+        logger.exception("Error parsing upload")
+        return _error_json(413, "File too large or invalid upload")
+
+    if requested_model and requested_model != model_name:
         return _error_json(400, f"Only configured model '{model_name}' is available")
 
-    session_key = f"api:{body['session_id']}" if body.get("session_id") else API_SESSION_KEY
+    session_key = f"api:{session_id}" if session_id else API_SESSION_KEY
     session_locks: dict[str, asyncio.Lock] = request.app["session_locks"]
     session_lock = session_locks.setdefault(session_key, asyncio.Lock())
 
-    logger.info("API request session_key={} content={}", session_key, user_content[:80])
+    logger.info(
+        "API request session_key={} media={} text={} stream={}",
+        session_key, len(media_paths), text[:80], stream,
+    )
+    # -- streaming path --
+    if stream:
+        resp = web.StreamResponse()
+        resp.content_type = "text/event-stream"
+        resp.headers["Cache-Control"] = "no-cache"
+        resp.headers["Connection"] = "keep-alive"
+        resp.enable_compression()
+        await resp.prepare(request)
 
+        chunk_id = f"chatcmpl-{uuid.uuid4().hex[:12]}"
+        queue: asyncio.Queue[str | None] = asyncio.Queue()
+
+        async def _on_stream(token: str) -> None:
+            await queue.put(token)
+
+        async def _on_stream_end(*_a: Any, **_kw: Any) -> None:
+            await queue.put(None)
+
+        async def _run() -> None:
+            try:
+                async with session_lock:
+                    await asyncio.wait_for(
+                        agent_loop.process_direct(
+                            content=text,
+                            media=media_paths if media_paths else None,
+                            session_key=session_key,
+                            channel="api",
+                            chat_id=API_CHAT_ID,
+                            on_stream=_on_stream,
+                            on_stream_end=_on_stream_end,
+                        ),
+                        timeout=timeout_s,
+                    )
+            except Exception:
+                logger.exception("Streaming error for session {}", session_key)
+                await queue.put(None)
+
+        task = asyncio.create_task(_run())
+        try:
+            while True:
+                token = await queue.get()
+                if token is None:
+                    break
+                await resp.write(_sse_chunk(token, model_name, chunk_id))
+        finally:
+            task.cancel()
+
+        await resp.write(_sse_chunk("", model_name, chunk_id, finish_reason="stop"))
+        await resp.write(_SSE_DONE)
+        return resp
+
+    # -- non-streaming path (original logic) --
     _FALLBACK = EMPTY_FINAL_RESPONSE_MESSAGE
 
     try:
@@ -107,7 +304,8 @@ async def handle_chat_completions(request: web.Request) -> web.Response:
             try:
                 response = await asyncio.wait_for(
                     agent_loop.process_direct(
-                        content=user_content,
+                        content=text,
+                        media=media_paths if media_paths else None,
                         session_key=session_key,
                         channel="api",
                         chat_id=API_CHAT_ID,
@@ -117,13 +315,11 @@ async def handle_chat_completions(request: web.Request) -> web.Response:
                 response_text = _response_text(response)
 
                 if not response_text or not response_text.strip():
-                    logger.warning(
-                        "Empty response for session {}, retrying",
-                        session_key,
-                    )
+                    logger.warning("Empty response for session {}, retrying", session_key)
                     retry_response = await asyncio.wait_for(
                         agent_loop.process_direct(
-                            content=user_content,
+                            content=text,
+                            media=media_paths if media_paths else None,
                             session_key=session_key,
                             channel="api",
                             chat_id=API_CHAT_ID,
@@ -132,10 +328,7 @@ async def handle_chat_completions(request: web.Request) -> web.Response:
                     )
                     response_text = _response_text(retry_response)
                     if not response_text or not response_text.strip():
-                        logger.warning(
-                            "Empty response after retry for session {}, using fallback",
-                            session_key,
-                        )
+                        logger.warning("Empty response after retry, using fallback")
                         response_text = _FALLBACK
 
             except asyncio.TimeoutError:
@@ -153,17 +346,19 @@ async def handle_chat_completions(request: web.Request) -> web.Response:
 async def handle_models(request: web.Request) -> web.Response:
     """GET /v1/models"""
     model_name = request.app.get("model_name", "nanobot")
-    return web.json_response({
-        "object": "list",
-        "data": [
-            {
-                "id": model_name,
-                "object": "model",
-                "created": 0,
-                "owned_by": "nanobot",
-            }
-        ],
-    })
+    return web.json_response(
+        {
+            "object": "list",
+            "data": [
+                {
+                    "id": model_name,
+                    "object": "model",
+                    "created": 0,
+                    "owned_by": "nanobot",
+                }
+            ],
+        }
+    )
 
 
 async def handle_health(request: web.Request) -> web.Response:
@@ -175,7 +370,10 @@ async def handle_health(request: web.Request) -> web.Response:
 # App factory
 # ---------------------------------------------------------------------------
 
-def create_app(agent_loop, model_name: str = "nanobot", request_timeout: float = 120.0) -> web.Application:
+
+def create_app(
+    agent_loop, model_name: str = "nanobot", request_timeout: float = 120.0
+) -> web.Application:
     """Create the aiohttp application.
 
     Args:
@@ -183,7 +381,7 @@ def create_app(agent_loop, model_name: str = "nanobot", request_timeout: float =
         model_name: Model name reported in responses.
         request_timeout: Per-request timeout in seconds.
     """
-    app = web.Application()
+    app = web.Application(client_max_size=20 * 1024 * 1024)  # 20MB for base64 images
     app["agent_loop"] = agent_loop
     app["model_name"] = model_name
     app["request_timeout"] = request_timeout

nanobot/channels/base.py

@@ -24,6 +24,7 @@ class BaseChannel(ABC):
     display_name: str = "Base"
     transcription_provider: str = "groq"
     transcription_api_key: str = ""
+    transcription_api_base: str = ""
 
     def __init__(self, config: Any, bus: MessageBus):
         """
@@ -44,10 +45,16 @@ class BaseChannel(ABC):
         try:
             if self.transcription_provider == "openai":
                 from nanobot.providers.transcription import OpenAITranscriptionProvider
-                provider = OpenAITranscriptionProvider(api_key=self.transcription_api_key)
+                provider = OpenAITranscriptionProvider(
+                    api_key=self.transcription_api_key,
+                    api_base=self.transcription_api_base or None,
+                )
             else:
                 from nanobot.providers.transcription import GroqTranscriptionProvider
-                provider = GroqTranscriptionProvider(api_key=self.transcription_api_key)
+                provider = GroqTranscriptionProvider(
+                    api_key=self.transcription_api_key,
+                    api_base=self.transcription_api_base or None,
+                )
             return await provider.transcribe(file_path)
         except Exception as e:
             logger.warning("{}: audio transcription failed: {}", self.name, e)
@@ -116,7 +123,13 @@ class BaseChannel(ABC):
 
     def is_allowed(self, sender_id: str) -> bool:
         """Check if *sender_id* is permitted.  Empty list → deny all; ``"*"`` → allow all."""
-        allow_list = getattr(self.config, "allow_from", [])
+        if isinstance(self.config, dict):
+            if "allow_from" in self.config:
+                allow_list = self.config.get("allow_from")
+            else:
+                allow_list = self.config.get("allowFrom", [])
+        else:
+            allow_list = getattr(self.config, "allow_from", [])
         if not allow_list:
             logger.warning("{}: allow_from is empty — all access denied", self.name)
             return False

nanobot/channels/dingtalk.py

@@ -5,6 +5,8 @@ import json
 import mimetypes
 import os
 import time
+import zipfile
+from io import BytesIO
 from pathlib import Path
 from typing import Any
 from urllib.parse import unquote, urlparse
@@ -171,6 +173,7 @@ class DingTalkChannel(BaseChannel):
     _IMAGE_EXTS = {".jpg", ".jpeg", ".png", ".gif", ".bmp", ".webp"}
     _AUDIO_EXTS = {".amr", ".mp3", ".wav", ".ogg", ".m4a", ".aac"}
     _VIDEO_EXTS = {".mp4", ".mov", ".avi", ".mkv", ".webm"}
+    _ZIP_BEFORE_UPLOAD_EXTS = {".htm", ".html"}
 
     @classmethod
     def default_config(cls) -> dict[str, Any]:
@@ -287,6 +290,31 @@ class DingTalkChannel(BaseChannel):
         name = os.path.basename(urlparse(media_ref).path)
         return name or {"image": "image.jpg", "voice": "audio.amr", "video": "video.mp4"}.get(upload_type, "file.bin")
 
+    @staticmethod
+    def _zip_bytes(filename: str, data: bytes) -> tuple[bytes, str, str]:
+        stem = Path(filename).stem or "attachment"
+        safe_name = filename or "attachment.bin"
+        zip_name = f"{stem}.zip"
+        buffer = BytesIO()
+        with zipfile.ZipFile(buffer, mode="w", compression=zipfile.ZIP_DEFLATED) as archive:
+            archive.writestr(safe_name, data)
+        return buffer.getvalue(), zip_name, "application/zip"
+
+    def _normalize_upload_payload(
+        self,
+        filename: str,
+        data: bytes,
+        content_type: str | None,
+    ) -> tuple[bytes, str, str | None]:
+        ext = Path(filename).suffix.lower()
+        if ext in self._ZIP_BEFORE_UPLOAD_EXTS or content_type == "text/html":
+            logger.info(
+                "DingTalk does not accept raw HTML attachments, zipping {} before upload",
+                filename,
+            )
+            return self._zip_bytes(filename, data)
+        return data, filename, content_type
+
     async def _read_media_bytes(
         self,
         media_ref: str,
@@ -309,6 +337,9 @@ class DingTalkChannel(BaseChannel):
                 content_type = (resp.headers.get("content-type") or "").split(";")[0].strip()
                 filename = self._guess_filename(media_ref, self._guess_upload_type(media_ref))
                 return resp.content, filename, content_type or None
+            except httpx.TransportError as e:
+                logger.error("DingTalk media download network error ref={} err={}", media_ref, e)
+                raise
             except Exception as e:
                 logger.error("DingTalk media download error ref={} err={}", media_ref, e)
                 return None, None, None
@@ -360,6 +391,9 @@ class DingTalkChannel(BaseChannel):
                 logger.error("DingTalk media upload missing media_id body={}", text[:500])
                 return None
             return str(media_id)
+        except httpx.TransportError as e:
+            logger.error("DingTalk media upload network error type={} err={}", media_type, e)
+            raise
         except Exception as e:
             logger.error("DingTalk media upload error type={} err={}", media_type, e)
             return None
@@ -409,6 +443,9 @@ class DingTalkChannel(BaseChannel):
                 return False
             logger.debug("DingTalk message sent to {} with msgKey={}", chat_id, msg_key)
             return True
+        except httpx.TransportError as e:
+            logger.error("DingTalk network error sending message msgKey={} err={}", msg_key, e)
+            raise
         except Exception as e:
             logger.error("Error sending DingTalk message msgKey={} err={}", msg_key, e)
             return False
@@ -444,6 +481,7 @@ class DingTalkChannel(BaseChannel):
             return False
 
         filename = filename or self._guess_filename(media_ref, upload_type)
+        data, filename, content_type = self._normalize_upload_payload(filename, data, content_type)
         file_type = Path(filename).suffix.lower().lstrip(".")
         if not file_type:
             guessed = mimetypes.guess_extension(content_type or "")

nanobot/channels/discord.py

@@ -4,6 +4,8 @@ from __future__ import annotations
 
 import asyncio
 import importlib.util
+import time
+from dataclasses import dataclass
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Literal
 
@@ -20,6 +22,7 @@ from nanobot.utils.helpers import safe_filename, split_message
 
 DISCORD_AVAILABLE = importlib.util.find_spec("discord") is not None
 if TYPE_CHECKING:
+    import aiohttp
     import discord
     from discord import app_commands
     from discord.abc import Messageable
@@ -34,17 +37,32 @@ MAX_MESSAGE_LEN = 2000  # Discord message character limit
 TYPING_INTERVAL_S = 8
 
 
+@dataclass
+class _StreamBuf:
+    """Per-chat streaming accumulator for progressive Discord message edits."""
+
+    text: str = ""
+    message: Any | None = None
+    last_edit: float = 0.0
+    stream_id: str | None = None
+
+
 class DiscordConfig(Base):
     """Discord channel configuration."""
 
     enabled: bool = False
     token: str = ""
     allow_from: list[str] = Field(default_factory=list)
+    allow_channels: list[str] = Field(default_factory=list)  # Allowed channel IDs (empty = all)
     intents: int = 37377
     group_policy: Literal["mention", "open"] = "mention"
     read_receipt_emoji: str = "👀"
     working_emoji: str = "🔧"
     working_emoji_delay: float = 2.0
+    streaming: bool = True
+    proxy: str | None = None
+    proxy_username: str | None = None
+    proxy_password: str | None = None
 
 
 if DISCORD_AVAILABLE:
@@ -52,8 +70,15 @@ if DISCORD_AVAILABLE:
     class DiscordBotClient(discord.Client):
         """discord.py client that forwards events to the channel."""
 
-        def __init__(self, channel: DiscordChannel, *, intents: discord.Intents) -> None:
-            super().__init__(intents=intents)
+        def __init__(
+            self,
+            channel: DiscordChannel,
+            *,
+            intents: discord.Intents,
+            proxy: str | None = None,
+            proxy_auth: aiohttp.BasicAuth | None = None,
+        ) -> None:
+            super().__init__(intents=intents, proxy=proxy, proxy_auth=proxy_auth)
             self._channel = channel
             self.tree = app_commands.CommandTree(self)
             self._register_app_commands()
@@ -117,6 +142,7 @@ if DISCORD_AVAILABLE:
             )
 
             for name, description, command_text in commands:
+
                 @self.tree.command(name=name, description=description)
                 async def command_handler(
                     interaction: discord.Interaction,
@@ -173,7 +199,9 @@ if DISCORD_AVAILABLE:
                 else:
                     failed_media.append(Path(media_path).name)
 
-            for index, chunk in enumerate(self._build_chunks(msg.content or "", failed_media, sent_media)):
+            for index, chunk in enumerate(
+                self._build_chunks(msg.content or "", failed_media, sent_media)
+            ):
                 kwargs: dict[str, Any] = {"content": chunk}
                 if index == 0 and reference is not None and not sent_media:
                     kwargs["reference"] = reference
@@ -242,6 +270,7 @@ class DiscordChannel(BaseChannel):
 
     name = "discord"
     display_name = "Discord"
+    _STREAM_EDIT_INTERVAL = 0.8
 
     @classmethod
     def default_config(cls) -> dict[str, Any]:
@@ -263,6 +292,7 @@ class DiscordChannel(BaseChannel):
         self._bot_user_id: str | None = None
         self._pending_reactions: dict[str, Any] = {}  # chat_id -> message object
         self._working_emoji_tasks: dict[str, asyncio.Task[None]] = {}
+        self._stream_bufs: dict[str, _StreamBuf] = {}
 
     async def start(self) -> None:
         """Start the Discord client."""
@@ -277,7 +307,29 @@ class DiscordChannel(BaseChannel):
         try:
             intents = discord.Intents.none()
             intents.value = self.config.intents
-            self._client = DiscordBotClient(self, intents=intents)
+
+            proxy_auth = None
+            has_user = bool(self.config.proxy_username)
+            has_pass = bool(self.config.proxy_password)
+            if has_user and has_pass:
+                import aiohttp
+
+                proxy_auth = aiohttp.BasicAuth(
+                    login=self.config.proxy_username,
+                    password=self.config.proxy_password,
+                )
+            elif has_user != has_pass:
+                logger.warning(
+                    "Discord proxy auth incomplete: both proxy_username and "
+                    "proxy_password must be set; ignoring partial credentials",
+                )
+
+            self._client = DiscordBotClient(
+                self,
+                intents=intents,
+                proxy=self.config.proxy,
+                proxy_auth=proxy_auth,
+            )
         except Exception as e:
             logger.error("Failed to initialize Discord client: {}", e)
             self._client = None
@@ -315,11 +367,71 @@ class DiscordChannel(BaseChannel):
             await client.send_outbound(msg)
         except Exception as e:
             logger.error("Error sending Discord message: {}", e)
+            raise
         finally:
             if not is_progress:
                 await self._stop_typing(msg.chat_id)
                 await self._clear_reactions(msg.chat_id)
 
+    async def send_delta(
+        self, chat_id: str, delta: str, metadata: dict[str, Any] | None = None
+    ) -> None:
+        """Progressive Discord delivery: send once, then edit until the stream ends."""
+        client = self._client
+        if client is None or not client.is_ready():
+            logger.warning("Discord client not ready; dropping stream delta")
+            return
+
+        meta = metadata or {}
+        stream_id = meta.get("_stream_id")
+
+        if meta.get("_stream_end"):
+            buf = self._stream_bufs.get(chat_id)
+            if not buf or buf.message is None or not buf.text:
+                return
+            if stream_id is not None and buf.stream_id is not None and buf.stream_id != stream_id:
+                return
+            await self._finalize_stream(chat_id, buf)
+            return
+
+        buf = self._stream_bufs.get(chat_id)
+        if buf is None or (
+            stream_id is not None and buf.stream_id is not None and buf.stream_id != stream_id
+        ):
+            buf = _StreamBuf(stream_id=stream_id)
+            self._stream_bufs[chat_id] = buf
+        elif buf.stream_id is None:
+            buf.stream_id = stream_id
+
+        buf.text += delta
+        if not buf.text.strip():
+            return
+
+        target = await self._resolve_channel(chat_id)
+        if target is None:
+            logger.warning("Discord stream target {} unavailable", chat_id)
+            return
+
+        now = time.monotonic()
+        if buf.message is None:
+            try:
+                buf.message = await target.send(content=buf.text)
+                buf.last_edit = now
+            except Exception as e:
+                logger.warning("Discord stream initial send failed: {}", e)
+                raise
+            return
+
+        if (now - buf.last_edit) < self._STREAM_EDIT_INTERVAL:
+            return
+
+        try:
+            await buf.message.edit(content=DiscordBotClient._build_chunks(buf.text, [], False)[0])
+            buf.last_edit = now
+        except Exception as e:
+            logger.warning("Discord stream edit failed: {}", e)
+            raise
+
     async def _handle_discord_message(self, message: discord.Message) -> None:
         """Handle incoming Discord messages from discord.py."""
         if message.author.bot:
@@ -339,7 +451,6 @@ class DiscordChannel(BaseChannel):
         await self._start_typing(message.channel)
 
         # Add read receipt reaction immediately, working emoji after delay
-        channel_id = self._channel_key(message.channel)
         try:
             await message.add_reaction(self.config.read_receipt_emoji)
             self._pending_reactions[channel_id] = message
@@ -373,6 +484,47 @@ class DiscordChannel(BaseChannel):
         """Backward-compatible alias for legacy tests/callers."""
         await self._handle_discord_message(message)
 
+    async def _resolve_channel(self, chat_id: str) -> Any | None:
+        """Resolve a Discord channel from cache first, then network fetch."""
+        client = self._client
+        if client is None or not client.is_ready():
+            return None
+        channel_id = int(chat_id)
+        channel = client.get_channel(channel_id)
+        if channel is not None:
+            return channel
+        try:
+            return await client.fetch_channel(channel_id)
+        except Exception as e:
+            logger.warning("Discord channel {} unavailable: {}", chat_id, e)
+            return None
+
+    async def _finalize_stream(self, chat_id: str, buf: _StreamBuf) -> None:
+        """Commit the final streamed content and flush overflow chunks."""
+        chunks = DiscordBotClient._build_chunks(buf.text, [], False)
+        if not chunks:
+            self._stream_bufs.pop(chat_id, None)
+            return
+
+        try:
+            await buf.message.edit(content=chunks[0])
+        except Exception as e:
+            logger.warning("Discord final stream edit failed: {}", e)
+            raise
+
+        target = getattr(buf.message, "channel", None) or await self._resolve_channel(chat_id)
+        if target is None:
+            logger.warning("Discord stream follow-up target {} unavailable", chat_id)
+            self._stream_bufs.pop(chat_id, None)
+            return
+
+        for extra_chunk in chunks[1:]:
+            await target.send(content=extra_chunk)
+
+        self._stream_bufs.pop(chat_id, None)
+        await self._stop_typing(chat_id)
+        await self._clear_reactions(chat_id)
+
     def _should_accept_inbound(
         self,
         message: discord.Message,
@@ -382,6 +534,12 @@ class DiscordChannel(BaseChannel):
         """Check if inbound Discord message should be processed."""
         if not self.is_allowed(sender_id):
             return False
+        # Channel-based filtering: only respond in allowed channels
+        allow_channels = self.config.allow_channels
+        if allow_channels:
+            channel_id = self._channel_key(message.channel)
+            if channel_id not in allow_channels:
+                return False
         if message.guild is not None and not self._should_respond_in_group(message, content):
             return False
         return True
@@ -423,7 +581,11 @@ class DiscordChannel(BaseChannel):
     @staticmethod
     def _build_inbound_metadata(message: discord.Message) -> dict[str, str | None]:
         """Build metadata for inbound Discord messages."""
-        reply_to = str(message.reference.message_id) if message.reference and message.reference.message_id else None
+        reply_to = (
+            str(message.reference.message_id)
+            if message.reference and message.reference.message_id
+            else None
+        )
         return {
             "message_id": str(message.id),
             "guild_id": str(message.guild.id) if message.guild else None,
@@ -438,7 +600,9 @@ class DiscordChannel(BaseChannel):
         if self.config.group_policy == "mention":
             bot_user_id = self._bot_user_id
             if bot_user_id is None:
-                logger.debug("Discord message in {} ignored (bot identity unavailable)", message.channel.id)
+                logger.debug(
+                    "Discord message in {} ignored (bot identity unavailable)", message.channel.id
+                )
                 return False
 
             if any(str(user.id) == bot_user_id for user in message.mentions):
@@ -480,7 +644,6 @@ class DiscordChannel(BaseChannel):
         except asyncio.CancelledError:
             pass
 
-
     async def _clear_reactions(self, chat_id: str) -> None:
         """Remove all pending reactions after bot replies."""
         # Cancel delayed working emoji if it hasn't fired yet
@@ -507,6 +670,7 @@ class DiscordChannel(BaseChannel):
     async def _reset_runtime_state(self, close_client: bool) -> None:
         """Reset client and typing state."""
         await self._cancel_all_typing()
+        self._stream_bufs.clear()
         if close_client and self._client is not None and not self._client.is_closed():
             try:
                 await self._client.close()

nanobot/channels/feishu.py

@@ -1290,7 +1290,6 @@ class FeishuChannel(BaseChannel):
 
         Supported metadata keys:
             _stream_end: Finalize the streaming card.
-            _resuming:   Mid-turn pause – flush but keep the buffer alive.
             _tool_hint:  Delta is a formatted tool hint (for display only).
             message_id:  Original message id (used with _stream_end for reaction cleanup).
             reaction_id: Reaction id to remove on stream end.
@@ -1309,50 +1308,44 @@ class FeishuChannel(BaseChannel):
                 if self.config.done_emoji and message_id:
                     await self._add_reaction(message_id, self.config.done_emoji)
 
-            resuming = meta.get("_resuming", False)
-            if resuming:
-                # Mid-turn pause (e.g. tool call between streaming segments).
-                # Flush current text to card but keep the buffer alive so the
-                # next segment appends to the same card.
-                buf = self._stream_bufs.get(chat_id)
-                if buf and buf.card_id and buf.text:
-                    buf.sequence += 1
-                    await loop.run_in_executor(
-                        None, self._stream_update_text_sync, buf.card_id, buf.text, buf.sequence,
-                    )
-                return
-
             buf = self._stream_bufs.pop(chat_id, None)
             if not buf or not buf.text:
                 return
+            # Try to finalize via streaming card; if that fails (e.g.
+            # streaming mode was closed by Feishu due to timeout), fall
+            # back to sending a regular interactive card.
             if buf.card_id:
                 buf.sequence += 1
-                await loop.run_in_executor(
+                ok = await loop.run_in_executor(
                     None,
                     self._stream_update_text_sync,
                     buf.card_id,
                     buf.text,
                     buf.sequence,
                 )
-                # Required so the chat list preview exits the streaming placeholder (Feishu streaming card docs).
-                buf.sequence += 1
-                await loop.run_in_executor(
-                    None,
-                    self._close_streaming_mode_sync,
-                    buf.card_id,
-                    buf.sequence,
-                )
-            else:
-                for chunk in self._split_elements_by_table_limit(
-                    self._build_card_elements(buf.text)
-                ):
-                    card = json.dumps(
-                        {"config": {"wide_screen_mode": True}, "elements": chunk},
-                        ensure_ascii=False,
-                    )
+                if ok:
+                    buf.sequence += 1
                     await loop.run_in_executor(
-                        None, self._send_message_sync, rid_type, chat_id, "interactive", card
+                        None,
+                        self._close_streaming_mode_sync,
+                        buf.card_id,
+                        buf.sequence,
                     )
+                    return
+                logger.warning(
+                    "Streaming card {} final update failed, falling back to regular card",
+                    buf.card_id,
+                )
+            for chunk in self._split_elements_by_table_limit(
+                self._build_card_elements(buf.text)
+            ):
+                card = json.dumps(
+                    {"config": {"wide_screen_mode": True}, "elements": chunk},
+                    ensure_ascii=False,
+                )
+                await loop.run_in_executor(
+                    None, self._send_message_sync, rid_type, chat_id, "interactive", card
+                )
             return
 
         # --- accumulate delta ---
@@ -1404,14 +1397,21 @@ class FeishuChannel(BaseChannel):
                 if buf and buf.card_id:
                     # Delegate to send_delta so tool hints get the same
                     # throttling (and card creation) as regular text deltas.
-                    lines = self.__class__._format_tool_hint_lines(hint).split("\n")
-                    delta = "\n\n" + "\n".join(
-                        f"{self.config.tool_hint_prefix} {ln}" for ln in lines if ln.strip()
-                    ) + "\n\n"
-                    await self.send_delta(msg.chat_id, delta)
+                    await self.send_delta(
+                        msg.chat_id,
+                        "\n\n" + self._format_tool_hint_delta(hint) + "\n\n",
+                    )
                     return
-                await self._send_tool_hint_card(
-                    receive_id_type, msg.chat_id, hint
+                # No active streaming card — send as a regular
+                # interactive card with the same 🔧 prefix style.
+                card = json.dumps(
+                    {"config": {"wide_screen_mode": True}, "elements": [
+                        {"tag": "markdown", "content": self._format_tool_hint_delta(hint)},
+                    ]},
+                    ensure_ascii=False,
+                )
+                await loop.run_in_executor(
+                    None, self._send_message_sync, receive_id_type, msg.chat_id, "interactive", card
                 )
                 return
 
@@ -1708,33 +1708,9 @@ class FeishuChannel(BaseChannel):
 
         return "\n".join(part for part in parts if part)
 
-    async def _send_tool_hint_card(
-        self, receive_id_type: str, receive_id: str, tool_hint: str
-    ) -> None:
-        """Send tool hint as an interactive card with formatted code block.
-
-        Args:
-            receive_id_type: "chat_id" or "open_id"
-            receive_id: The target chat or user ID
-            tool_hint: Formatted tool hint string (e.g., 'web_search("q"), read_file("path")')
-        """
-        loop = asyncio.get_running_loop()
-
-        # Put each top-level tool call on its own line without altering commas inside arguments.
-        formatted_code = self.__class__._format_tool_hint_lines(tool_hint)
-
-        card = {
-            "config": {"wide_screen_mode": True},
-            "elements": [
-                {"tag": "markdown", "content": f"**Tool Calls**\n\n```text\n{formatted_code}\n```"}
-            ],
-        }
-
-        await loop.run_in_executor(
-            None,
-            self._send_message_sync,
-            receive_id_type,
-            receive_id,
-            "interactive",
-            json.dumps(card, ensure_ascii=False),
+    def _format_tool_hint_delta(self, tool_hint: str) -> str:
+        """Format a tool hint string with the 🔧 prefix for each line."""
+        lines = self.__class__._format_tool_hint_lines(tool_hint).split("\n")
+        return "\n".join(
+            f"{self.config.tool_hint_prefix} {ln}" for ln in lines if ln.strip()
         )

nanobot/channels/manager.py

@@ -41,6 +41,7 @@ class ChannelManager:
 
         transcription_provider = self.config.channels.transcription_provider
         transcription_key = self._resolve_transcription_key(transcription_provider)
+        transcription_base = self._resolve_transcription_base(transcription_provider)
 
         for name, cls in discover_all().items():
             section = getattr(self.config.channels, name, None)
@@ -57,6 +58,7 @@ class ChannelManager:
                 channel = cls(section, self.bus)
                 channel.transcription_provider = transcription_provider
                 channel.transcription_api_key = transcription_key
+                channel.transcription_api_base = transcription_base
                 self.channels[name] = channel
                 logger.info("{} channel enabled", cls.display_name)
             except Exception as e:
@@ -73,9 +75,26 @@ class ChannelManager:
         except AttributeError:
             return ""
 
+    def _resolve_transcription_base(self, provider: str) -> str:
+        """Pick the API base URL for the configured transcription provider."""
+        try:
+            if provider == "openai":
+                return self.config.providers.openai.api_base or ""
+            return self.config.providers.groq.api_base or ""
+        except AttributeError:
+            return ""
+
     def _validate_allow_from(self) -> None:
         for name, ch in self.channels.items():
-            if getattr(ch.config, "allow_from", None) == []:
+            cfg = ch.config
+            if isinstance(cfg, dict):
+                if "allow_from" in cfg:
+                    allow = cfg.get("allow_from")
+                else:
+                    allow = cfg.get("allowFrom")
+            else:
+                allow = getattr(cfg, "allow_from", None)
+            if allow == []:
                 raise SystemExit(
                     f'Error: "{name}" has empty allowFrom (denies all). '
                     f'Set ["*"] to allow everyone, or add specific user IDs.'

nanobot/channels/msteams.py

@@ -17,6 +17,7 @@ import importlib.util
 import json
 import re
 import threading
+import time
 from dataclasses import dataclass
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
 from typing import TYPE_CHECKING, Any
@@ -31,7 +32,10 @@ from nanobot.channels.base import BaseChannel
 from nanobot.config.paths import get_workspace_path
 from nanobot.config.schema import Base
 
-MSTEAMS_AVAILABLE = importlib.util.find_spec("jwt") is not None
+MSTEAMS_AVAILABLE = (
+    importlib.util.find_spec("jwt") is not None
+    and importlib.util.find_spec("cryptography") is not None
+)
 
 if TYPE_CHECKING:
     import jwt
@@ -53,12 +57,7 @@ class MSTeamsConfig(Base):
     allow_from: list[str] = Field(default_factory=list)
     reply_in_thread: bool = True
     mention_only_response: str = "Hi — what can I help with?"
-    validate_inbound_auth: bool = False
-    restart_notify_enabled: bool = False
-    restart_notify_pre_message: str = (
-        "Nanobot agent initiated a gateway restart. I will message again when the gateway is back online."
-    )
-    restart_notify_post_message: str = "Nanobot gateway is back online."
+    validate_inbound_auth: bool = True
 
 
 @dataclass
@@ -115,6 +114,13 @@ class MSTeamsChannel(BaseChannel):
             logger.error("MSTeams app_id/app_password not configured")
             return
 
+        if not self.config.validate_inbound_auth:
+            logger.warning(
+                "MSTeams inbound auth validation was explicitly DISABLED in config. "
+                "Anyone who knows the webhook URL can send messages as any user. "
+                "Only disable this for local development or controlled testing."
+            )
+
         self._loop = asyncio.get_running_loop()
         self._http = httpx.AsyncClient(timeout=30.0)
         self._running = True
@@ -268,6 +274,14 @@ class MSTeamsChannel(BaseChannel):
                 logger.debug("MSTeams ignoring empty message after Teams text sanitization")
                 return
 
+        if not self.is_allowed(sender_id):
+            logger.warning(
+                "Access denied for sender {} on channel {}. "
+                "Add them to allowFrom list in config to grant access.",
+                sender_id, self.name,
+            )
+            return
+
         self._conversation_refs[conversation_id] = ConversationRef(
             service_url=service_url,
             conversation_id=conversation_id,
@@ -305,7 +319,7 @@ class MSTeamsChannel(BaseChannel):
         while preview_lines and not preview_lines[0]:
             preview_lines.pop(0)
         first_line = preview_lines[0] if preview_lines else ""
-        looks_like_quote_wrapper = first_line.lower().startswith("replying to ") or first_line.startswith("FWDIOC-BOT")
+        looks_like_quote_wrapper = first_line.lower().startswith("replying to ") or first_line.startswith("Reply wrapper")
 
         if reply_to_id or channel_data.get("messageType") == "reply" or looks_like_quote_wrapper:
             text = self._normalize_teams_reply_quote(text)
@@ -338,10 +352,10 @@ class MSTeamsChannel(BaseChannel):
             reply = "\n".join(lines[1:]).strip()
             return self._format_reply_with_quote(quoted, reply)
 
-        # Observed FWDIOC relay wrapper where the quoted content is surfaced after a
-        # synthetic "FWDIOC-BOT" header, sometimes with a blank line separating quote
+        # Observed reply wrapper where the quoted content is surfaced after a
+        # synthetic "Reply wrapper" header, sometimes with a blank line separating quote
         # and reply, and sometimes as a compact line-based fallback shape.
-        if lines and lines[0].strip().startswith("FWDIOC-BOT"):
+        if lines and lines[0].strip().startswith("Reply wrapper"):
             body = normalized_newlines.split("\n", 1)[1] if "\n" in normalized_newlines else ""
             body = body.lstrip()
             parts = re.split(r"\n\s*\n", body, maxsplit=1)
@@ -358,22 +372,24 @@ class MSTeamsChannel(BaseChannel):
                 if quoted and reply:
                     return self._format_reply_with_quote(quoted, reply)
 
-        # Observed compact fallback where the relay flattens everything into one line
-        # and appends the literal reply text marker at the end.
+        # Observed compact fallback where the relay flattens quote and reply into
+        # a single line after the synthetic Reply wrapper prefix.
         compact = re.sub(r"\s+", " ", normalized_newlines).strip()
-        if compact.startswith("FWDIOC-BOT "):
-            compact = compact[len("FWDIOC-BOT ") :].strip()
-
-        marker = " Reply with quote test"
-        if compact.endswith(marker):
-            quoted = compact[: -len(marker)].strip()
-            reply = marker.strip()
-            return self._format_reply_with_quote(quoted, reply)
+        if compact.startswith("Reply wrapper "):
+            compact = compact[len("Reply wrapper ") :].strip()
+            for boundary in (". ", "! ", "? ", "… "):
+                idx = compact.rfind(boundary)
+                if idx == -1:
+                    continue
+                quoted = compact[: idx + 1].strip()
+                reply = compact[idx + len(boundary) :].strip()
+                if quoted and reply and len(reply) <= 160:
+                    return self._format_reply_with_quote(quoted, reply)
 
         return cleaned
 
     def _format_reply_with_quote(self, quoted: str, reply: str) -> str:
-        """Format a quoted reply for the model without Teams wrapper noise."""
+        """Format a reply-with-context message for the model without Teams wrapper noise."""
         quoted = quoted.strip()
         reply = reply.strip()
         if quoted and reply:
@@ -426,7 +442,6 @@ class MSTeamsChannel(BaseChannel):
 
     async def _get_botframework_openid_config(self) -> dict[str, Any]:
         """Fetch and cache Bot Framework OpenID configuration."""
-        import time
 
         now = time.time()
         if self._botframework_openid_config and now < self._botframework_openid_config_expires_at:
@@ -443,7 +458,6 @@ class MSTeamsChannel(BaseChannel):
 
     async def _get_botframework_jwks(self) -> dict[str, Any]:
         """Fetch and cache Bot Framework JWKS."""
-        import time
 
         now = time.time()
         if self._botframework_jwks and now < self._botframework_jwks_expires_at:
@@ -497,7 +511,6 @@ class MSTeamsChannel(BaseChannel):
 
     async def _get_access_token(self) -> str:
         """Fetch an access token for Bot Framework / Azure Bot auth."""
-        import time
 
         now = time.time()
         if self._token and now < self._token_expires_at - 60:

nanobot/channels/qq.py

@@ -280,6 +280,9 @@ class QQChannel(BaseChannel):
                     msg_id=msg_id,
                     content=msg.content.strip(),
                 )
+        except (aiohttp.ClientError, OSError):
+            # Network / transport errors — propagate so ChannelManager can retry
+            raise
         except Exception:
             logger.exception("Error sending QQ message to chat_id={}", msg.chat_id)
 
@@ -362,7 +365,12 @@ class QQChannel(BaseChannel):
 
             logger.info("QQ media sent: {}", filename)
             return True
+        except (aiohttp.ClientError, OSError) as e:
+            # Network / transport errors — propagate for retry by caller
+            logger.warning("QQ send media network error filename={} err={}", filename, e)
+            raise
         except Exception as e:
+            # API-level or other non-network errors — return False so send() can fallback
             logger.error("QQ send media failed filename={} err={}", filename, e)
             return False
 

nanobot/channels/slack.py

@@ -5,6 +5,7 @@ import re
 from typing import Any
 
 from loguru import logger
+from pydantic import Field
 from slack_sdk.socket_mode.request import SocketModeRequest
 from slack_sdk.socket_mode.response import SocketModeResponse
 from slack_sdk.socket_mode.websockets import SocketModeClient
@@ -13,8 +14,6 @@ from slackify_markdown import slackify_markdown
 
 from nanobot.bus.events import OutboundMessage
 from nanobot.bus.queue import MessageBus
-from pydantic import Field
-
 from nanobot.channels.base import BaseChannel
 from nanobot.config.schema import Base
 
@@ -50,6 +49,9 @@ class SlackChannel(BaseChannel):
 
     name = "slack"
     display_name = "Slack"
+    _SLACK_ID_RE = re.compile(r"^[CDGUW][A-Z0-9]{2,}$")
+    _SLACK_CHANNEL_REF_RE = re.compile(r"^<#([A-Z0-9]+)(?:\|[^>]+)?>$")
+    _SLACK_USER_REF_RE = re.compile(r"^<@([A-Z0-9]+)(?:\|[^>]+)?>$")
 
     @classmethod
     def default_config(cls) -> dict[str, Any]:
@@ -63,6 +65,7 @@ class SlackChannel(BaseChannel):
         self._web_client: AsyncWebClient | None = None
         self._socket_client: SocketModeClient | None = None
         self._bot_user_id: str | None = None
+        self._target_cache: dict[str, str] = {}
 
     async def start(self) -> None:
         """Start the Slack Socket Mode client."""
@@ -113,17 +116,23 @@ class SlackChannel(BaseChannel):
             logger.warning("Slack client not running")
             return
         try:
+            target_chat_id = await self._resolve_target_chat_id(msg.chat_id)
             slack_meta = msg.metadata.get("slack", {}) if msg.metadata else {}
             thread_ts = slack_meta.get("thread_ts")
             channel_type = slack_meta.get("channel_type")
+            origin_chat_id = str((slack_meta.get("event", {}) or {}).get("channel") or msg.chat_id)
             # Slack DMs don't use threads; channel/group replies may keep thread_ts.
-            thread_ts_param = thread_ts if thread_ts and channel_type != "im" else None
+            thread_ts_param = (
+                thread_ts
+                if thread_ts and channel_type != "im" and target_chat_id == origin_chat_id
+                else None
+            )
 
             # Slack rejects empty text payloads. Keep media-only messages media-only,
             # but send a single blank message when the bot has no text or files to send.
             if msg.content or not (msg.media or []):
                 await self._web_client.chat_postMessage(
-                    channel=msg.chat_id,
+                    channel=target_chat_id,
                     text=self._to_mrkdwn(msg.content) if msg.content else " ",
                     thread_ts=thread_ts_param,
                 )
@@ -131,7 +140,7 @@ class SlackChannel(BaseChannel):
             for media_path in msg.media or []:
                 try:
                     await self._web_client.files_upload_v2(
-                        channel=msg.chat_id,
+                        channel=target_chat_id,
                         file=media_path,
                         thread_ts=thread_ts_param,
                     )
@@ -141,12 +150,123 @@ class SlackChannel(BaseChannel):
             # Update reaction emoji when the final (non-progress) response is sent
             if not (msg.metadata or {}).get("_progress"):
                 event = slack_meta.get("event", {})
-                await self._update_react_emoji(msg.chat_id, event.get("ts"))
+                await self._update_react_emoji(origin_chat_id, event.get("ts"))
 
         except Exception as e:
             logger.error("Error sending Slack message: {}", e)
             raise
 
+    async def _resolve_target_chat_id(self, target: str) -> str:
+        """Resolve human-friendly Slack targets to concrete IDs when needed."""
+        if not self._web_client:
+            return target
+
+        target = target.strip()
+        if not target:
+            return target
+
+        if match := self._SLACK_CHANNEL_REF_RE.fullmatch(target):
+            return match.group(1)
+        if match := self._SLACK_USER_REF_RE.fullmatch(target):
+            return await self._open_dm_for_user(match.group(1))
+        if self._SLACK_ID_RE.fullmatch(target):
+            if target.startswith(("U", "W")):
+                return await self._open_dm_for_user(target)
+            return target
+
+        if target.startswith("#"):
+            return await self._resolve_channel_name(target[1:])
+        if target.startswith("@"):
+            return await self._resolve_user_handle(target[1:])
+
+        try:
+            return await self._resolve_channel_name(target)
+        except ValueError:
+            return await self._resolve_user_handle(target)
+
+    async def _resolve_channel_name(self, name: str) -> str:
+        normalized = self._normalize_target_name(name)
+        if not normalized:
+            raise ValueError("Slack target channel name is empty")
+
+        cache_key = f"channel:{normalized}"
+        if cache_key in self._target_cache:
+            return self._target_cache[cache_key]
+
+        cursor: str | None = None
+        while True:
+            response = await self._web_client.conversations_list(
+                types="public_channel,private_channel",
+                exclude_archived=True,
+                limit=200,
+                cursor=cursor,
+            )
+            for channel in response.get("channels", []):
+                if self._normalize_target_name(str(channel.get("name") or "")) == normalized:
+                    channel_id = str(channel.get("id") or "")
+                    if channel_id:
+                        self._target_cache[cache_key] = channel_id
+                        return channel_id
+            cursor = ((response.get("response_metadata") or {}).get("next_cursor") or "").strip()
+            if not cursor:
+                break
+
+        raise ValueError(
+            f"Slack channel '{name}' was not found. Use a joined channel name like "
+            f"'#general' or a concrete channel ID."
+        )
+
+    async def _resolve_user_handle(self, handle: str) -> str:
+        normalized = self._normalize_target_name(handle)
+        if not normalized:
+            raise ValueError("Slack target user handle is empty")
+
+        cache_key = f"user:{normalized}"
+        if cache_key in self._target_cache:
+            return self._target_cache[cache_key]
+
+        cursor: str | None = None
+        while True:
+            response = await self._web_client.users_list(limit=200, cursor=cursor)
+            for member in response.get("members", []):
+                if self._member_matches_handle(member, normalized):
+                    user_id = str(member.get("id") or "")
+                    if not user_id:
+                        continue
+                    dm_id = await self._open_dm_for_user(user_id)
+                    self._target_cache[cache_key] = dm_id
+                    return dm_id
+            cursor = ((response.get("response_metadata") or {}).get("next_cursor") or "").strip()
+            if not cursor:
+                break
+
+        raise ValueError(
+            f"Slack user '{handle}' was not found. Use '@name' or a concrete DM/channel ID."
+        )
+
+    async def _open_dm_for_user(self, user_id: str) -> str:
+        response = await self._web_client.conversations_open(users=user_id)
+        channel_id = str(((response.get("channel") or {}).get("id")) or "")
+        if not channel_id:
+            raise ValueError(f"Slack DM target for user '{user_id}' could not be opened.")
+        return channel_id
+
+    @staticmethod
+    def _normalize_target_name(value: str) -> str:
+        return value.strip().lstrip("#@").lower()
+
+    @classmethod
+    def _member_matches_handle(cls, member: dict[str, Any], normalized: str) -> bool:
+        profile = member.get("profile") or {}
+        candidates = {
+            str(member.get("name") or ""),
+            str(profile.get("display_name") or ""),
+            str(profile.get("display_name_normalized") or ""),
+            str(profile.get("real_name") or ""),
+            str(profile.get("real_name_normalized") or ""),
+        }
+        return normalized in {cls._normalize_target_name(candidate) for candidate in candidates if candidate}
+
     async def _on_socket_request(
         self,
         client: SocketModeClient,

nanobot/channels/telegram.py

@@ -166,6 +166,7 @@ def _markdown_to_telegram_html(text: str) -> str:
 
 _SEND_MAX_RETRIES = 3
 _SEND_RETRY_BASE_DELAY = 0.5  # seconds, doubled each retry
+_STREAM_EDIT_INTERVAL_DEFAULT = 0.6  # min seconds between edit_message_text calls
 
 
 @dataclass
@@ -190,6 +191,7 @@ class TelegramConfig(Base):
     connection_pool_size: int = 32
     pool_timeout: float = 5.0
     streaming: bool = True
+    stream_edit_interval: float = Field(default=_STREAM_EDIT_INTERVAL_DEFAULT, ge=0.1)
 
 
 class TelegramChannel(BaseChannel):
@@ -219,8 +221,6 @@ class TelegramChannel(BaseChannel):
     def default_config(cls) -> dict[str, Any]:
         return TelegramConfig().model_dump(by_alias=True)
 
-    _STREAM_EDIT_INTERVAL = 0.6  # min seconds between edit_message_text calls
-
     def __init__(self, config: Any, bus: MessageBus):
         if isinstance(config, dict):
             config = TelegramConfig.model_validate(config)
@@ -316,10 +316,10 @@ class TelegramChannel(BaseChannel):
         )
         self._app.add_handler(MessageHandler(filters.Regex(r"^/help(?:@\w+)?$"), self._on_help))
 
-        # Add message handler for text, photos, voice, documents
+        # Add message handler for text, photos, voice, documents, and locations
         self._app.add_handler(
             MessageHandler(
-                (filters.TEXT | filters.PHOTO | filters.VOICE | filters.AUDIO | filters.Document.ALL)
+                (filters.TEXT | filters.PHOTO | filters.VOICE | filters.AUDIO | filters.Document.ALL | filters.LOCATION)
                 & ~filters.COMMAND,
                 self._on_message
             )
@@ -520,7 +520,10 @@ class TelegramChannel(BaseChannel):
                 reply_parameters=reply_params,
                 **(thread_kwargs or {}),
             )
-        except Exception as e:
+        except BadRequest as e:
+            # Only fall back to plain text on actual HTML parse/format errors.
+            # Network errors (TimedOut, NetworkError) should propagate immediately
+            # to avoid doubling connection demand during pool exhaustion.
             logger.warning("HTML parse failed, falling back to plain text: {}", e)
             try:
                 await self._call_with_retry(
@@ -567,7 +570,10 @@ class TelegramChannel(BaseChannel):
                     chat_id=int_chat_id, message_id=buf.message_id,
                     text=html, parse_mode="HTML",
                 )
-            except Exception as e:
+            except BadRequest as e:
+                # Only fall back to plain text on actual HTML parse/format errors.
+                # Network errors (TimedOut, NetworkError) should propagate immediately
+                # to avoid doubling connection demand during pool exhaustion.
                 if self._is_not_modified_error(e):
                     logger.debug("Final stream edit already applied for {}", chat_id)
                     self._stream_bufs.pop(chat_id, None)
@@ -619,7 +625,7 @@ class TelegramChannel(BaseChannel):
             except Exception as e:
                 logger.warning("Stream initial send failed: {}", e)
                 raise  # Let ChannelManager handle retry
-        elif (now - buf.last_edit) >= self._STREAM_EDIT_INTERVAL:
+        elif (now - buf.last_edit) >= self.config.stream_edit_interval:
             try:
                 await self._call_with_retry(
                     self._app.bot.edit_message_text,
@@ -884,6 +890,12 @@ class TelegramChannel(BaseChannel):
         if message.caption:
             content_parts.append(message.caption)
 
+        # Location content
+        if message.location:
+            lat = message.location.latitude
+            lon = message.location.longitude
+            content_parts.append(f"[location: {lat}, {lon}]")
+
         # Download current message media
         current_media_paths, current_media_parts = await self._download_message_media(
             message, add_failure_content=True

nanobot/channels/wecom.py

@@ -302,13 +302,22 @@ class WecomChannel(BaseChannel):
 
             elif msg_type == "mixed":
                 # Mixed content contains multiple message items
-                msg_items = body.get("mixed", {}).get("item", [])
+                msg_items = body.get("mixed", {}).get("msg_item", [])
                 for item in msg_items:
-                    item_type = item.get("type", "")
+                    item_type = item.get("msgtype", "")
                     if item_type == "text":
                         text = item.get("text", {}).get("content", "")
                         if text:
                             content_parts.append(text)
+                    elif item_type == "image":
+                        file_url = item.get("image", {}).get("url", "")
+                        aes_key = item.get("image", {}).get("aeskey", "")
+                        if file_url and aes_key:
+                            file_path = await self._download_and_save_media(file_url, aes_key, "image")
+                            if file_path:
+                                filename = os.path.basename(file_path)
+                                content_parts.append(f"[image: {filename}]")
+                                media_paths.append(file_path)
                     else:
                         content_parts.append(MSG_TYPE_MAP.get(item_type, f"[{item_type}]"))
 

nanobot/channels/weixin.py

@@ -985,7 +985,43 @@ class WeixinChannel(BaseChannel):
             for media_path in (msg.media or []):
                 try:
                     await self._send_media_file(msg.chat_id, media_path, ctx_token)
+                except (httpx.TimeoutException, httpx.TransportError) as net_err:
+                    # Network/transport errors: do NOT fall back to text —
+                    # the text send would also likely fail, and the outer
+                    # except will re-raise so ChannelManager retries properly.
+                    logger.error(
+                        "Network error sending WeChat media {}: {}",
+                        media_path,
+                        net_err,
+                    )
+                    raise
+                except httpx.HTTPStatusError as http_err:
+                    status_code = (
+                        http_err.response.status_code
+                        if http_err.response is not None
+                        else 0
+                    )
+                    if status_code >= 500:
+                        # Server-side / retryable HTTP error — same as network.
+                        logger.error(
+                            "Server error ({} {}) sending WeChat media {}: {}",
+                            status_code,
+                            http_err.response.reason_phrase
+                            if http_err.response is not None
+                            else "",
+                            media_path,
+                            http_err,
+                        )
+                        raise
+                    # 4xx client errors are NOT retryable — fall back to text.
+                    filename = Path(media_path).name
+                    logger.error("Failed to send WeChat media {}: {}", media_path, http_err)
+                    await self._send_text(
+                        msg.chat_id, f"[Failed to send: {filename}]", ctx_token,
+                    )
                 except Exception as e:
+                    # Non-network errors (format, file-not-found, etc.):
+                    # notify the user via text fallback.
                     filename = Path(media_path).name
                     logger.error("Failed to send WeChat media {}: {}", media_path, e)
                     # Notify user about failure via text

nanobot/cli/commands.py

@@ -590,7 +590,11 @@ def serve(
         mcp_servers=runtime_config.tools.mcp_servers,
         channels_config=runtime_config.channels,
         timezone=runtime_config.agents.defaults.timezone,
+        unified_session=runtime_config.agents.defaults.unified_session,
+        disabled_skills=runtime_config.agents.defaults.disabled_skills,
         session_ttl_minutes=runtime_config.agents.defaults.session_ttl_minutes,
+        consolidation_ratio=runtime_config.agents.defaults.consolidation_ratio,
+        tools_config=runtime_config.tools,
     )
 
     model_name = runtime_config.agents.defaults.model
@@ -682,7 +686,11 @@ def gateway(
         mcp_servers=config.tools.mcp_servers,
         channels_config=config.channels,
         timezone=config.agents.defaults.timezone,
+        unified_session=config.agents.defaults.unified_session,
+        disabled_skills=config.agents.defaults.disabled_skills,
         session_ttl_minutes=config.agents.defaults.session_ttl_minutes,
+        consolidation_ratio=config.agents.defaults.consolidation_ratio,
+        tools_config=config.tools,
     )
 
     # Set cron callback (needs agent)
@@ -725,7 +733,7 @@ def gateway(
         response = resp.content if resp else ""
 
         message_tool = agent.tools.get("message")
-        if isinstance(message_tool, MessageTool) and message_tool._sent_in_turn:
+        if job.payload.deliver and isinstance(message_tool, MessageTool) and message_tool._sent_in_turn:
             return response
 
         if job.payload.deliver and job.payload.to and response:
@@ -817,12 +825,55 @@ def gateway(
 
     console.print(f"[green]✓[/green] Heartbeat: every {hb_cfg.interval_s}s")
 
+    async def _health_server(host: str, health_port: int):
+        """Lightweight HTTP health endpoint on the gateway port."""
+        import json as _json
+
+        async def handle(reader, writer):
+            try:
+                data = await asyncio.wait_for(reader.read(4096), timeout=5)
+            except (asyncio.TimeoutError, ConnectionError):
+                writer.close()
+                return
+
+            request_line = data.split(b"\r\n", 1)[0].decode("utf-8", errors="replace")
+            method, path = "", ""
+            parts = request_line.split(" ")
+            if len(parts) >= 2:
+                method, path = parts[0], parts[1]
+
+            if method == "GET" and path == "/health":
+                body = _json.dumps({"status": "ok"})
+                resp = (
+                    f"HTTP/1.0 200 OK\r\n"
+                    f"Content-Type: application/json\r\n"
+                    f"Content-Length: {len(body)}\r\n"
+                    f"\r\n{body}"
+                )
+            else:
+                body = "Not Found"
+                resp = (
+                    f"HTTP/1.0 404 Not Found\r\n"
+                    f"Content-Type: text/plain\r\n"
+                    f"Content-Length: {len(body)}\r\n"
+                    f"\r\n{body}"
+                )
+
+            writer.write(resp.encode())
+            await writer.drain()
+            writer.close()
+
+        server = await asyncio.start_server(handle, host, health_port)
+        console.print(f"[green]✓[/green] Health endpoint: http://{host}:{health_port}/health")
+        async with server:
+            await server.serve_forever()
     # Register Dream system job (always-on, idempotent on restart)
     dream_cfg = config.agents.defaults.dream
     if dream_cfg.model_override:
         agent.dream.model = dream_cfg.model_override
     agent.dream.max_batch_size = dream_cfg.max_batch_size
     agent.dream.max_iterations = dream_cfg.max_iterations
+    agent.dream.annotate_line_ages = dream_cfg.annotate_line_ages
     from nanobot.cron.types import CronJob, CronPayload
     cron.register_system_job(CronJob(
         id="dream",
@@ -839,6 +890,7 @@ def gateway(
             await asyncio.gather(
                 agent.run(),
                 channels.start_all(),
+                _health_server(config.gateway.host, port),
             )
         except KeyboardInterrupt:
             console.print("\nShutting down...")
@@ -914,7 +966,11 @@ def agent(
         mcp_servers=config.tools.mcp_servers,
         channels_config=config.channels,
         timezone=config.agents.defaults.timezone,
+        unified_session=config.agents.defaults.unified_session,
+        disabled_skills=config.agents.defaults.disabled_skills,
         session_ttl_minutes=config.agents.defaults.session_ttl_minutes,
+        consolidation_ratio=config.agents.defaults.consolidation_ratio,
+        tools_config=config.tools,
     )
     restart_notice = consume_restart_notice_from_env()
     if restart_notice and should_show_cli_restart_notice(restart_notice, session_id):
@@ -958,7 +1014,7 @@ def agent(
         # Interactive mode — route through bus like other channels
         from nanobot.bus.events import InboundMessage
         _init_prompt_session()
-        console.print(f"{__logo__} Interactive mode (type [bold]exit[/bold] or [bold]Ctrl+C[/bold] to quit)\n")
+        console.print(f"{__logo__} Interactive mode [bold blue]({config.agents.defaults.model})[/bold blue] — type [bold]exit[/bold] or [bold]Ctrl+C[/bold] to quit\n")
 
         if ":" in session_id:
             cli_channel, cli_chat_id = session_id.split(":", 1)
@@ -1119,7 +1175,7 @@ def channels_status(
 
     table = Table(title="Channel Status")
     table.add_column("Channel", style="cyan")
-    table.add_column("Enabled", style="green")
+    table.add_column("Enabled")
 
     for name, cls in sorted(discover_all().items()):
         section = getattr(config.channels, name, None)
@@ -1254,7 +1310,7 @@ def plugins_list():
     table = Table(title="Channel Plugins")
     table.add_column("Name", style="cyan")
     table.add_column("Source", style="magenta")
-    table.add_column("Enabled", style="green")
+    table.add_column("Enabled")
 
     for name in sorted(all_channels):
         cls = all_channels[name]

nanobot/cli/stream.py

@@ -102,7 +102,7 @@ class StreamRenderer:
             self._live = Live(self._render(), console=c, auto_refresh=False)
             self._live.start()
         now = time.monotonic()
-        if "\n" in delta or (now - self._t) > 0.05:
+        if (now - self._t) > 0.15:
             self._live.update(self._render())
             self._live.refresh()
             self._t = now

nanobot/command/builtin.py

@@ -74,6 +74,12 @@ async def cmd_status(ctx: CommandContext) -> OutboundMessage:
             search_usage_text = usage.format()
     except Exception:
         pass  # Never let usage fetch break /status
+    active_tasks = loop._active_tasks.get(ctx.key, [])
+    task_count = sum(1 for t in active_tasks if not t.done())
+    try:
+        task_count += loop.subagents.get_running_count_by_session(ctx.key)
+    except Exception:
+        pass
     return OutboundMessage(
         channel=ctx.msg.channel,
         chat_id=ctx.msg.chat_id,
@@ -84,6 +90,10 @@ async def cmd_status(ctx: CommandContext) -> OutboundMessage:
             session_msg_count=len(session.get_history(max_messages=0)),
             context_tokens_estimate=ctx_est,
             search_usage_text=search_usage_text,
+            active_task_count=task_count,
+            max_completion_tokens=getattr(
+                getattr(loop.provider, "generation", None), "max_tokens", 8192
+            ),
         ),
         metadata={**dict(ctx.msg.metadata or {}), "render_as": "text"},
     )

nanobot/config/loader.py

@@ -117,4 +117,19 @@ def _migrate_config(data: dict) -> dict:
     exec_cfg = tools.get("exec", {})
     if "restrictToWorkspace" in exec_cfg and "restrictToWorkspace" not in tools:
         tools["restrictToWorkspace"] = exec_cfg.pop("restrictToWorkspace")
+
+    # Move tools.myEnabled / tools.mySet → tools.my.{enable, allowSet}.
+    # The old flat keys shipped in the initial MyTool landing; wrapping them in a
+    # sub-config keeps `web` / `exec` / `my` symmetric and gives room to grow.
+    if "myEnabled" in tools or "mySet" in tools:
+        my_cfg = tools.setdefault("my", {})
+        if "myEnabled" in tools and "enable" not in my_cfg:
+            my_cfg["enable"] = tools.pop("myEnabled")
+        else:
+            tools.pop("myEnabled", None)
+        if "mySet" in tools and "allowSet" not in my_cfg:
+            my_cfg["allowSet"] = tools.pop("mySet")
+        else:
+            tools.pop("mySet", None)
+
     return data

nanobot/config/schema.py

@@ -43,7 +43,12 @@ class DreamConfig(Base):
         validation_alias=AliasChoices("modelOverride", "model", "model_override"),
     )  # Optional Dream-specific model override
     max_batch_size: int = Field(default=20, ge=1)  # Max history entries per run
-    max_iterations: int = Field(default=10, ge=1)  # Max tool calls per Phase 2
+    # Bumped from 10 to 15 in #3212 (exp002: +30% dedup, no accuracy loss; >15 plateaus).
+    max_iterations: int = Field(default=15, ge=1)  # Max tool calls per Phase 2
+    # Per-line git-blame age annotation in Phase 1 prompt (see #3212). Default
+    # on — set to False to feed MEMORY.md raw if a specific LLM reacts poorly
+    # to the `← Nd` suffix or you want deterministic, git-independent prompts.
+    annotate_line_ages: bool = True
 
     def build_schedule(self, timezone: str) -> CronSchedule:
         """Build the runtime schedule, preferring the legacy cron override if present."""
@@ -74,9 +79,23 @@ class AgentDefaults(Base):
     max_tool_iterations: int = 200
     max_tool_result_chars: int = 16_000
     provider_retry_mode: Literal["standard", "persistent"] = "standard"
-    reasoning_effort: str | None = None  # low / medium / high - enables LLM thinking mode
+    reasoning_effort: str | None = None  # low / medium / high / adaptive - enables LLM thinking mode
     timezone: str = "UTC"  # IANA timezone, e.g. "Asia/Shanghai", "America/New_York"
-    session_ttl_minutes: int = Field(default=0, ge=0)  # Auto /new after idle (0 = disabled)
+    unified_session: bool = False  # Share one session across all channels (single-user multi-device)
+    disabled_skills: list[str] = Field(default_factory=list)  # Skill names to exclude from loading (e.g. ["summarize", "skill-creator"])
+    session_ttl_minutes: int = Field(
+        default=0,
+        ge=0,
+        validation_alias=AliasChoices("idleCompactAfterMinutes", "sessionTtlMinutes"),
+        serialization_alias="idleCompactAfterMinutes",
+    )  # Auto-compact idle threshold in minutes (0 = disabled)
+    consolidation_ratio: float = Field(
+        default=0.5,
+        ge=0.1,
+        le=0.95,
+        validation_alias=AliasChoices("consolidationRatio"),
+        serialization_alias="consolidationRatio",
+    )  # Consolidation target ratio (0.5 = 50% of budget retained after compression)
     dream: DreamConfig = Field(default_factory=DreamConfig)
 
 
@@ -89,7 +108,7 @@ class AgentsConfig(Base):
 class ProviderConfig(Base):
     """LLM provider configuration."""
 
-    api_key: str = ""
+    api_key: str | None = None
     api_base: str | None = None
     extra_headers: dict[str, str] | None = None  # Custom headers (e.g. APP-Code for AiHubMix)
 
@@ -108,10 +127,12 @@ class ProvidersConfig(Base):
     dashscope: ProviderConfig = Field(default_factory=ProviderConfig)
     vllm: ProviderConfig = Field(default_factory=ProviderConfig)
     ollama: ProviderConfig = Field(default_factory=ProviderConfig)  # Ollama local models
+    lm_studio: ProviderConfig = Field(default_factory=ProviderConfig)  # LM Studio local models
     ovms: ProviderConfig = Field(default_factory=ProviderConfig)  # OpenVINO Model Server (OVMS)
     gemini: ProviderConfig = Field(default_factory=ProviderConfig)
     moonshot: ProviderConfig = Field(default_factory=ProviderConfig)
     minimax: ProviderConfig = Field(default_factory=ProviderConfig)
+    minimax_anthropic: ProviderConfig = Field(default_factory=ProviderConfig)  # MiniMax Anthropic endpoint (thinking)
     mistral: ProviderConfig = Field(default_factory=ProviderConfig)
     stepfun: ProviderConfig = Field(default_factory=ProviderConfig)  # Step Fun (阶跃星辰)
     xiaomi_mimo: ProviderConfig = Field(default_factory=ProviderConfig)  # Xiaomi MIMO (小米)
@@ -145,7 +166,7 @@ class ApiConfig(Base):
 class GatewayConfig(Base):
     """Gateway/server configuration."""
 
-    host: str = "0.0.0.0"
+    host: str = "127.0.0.1"  # Safer default: local-only bind.
     port: int = 18790
     heartbeat: HeartbeatConfig = Field(default_factory=HeartbeatConfig)
 
@@ -153,7 +174,7 @@ class GatewayConfig(Base):
 class WebSearchConfig(Base):
     """Web search tool configuration."""
 
-    provider: str = "duckduckgo"  # brave, tavily, duckduckgo, searxng, jina
+    provider: str = "duckduckgo"  # brave, tavily, duckduckgo, searxng, jina, kagi
     api_key: str = ""
     base_url: str = ""  # SearXNG base URL
     max_results: int = 5
@@ -177,6 +198,7 @@ class ExecToolConfig(Base):
     timeout: int = 60
     path_append: str = ""
     sandbox: str = ""  # sandbox backend: "" (none) or "bwrap"
+    allowed_env_keys: list[str] = Field(default_factory=list)  # Env var names to pass through to subprocess (e.g. ["GOPATH", "JAVA_HOME"])
 
 class MCPServerConfig(Base):
     """MCP server connection configuration (stdio or HTTP)."""
@@ -190,11 +212,19 @@ class MCPServerConfig(Base):
     tool_timeout: int = 30  # seconds before a tool call is cancelled
     enabled_tools: list[str] = Field(default_factory=lambda: ["*"])  # Only register these tools; accepts raw MCP names or wrapped mcp_<server>_<tool> names; ["*"] = all tools; [] = no tools
 
+class MyToolConfig(Base):
+    """Self-inspection tool configuration."""
+
+    enable: bool = True  # register the `my` tool (agent runtime state inspection)
+    allow_set: bool = False  # let `my` modify loop state (read-only if False)
+
+
 class ToolsConfig(Base):
     """Tools configuration."""
 
     web: WebToolsConfig = Field(default_factory=WebToolsConfig)
     exec: ExecToolConfig = Field(default_factory=ExecToolConfig)
+    my: MyToolConfig = Field(default_factory=MyToolConfig)
     restrict_to_workspace: bool = False  # restrict all tool access to workspace directory
     mcp_servers: dict[str, MCPServerConfig] = Field(default_factory=dict)
     ssrf_whitelist: list[str] = Field(default_factory=list)  # CIDR ranges to exempt from SSRF blocking (e.g. ["100.64.0.0/10"] for Tailscale)

nanobot/cron/service.py

@@ -4,10 +4,12 @@ import asyncio
 import json
 import time
 import uuid
+from dataclasses import asdict
 from datetime import datetime
 from pathlib import Path
 from typing import Any, Callable, Coroutine, Literal
 
+from filelock import FileLock
 from loguru import logger
 
 from nanobot.cron.types import CronJob, CronJobState, CronPayload, CronRunRecord, CronSchedule, CronStore
@@ -69,28 +71,26 @@ class CronService:
         self,
         store_path: Path,
         on_job: Callable[[CronJob], Coroutine[Any, Any, str | None]] | None = None,
+        max_sleep_ms: int = 300_000,  # 5 minutes
     ):
         self.store_path = store_path
+        self._action_path = store_path.parent / "action.jsonl"
+        self._lock = FileLock(str(self._action_path.parent) + ".lock")
         self.on_job = on_job
         self._store: CronStore | None = None
-        self._last_mtime: float = 0.0
         self._timer_task: asyncio.Task | None = None
         self._running = False
+        self._timer_active = False
+        self.max_sleep_ms = max_sleep_ms
 
-    def _load_store(self) -> CronStore:
-        """Load jobs from disk. Reloads automatically if file was modified externally."""
-        if self._store and self.store_path.exists():
-            mtime = self.store_path.stat().st_mtime
-            if mtime != self._last_mtime:
-                logger.info("Cron: jobs.json modified externally, reloading")
-                self._store = None
-        if self._store:
-            return self._store
-
+    def _load_jobs(self) -> tuple[list[CronJob], int]:
+        jobs = []
+        version = 1
         if self.store_path.exists():
             try:
                 data = json.loads(self.store_path.read_text(encoding="utf-8"))
                 jobs = []
+                version = data.get("version", 1)
                 for j in data.get("jobs", []):
                     jobs.append(CronJob(
                         id=j["id"],
@@ -129,12 +129,57 @@ class CronService:
                         updated_at_ms=j.get("updatedAtMs", 0),
                         delete_after_run=j.get("deleteAfterRun", False),
                     ))
-                self._store = CronStore(jobs=jobs)
             except Exception as e:
                 logger.warning("Failed to load cron store: {}", e)
-                self._store = CronStore()
-        else:
-            self._store = CronStore()
+        return jobs, version
+
+    def _merge_action(self):
+        if not self._action_path.exists():
+            return
+
+        jobs_map = {j.id: j for j in self._store.jobs}
+        def _update(params: dict):
+            j = CronJob.from_dict(params)
+            jobs_map[j.id] = j
+
+        def _del(params: dict):
+            if job_id := params.get("job_id"):
+                jobs_map.pop(job_id)
+
+        with self._lock:
+            with open(self._action_path, "r", encoding="utf-8") as f:
+                changed = False
+                for line in f:
+                    try:
+                        line = line.strip()
+                        action = json.loads(line)
+                        if "action" not in action:
+                            continue
+                        if action["action"] == "del":
+                            _del(action.get("params", {}))
+                        else:
+                            _update(action.get("params", {}))
+                        changed = True
+                    except Exception as exp:
+                        logger.debug(f"load action line error: {exp}")
+                        continue
+            self._store.jobs = list(jobs_map.values())
+            if self._running and changed:
+                self._action_path.write_text("", encoding="utf-8")
+                self._save_store()
+        return
+
+    def _load_store(self) -> CronStore:
+        """Load jobs from disk. Reloads automatically if file was modified externally.
+        - Reload every time because it needs to merge operations on the jobs object from other instances.
+        - During _on_timer execution, return the existing store to prevent concurrent
+          _load_store calls (e.g. from list_jobs polling) from replacing it mid-execution.
+        """
+        if self._timer_active and self._store:
+            return self._store
+        jobs, version = self._load_jobs()
+        self._store = CronStore(version=version, jobs=jobs)
+        self._merge_action()
 
         return self._store
 
@@ -190,8 +235,7 @@ class CronService:
         }
 
         self.store_path.write_text(json.dumps(data, indent=2, ensure_ascii=False), encoding="utf-8")
-        self._last_mtime = self.store_path.stat().st_mtime
-    
+
     async def start(self) -> None:
         """Start the cron service."""
         self._running = True
@@ -230,11 +274,14 @@ class CronService:
         if self._timer_task:
             self._timer_task.cancel()
 
-        next_wake = self._get_next_wake_ms()
-        if not next_wake or not self._running:
+        if not self._running:
             return
 
-        delay_ms = max(0, next_wake - _now_ms())
+        next_wake = self._get_next_wake_ms()
+        if next_wake is None:
+            delay_ms = self.max_sleep_ms
+        else:
+            delay_ms = min(self.max_sleep_ms, max(0, next_wake - _now_ms()))
         delay_s = delay_ms / 1000
 
         async def tick():
@@ -248,18 +295,23 @@ class CronService:
         """Handle timer tick - run due jobs."""
         self._load_store()
         if not self._store:
+            self._arm_timer()
             return
 
-        now = _now_ms()
-        due_jobs = [
-            j for j in self._store.jobs
-            if j.enabled and j.state.next_run_at_ms and now >= j.state.next_run_at_ms
-        ]
+        self._timer_active = True
+        try:
+            now = _now_ms()
+            due_jobs = [
+                j for j in self._store.jobs
+                if j.enabled and j.state.next_run_at_ms and now >= j.state.next_run_at_ms
+            ]
 
-        for job in due_jobs:
-            await self._execute_job(job)
+            for job in due_jobs:
+                await self._execute_job(job)
 
-        self._save_store()
+            self._save_store()
+        finally:
+            self._timer_active = False
         self._arm_timer()
 
     async def _execute_job(self, job: CronJob) -> None:
@@ -303,6 +355,13 @@ class CronService:
             # Compute next run
             job.state.next_run_at_ms = _compute_next_run(job.schedule, _now_ms())
 
+    def _append_action(self, action: Literal["add", "del", "update"], params: dict):
+        self.store_path.parent.mkdir(parents=True, exist_ok=True)
+        with self._lock:
+            with open(self._action_path, "a", encoding="utf-8") as f:
+                f.write(json.dumps({"action": action, "params": params}, ensure_ascii=False) + "\n")
+
+
     # ========== Public API ==========
 
     def list_jobs(self, include_disabled: bool = False) -> list[CronJob]:
@@ -322,7 +381,6 @@ class CronService:
         delete_after_run: bool = False,
     ) -> CronJob:
         """Add a new job."""
-        store = self._load_store()
         _validate_schedule_for_add(schedule)
         now = _now_ms()
 
@@ -343,10 +401,13 @@ class CronService:
             updated_at_ms=now,
             delete_after_run=delete_after_run,
         )
-
-        store.jobs.append(job)
-        self._save_store()
-        self._arm_timer()
+        if self._running:
+            store = self._load_store()
+            store.jobs.append(job)
+            self._save_store()
+            self._arm_timer()
+        else:
+            self._append_action("add", asdict(job))
 
         logger.info("Cron: added job '{}' ({})", name, job.id)
         return job
@@ -380,8 +441,11 @@ class CronService:
         removed = len(store.jobs) < before
 
         if removed:
-            self._save_store()
-            self._arm_timer()
+            if self._running:
+                self._save_store()
+                self._arm_timer()
+            else:
+                self._append_action("del", {"job_id": job_id})
             logger.info("Cron: removed job {}", job_id)
             return "removed"
 
@@ -398,23 +462,85 @@ class CronService:
                     job.state.next_run_at_ms = _compute_next_run(job.schedule, _now_ms())
                 else:
                     job.state.next_run_at_ms = None
-                self._save_store()
-                self._arm_timer()
+                if self._running:
+                    self._save_store()
+                    self._arm_timer()
+                else:
+                    self._append_action("update", asdict(job))
                 return job
         return None
 
-    async def run_job(self, job_id: str, force: bool = False) -> bool:
-        """Manually run a job."""
+    def update_job(
+        self,
+        job_id: str,
+        *,
+        name: str | None = None,
+        schedule: CronSchedule | None = None,
+        message: str | None = None,
+        deliver: bool | None = None,
+        channel: str | None = ...,
+        to: str | None = ...,
+        delete_after_run: bool | None = None,
+    ) -> CronJob | Literal["not_found", "protected"]:
+        """Update mutable fields of an existing job. System jobs cannot be updated.
+
+        For ``channel`` and ``to``, pass an explicit value (including ``None``)
+        to update; omit (sentinel ``...``) to leave unchanged.
+        """
         store = self._load_store()
-        for job in store.jobs:
-            if job.id == job_id:
-                if not force and not job.enabled:
-                    return False
-                await self._execute_job(job)
-                self._save_store()
+        job = next((j for j in store.jobs if j.id == job_id), None)
+        if job is None:
+            return "not_found"
+        if job.payload.kind == "system_event":
+            return "protected"
+
+        if schedule is not None:
+            _validate_schedule_for_add(schedule)
+            job.schedule = schedule
+        if name is not None:
+            job.name = name
+        if message is not None:
+            job.payload.message = message
+        if deliver is not None:
+            job.payload.deliver = deliver
+        if channel is not ...:
+            job.payload.channel = channel
+        if to is not ...:
+            job.payload.to = to
+        if delete_after_run is not None:
+            job.delete_after_run = delete_after_run
+
+        job.updated_at_ms = _now_ms()
+        if job.enabled:
+            job.state.next_run_at_ms = _compute_next_run(job.schedule, _now_ms())
+
+        if self._running:
+            self._save_store()
+            self._arm_timer()
+        else:
+            self._append_action("update", asdict(job))
+
+        logger.info("Cron: updated job '{}' ({})", job.name, job.id)
+        return job
+
+    async def run_job(self, job_id: str, force: bool = False) -> bool:
+        """Manually run a job without disturbing the service's running state."""
+        was_running = self._running
+        self._running = True
+        try:
+            store = self._load_store()
+            for job in store.jobs:
+                if job.id == job_id:
+                    if not force and not job.enabled:
+                        return False
+                    await self._execute_job(job)
+                    self._save_store()
+                    return True
+            return False
+        finally:
+            self._running = was_running
+            if was_running:
                 self._arm_timer()
-                return True
-        return False
 
     def get_job(self, job_id: str) -> CronJob | None:
         """Get a job by ID."""

nanobot/cron/types.py

@@ -61,6 +61,18 @@ class CronJob:
     updated_at_ms: int = 0
     delete_after_run: bool = False
 
+    @classmethod
+    def from_dict(cls, kwargs: dict):
+        state_kwargs = dict(kwargs.get("state", {}))
+        state_kwargs["run_history"] = [
+            record if isinstance(record, CronRunRecord) else CronRunRecord(**record)
+            for record in state_kwargs.get("run_history", [])
+        ]
+        kwargs["schedule"] = CronSchedule(**kwargs.get("schedule", {"kind": "every"}))
+        kwargs["payload"] = CronPayload(**kwargs.get("payload", {}))
+        kwargs["state"] = CronJobState(**state_kwargs)
+        return cls(**kwargs)
+
 
 @dataclass
 class CronStore:

nanobot/nanobot.py

@@ -81,7 +81,11 @@ class Nanobot:
             restrict_to_workspace=config.tools.restrict_to_workspace,
             mcp_servers=config.tools.mcp_servers,
             timezone=defaults.timezone,
+            unified_session=defaults.unified_session,
+            disabled_skills=defaults.disabled_skills,
             session_ttl_minutes=defaults.session_ttl_minutes,
+            consolidation_ratio=defaults.consolidation_ratio,
+            tools_config=config.tools,
         )
         return cls(loop)
 

nanobot/providers/anthropic_provider.py

@@ -380,9 +380,15 @@ class AnthropicProvider(LLMProvider):
         if system:
             kwargs["system"] = system
 
-        if thinking_enabled:
+        if reasoning_effort == "adaptive":
+            # Adaptive thinking: model decides when and how much to think
+            # Supported on claude-sonnet-4-6 and claude-opus-4-6.
+            # Also auto-enables interleaved thinking between tool calls.
+            kwargs["thinking"] = {"type": "adaptive"}
+            kwargs["temperature"] = 1.0
+        elif thinking_enabled:
             budget_map = {"low": 1024, "medium": 4096, "high": max(8192, max_tokens)}
-            budget = budget_map.get(reasoning_effort.lower(), 4096)  # type: ignore[union-attr]
+            budget = budget_map.get(reasoning_effort.lower(), 4096)
             kwargs["thinking"] = {"type": "enabled", "budget_tokens": budget}
             kwargs["max_tokens"] = max(max_tokens, budget + 4096)
             kwargs["temperature"] = 1.0

nanobot/providers/base.py

@@ -353,6 +353,64 @@ class LLMProvider(ABC):
         # Unknown 429 defaults to WAIT+retry.
         return True
 
+    @staticmethod
+    def _enforce_role_alternation(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Merge consecutive same-role messages and drop trailing assistant messages.
+
+        Some providers (OpenAI-compat, Azure, vLLM, Ollama, etc.) reject requests
+        where the last message is 'assistant' (prefill not supported) or two
+        consecutive non-system messages share the same role.
+        """
+        if not messages:
+            return messages
+
+        merged: list[dict[str, Any]] = []
+        for msg in messages:
+            role = msg.get("role")
+            if (
+                merged
+                and role != "system"
+                and role not in ("tool",)
+                and merged[-1].get("role") == role
+                and role in ("user", "assistant")
+            ):
+                prev = merged[-1]
+                if role == "assistant":
+                    prev_has_tools = bool(prev.get("tool_calls"))
+                    curr_has_tools = bool(msg.get("tool_calls"))
+                    if curr_has_tools:
+                        merged[-1] = dict(msg)
+                        continue
+                    if prev_has_tools:
+                        continue
+                prev_content = prev.get("content") or ""
+                curr_content = msg.get("content") or ""
+                if isinstance(prev_content, str) and isinstance(curr_content, str):
+                    prev["content"] = (prev_content + "\n\n" + curr_content).strip()
+                else:
+                    merged[-1] = dict(msg)
+            else:
+                merged.append(dict(msg))
+
+        last_popped = None
+        while merged and merged[-1].get("role") == "assistant":
+            last_popped = merged.pop()
+
+        # If removing trailing assistant messages left only system messages,
+        # the request would be invalid for most providers (e.g. Zhipu/GLM
+        # error 1214).  Recover by converting the last popped assistant
+        # message to a user message so the LLM can still see the content.
+        if (
+            merged
+            and last_popped is not None
+            and not any(m.get("role") in ("user", "tool") for m in merged)
+        ):
+            recovered = dict(last_popped)
+            recovered["role"] = "user"
+            merged.append(recovered)
+
+        return merged
+
     @staticmethod
     def _strip_image_content(messages: list[dict[str, Any]]) -> list[dict[str, Any]] | None:
         """Replace image_url blocks with text placeholder. Returns None if no images found."""
@@ -375,6 +433,26 @@ class LLMProvider(ABC):
                 result.append(msg)
         return result if found else None
 
+    @staticmethod
+    def _strip_image_content_inplace(messages: list[dict[str, Any]]) -> bool:
+        """Replace image_url blocks with text placeholder *in-place*.
+
+        Mutates the content lists of the original message dicts so that
+        callers holding references to those dicts also see the stripped
+        version.
+        """
+        found = False
+        for msg in messages:
+            content = msg.get("content")
+            if isinstance(content, list):
+                for i, b in enumerate(content):
+                    if isinstance(b, dict) and b.get("type") == "image_url":
+                        path = (b.get("_meta") or {}).get("path", "")
+                        placeholder = image_placeholder_text(path, empty="[image omitted]")
+                        content[i] = {"type": "text", "text": placeholder}
+                        found = True
+        return found
+
     async def _safe_chat(self, **kwargs: Any) -> LLMResponse:
         """Call chat() and convert unexpected exceptions to error responses."""
         try:
@@ -434,9 +512,9 @@ class LLMProvider(ABC):
         on_retry_wait: Callable[[str], Awaitable[None]] | None = None,
     ) -> LLMResponse:
         """Call chat_stream() with retry on transient provider failures."""
-        if max_tokens is self._SENTINEL:
+        if max_tokens is self._SENTINEL or max_tokens is None:
             max_tokens = self.generation.max_tokens
-        if temperature is self._SENTINEL:
+        if temperature is self._SENTINEL or temperature is None:
             temperature = self.generation.temperature
         if reasoning_effort is self._SENTINEL:
             reasoning_effort = self.generation.reasoning_effort
@@ -471,11 +549,14 @@ class LLMProvider(ABC):
 
         Parameters default to ``self.generation`` when not explicitly passed,
         so callers no longer need to thread temperature / max_tokens /
-        reasoning_effort through every layer.
+        reasoning_effort through every layer. Explicit ``None`` is also
+        normalized to the provider's generation defaults so that downstream
+        ``_build_kwargs`` never sees ``None`` for ``max_tokens`` / ``temperature``
+        (which would crash ``max(1, max_tokens)``).
         """
-        if max_tokens is self._SENTINEL:
+        if max_tokens is self._SENTINEL or max_tokens is None:
             max_tokens = self.generation.max_tokens
-        if temperature is self._SENTINEL:
+        if temperature is self._SENTINEL or temperature is None:
             temperature = self.generation.temperature
         if reasoning_effort is self._SENTINEL:
             reasoning_effort = self.generation.reasoning_effort
@@ -626,7 +707,12 @@ class LLMProvider(ABC):
                     )
                     retry_kw = dict(kw)
                     retry_kw["messages"] = stripped
-                    return await call(**retry_kw)
+                    result = await call(**retry_kw)
+                    # Permanently strip images from the original messages so
+                    # subsequent iterations do not repeat the error-retry cycle.
+                    if result.finish_reason != "error":
+                        self._strip_image_content_inplace(original_messages)
+                    return result
                 return response
 
             if persistent and identical_error_count >= self._PERSISTENT_IDENTICAL_ERROR_LIMIT:
@@ -635,9 +721,22 @@ class LLMProvider(ABC):
                     identical_error_count,
                     (response.content or "")[:120].lower(),
                 )
+                if on_retry_wait:
+                    await on_retry_wait(
+                        f"Persistent retry stopped after {identical_error_count} identical errors."
+                    )
                 return response
 
             if not persistent and attempt > len(delays):
+                logger.warning(
+                    "LLM request failed after {} retries, giving up: {}",
+                    attempt,
+                    (response.content or "")[:120].lower(),
+                )
+                if on_retry_wait:
+                    await on_retry_wait(
+                        f"Model request failed after {attempt} retries, giving up."
+                    )
                 break
 
             base_delay = delays[min(attempt - 1, len(delays) - 1)]

nanobot/providers/openai_compat_provider.py

@@ -3,16 +3,19 @@
 from __future__ import annotations
 
 import asyncio
+import json
 import hashlib
 import importlib.util
 import os
 import secrets
 import string
+import time
 import uuid
 from collections.abc import Awaitable, Callable
 from typing import TYPE_CHECKING, Any
 
 import json_repair
+from loguru import logger
 
 if os.environ.get("LANGFUSE_SECRET_KEY") and importlib.util.find_spec("langfuse"):
     from langfuse.openai import AsyncOpenAI
@@ -26,6 +29,12 @@ else:
     from openai import AsyncOpenAI
 
 from nanobot.providers.base import LLMProvider, LLMResponse, ToolCallRequest
+from nanobot.providers.openai_responses import (
+    consume_sdk_stream,
+    convert_messages,
+    convert_tools,
+    parse_response_output,
+)
 
 if TYPE_CHECKING:
     from nanobot.providers.registry import ProviderSpec
@@ -43,6 +52,29 @@ _DEFAULT_OPENROUTER_HEADERS = {
     "X-OpenRouter-Title": "nanobot",
     "X-OpenRouter-Categories": "cli-agent,personal-agent",
 }
+_KIMI_THINKING_MODELS: frozenset[str] = frozenset({
+    "kimi-k2.5",
+    "k2.6-code-preview",
+})
+
+
+def _is_kimi_thinking_model(model_name: str) -> bool:
+    """Return True if model_name refers to a Kimi thinking-capable model.
+
+    Supports two forms:
+    - Exact match: kimi-k2.5 in _KIMI_THINKING_MODELS
+    - Slug match:  moonshotai/kimi-k2.5 -> the part after the last "/"
+                   is checked against _KIMI_THINKING_MODELS
+
+    This covers both the native Moonshot provider (bare slug) and
+    OpenRouter-style names (``"publisher/slug"``).
+    """
+    name = model_name.lower()
+    if name in _KIMI_THINKING_MODELS:
+        return True
+    if "/" in name and name.rsplit("/", 1)[1] in _KIMI_THINKING_MODELS:
+        return True
+    return False
 
 
 def _short_tool_id() -> str:
@@ -113,6 +145,18 @@ def _uses_openrouter_attribution(spec: "ProviderSpec | None", api_base: str | No
     return bool(api_base and "openrouter" in api_base.lower())
 
 
+_RESPONSES_FAILURE_THRESHOLD = 3
+_RESPONSES_PROBE_INTERVAL_S = 300  # 5 minutes
+
+
+def _is_direct_openai_base(api_base: str | None) -> bool:
+    """Return True for direct OpenAI endpoints, not generic OpenAI-compatible gateways."""
+    if not api_base:
+        return True
+    normalized = api_base.strip().lower().rstrip("/")
+    return "api.openai.com" in normalized and "openrouter" not in normalized
+
+
 class OpenAICompatProvider(LLMProvider):
     """Unified provider for all OpenAI-compatible APIs.
 
@@ -137,6 +181,7 @@ class OpenAICompatProvider(LLMProvider):
             self._setup_env(api_key, api_base)
 
         effective_base = api_base or (spec.default_api_base if spec else None) or None
+        self._effective_base = effective_base
         default_headers = {"x-session-affinity": uuid.uuid4().hex}
         if _uses_openrouter_attribution(spec, effective_base):
             default_headers.update(_DEFAULT_OPENROUTER_HEADERS)
@@ -150,6 +195,11 @@ class OpenAICompatProvider(LLMProvider):
             max_retries=0,
         )
 
+        # Responses API circuit breaker: skip after repeated failures,
+        # probe again after _RESPONSES_PROBE_INTERVAL_S seconds.
+        self._responses_failures: dict[str, int] = {}
+        self._responses_tripped_at: dict[str, float] = {}
+
     def _setup_env(self, api_key: str, api_base: str | None) -> None:
         """Set environment variables based on provider spec."""
         spec = self._spec
@@ -207,6 +257,24 @@ class OpenAICompatProvider(LLMProvider):
             return tool_call_id
         return hashlib.sha1(tool_call_id.encode()).hexdigest()[:9]
 
+    @staticmethod
+    def _normalize_tool_call_arguments(arguments: Any) -> str:
+        """Force function.arguments into a valid JSON object string."""
+        if isinstance(arguments, str):
+            stripped = arguments.strip()
+            if not stripped:
+                return "{}"
+            try:
+                parsed = json_repair.loads(stripped)
+            except Exception:
+                return "{}"
+            if isinstance(parsed, dict):
+                return json.dumps(parsed, ensure_ascii=False)
+            return "{}"
+        if isinstance(arguments, dict):
+            return json.dumps(arguments, ensure_ascii=False)
+        return "{}"
+
     def _sanitize_messages(self, messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
         """Strip non-standard keys, normalize tool_call IDs."""
         sanitized = LLMProvider._sanitize_request_messages(messages, _ALLOWED_MSG_KEYS)
@@ -226,11 +294,25 @@ class OpenAICompatProvider(LLMProvider):
                         continue
                     tc_clean = dict(tc)
                     tc_clean["id"] = map_id(tc_clean.get("id"))
+                    function = tc_clean.get("function")
+                    if isinstance(function, dict):
+                        function_clean = dict(function)
+                        if "arguments" in function_clean:
+                            function_clean["arguments"] = self._normalize_tool_call_arguments(
+                                function_clean.get("arguments")
+                            )
+                        else:
+                            function_clean["arguments"] = "{}"
+                        tc_clean["function"] = function_clean
                     normalized.append(tc_clean)
                 clean["tool_calls"] = normalized
+                if clean.get("role") == "assistant":
+                    # Some OpenAI-compatible gateways reject assistant messages
+                    # that mix non-empty content with tool_calls.
+                    clean["content"] = None
             if "tool_call_id" in clean and clean["tool_call_id"]:
                 clean["tool_call_id"] = map_id(clean["tool_call_id"])
-        return sanitized
+        return self._enforce_role_alternation(sanitized)
 
     # ------------------------------------------------------------------
     # Build kwargs
@@ -315,12 +397,134 @@ class OpenAICompatProvider(LLMProvider):
             if extra:
                 kwargs.setdefault("extra_body", {}).update(extra)
 
+        # Model-level thinking injection for Kimi thinking-capable models.
+        # Strip any provider prefix (e.g. "moonshotai/") before the set lookup
+        # so that OpenRouter-style names like "moonshotai/kimi-k2.5" are handled
+        # identically to bare names like "kimi-k2.5".
+        if reasoning_effort is not None and _is_kimi_thinking_model(model_name):
+            thinking_enabled = reasoning_effort.lower() != "minimal"
+            kwargs.setdefault("extra_body", {}).update(
+                {"thinking": {"type": "enabled" if thinking_enabled else "disabled"}}
+            )
+
         if tools:
             kwargs["tools"] = tools
             kwargs["tool_choice"] = tool_choice or "auto"
 
         return kwargs
 
+    def _should_use_responses_api(
+        self,
+        model: str | None,
+        reasoning_effort: str | None,
+    ) -> bool:
+        """Use Responses API only for direct OpenAI requests that benefit from it."""
+        if self._spec and self._spec.name != "openai":
+            return False
+        if not _is_direct_openai_base(self._effective_base):
+            return False
+
+        model_name = (model or self.default_model).lower()
+        wants = False
+        if reasoning_effort and reasoning_effort.lower() != "none":
+            wants = True
+        elif any(token in model_name for token in ("gpt-5", "o1", "o3", "o4")):
+            wants = True
+        if not wants:
+            return False
+
+        # Circuit breaker: skip after repeated failures, probe periodically.
+        key = f"{model_name}:{reasoning_effort or ''}"
+        failures = self._responses_failures.get(key, 0)
+        if failures >= _RESPONSES_FAILURE_THRESHOLD:
+            tripped = self._responses_tripped_at.get(key, 0.0)
+            if (time.monotonic() - tripped) < _RESPONSES_PROBE_INTERVAL_S:
+                return False
+            # Half-open: allow one probe attempt
+        return True
+
+    def _record_responses_failure(self, model: str | None, reasoning_effort: str | None) -> None:
+        key = f"{(model or self.default_model).lower()}:{reasoning_effort or ''}"
+        count = self._responses_failures.get(key, 0) + 1
+        self._responses_failures[key] = count
+        if count >= _RESPONSES_FAILURE_THRESHOLD:
+            self._responses_tripped_at[key] = time.monotonic()
+            logger.warning(
+                "Responses API circuit open for {} — falling back to Chat Completions",
+                key,
+            )
+
+    def _record_responses_success(self, model: str | None, reasoning_effort: str | None) -> None:
+        key = f"{(model or self.default_model).lower()}:{reasoning_effort or ''}"
+        self._responses_failures.pop(key, None)
+        self._responses_tripped_at.pop(key, None)
+
+    @staticmethod
+    def _should_fallback_from_responses_error(e: Exception) -> bool:
+        """Fallback only for likely Responses API compatibility errors."""
+        response = getattr(e, "response", None)
+        status_code = getattr(e, "status_code", None)
+        if status_code is None and response is not None:
+            status_code = getattr(response, "status_code", None)
+        if status_code not in {400, 404, 422}:
+            return False
+
+        body = (
+            getattr(e, "body", None)
+            or getattr(e, "doc", None)
+            or getattr(response, "text", None)
+        )
+        body_text = str(body).lower() if body is not None else ""
+        compatibility_markers = (
+            "responses",
+            "response api",
+            "max_output_tokens",
+            "instructions",
+            "previous_response",
+            "unsupported",
+            "not supported",
+            "unknown parameter",
+            "unrecognized request argument",
+        )
+        return any(marker in body_text for marker in compatibility_markers)
+
+    def _build_responses_body(
+        self,
+        messages: list[dict[str, Any]],
+        tools: list[dict[str, Any]] | None,
+        model: str | None,
+        max_tokens: int,
+        temperature: float,
+        reasoning_effort: str | None,
+        tool_choice: str | dict[str, Any] | None,
+    ) -> dict[str, Any]:
+        """Build a Responses API body for direct OpenAI requests."""
+        model_name = model or self.default_model
+        sanitized_messages = self._sanitize_messages(self._sanitize_empty_content(messages))
+        instructions, input_items = convert_messages(sanitized_messages)
+
+        body: dict[str, Any] = {
+            "model": model_name,
+            "instructions": instructions or None,
+            "input": input_items,
+            "max_output_tokens": max(1, max_tokens),
+            "store": False,
+            "stream": False,
+        }
+
+        if self._supports_temperature(model_name, reasoning_effort):
+            body["temperature"] = temperature
+
+        if reasoning_effort and reasoning_effort.lower() != "none":
+            body["reasoning"] = {"effort": reasoning_effort}
+            body["include"] = ["reasoning.encrypted_content"]
+
+        if tools:
+            body["tools"] = convert_tools(tools)
+            body["tool_choice"] = tool_choice or "auto"
+
+        return body
+
     # ------------------------------------------------------------------
     # Response parsing
     # ------------------------------------------------------------------
@@ -455,7 +659,12 @@ class OpenAICompatProvider(LLMProvider):
             finish_reason = str(choice0.get("finish_reason") or "stop")
 
             raw_tool_calls: list[Any] = []
+            # StepFun Plan: fallback to reasoning field when content is empty
+            if not content and msg0.get("reasoning"):
+                content = self._extract_text_content(msg0.get("reasoning"))
             reasoning_content = msg0.get("reasoning_content")
+            if not reasoning_content and msg0.get("reasoning"):
+                reasoning_content = self._extract_text_content(msg0.get("reasoning"))
             for ch in choices:
                 ch_map = self._maybe_mapping(ch) or {}
                 m = self._maybe_mapping(ch_map.get("message")) or {}
@@ -511,6 +720,8 @@ class OpenAICompatProvider(LLMProvider):
                     finish_reason = ch.finish_reason
             if not content and m.content:
                 content = m.content
+            if not content and getattr(m, "reasoning", None):
+                content = m.reasoning
 
         tool_calls = []
         for tc in raw_tool_calls:
@@ -527,12 +738,16 @@ class OpenAICompatProvider(LLMProvider):
                 function_provider_specific_fields=fn_prov,
             ))
 
+        reasoning_content = getattr(msg, "reasoning_content", None) or None
+        if not reasoning_content and getattr(msg, "reasoning", None):
+            reasoning_content = msg.reasoning
+
         return LLMResponse(
             content=content,
             tool_calls=tool_calls,
             finish_reason=finish_reason or "stop",
             usage=self._extract_usage(response),
-            reasoning_content=getattr(msg, "reasoning_content", None) or None,
+            reasoning_content=reasoning_content,
         )
 
     @classmethod
@@ -593,6 +808,8 @@ class OpenAICompatProvider(LLMProvider):
                 if text:
                     content_parts.append(text)
                 text = cls._extract_text_content(delta.get("reasoning_content"))
+                if not text:
+                    text = cls._extract_text_content(delta.get("reasoning"))
                 if text:
                     reasoning_parts.append(text)
                 for idx, tc in enumerate(delta.get("tool_calls") or []):
@@ -611,6 +828,8 @@ class OpenAICompatProvider(LLMProvider):
                 content_parts.append(delta.content)
             if delta:
                 reasoning = getattr(delta, "reasoning_content", None)
+                if not reasoning:
+                    reasoning = getattr(delta, "reasoning", None)
                 if reasoning:
                     reasoning_parts.append(reasoning)
             for tc in (delta.tool_calls or []) if delta else []:
@@ -683,7 +902,12 @@ class OpenAICompatProvider(LLMProvider):
         }
 
     @staticmethod
-    def _handle_error(e: Exception) -> LLMResponse:
+    def _handle_error(
+        e: Exception,
+        *,
+        spec: ProviderSpec | None = None,
+        api_base: str | None = None,
+    ) -> LLMResponse:
         body = (
             getattr(e, "doc", None)
             or getattr(e, "body", None)
@@ -691,6 +915,15 @@ class OpenAICompatProvider(LLMProvider):
         )
         body_text = body if isinstance(body, str) else str(body) if body is not None else ""
         msg = f"Error: {body_text.strip()[:500]}" if body_text.strip() else f"Error calling LLM: {e}"
+
+        text = f"{body_text} {e}".lower()
+        if spec and spec.is_local and ("502" in text or "connection" in text or "refused" in text):
+            msg += (
+                "\nHint: this is a local model endpoint. Check that the local server is reachable at "
+                f"{api_base or spec.default_api_base}, and if you are using a proxy/tunnel, make sure it "
+                "can reach your local Ollama/vLLM service instead of routing localhost through the remote host."
+            )
+
         response = getattr(e, "response", None)
         retry_after = LLMProvider._extract_retry_after_from_headers(getattr(response, "headers", None))
         if retry_after is None:
@@ -716,14 +949,28 @@ class OpenAICompatProvider(LLMProvider):
         reasoning_effort: str | None = None,
         tool_choice: str | dict[str, Any] | None = None,
     ) -> LLMResponse:
-        kwargs = self._build_kwargs(
-            messages, tools, model, max_tokens, temperature,
-            reasoning_effort, tool_choice,
-        )
         try:
+            if self._should_use_responses_api(model, reasoning_effort):
+                try:
+                    body = self._build_responses_body(
+                        messages, tools, model, max_tokens, temperature,
+                        reasoning_effort, tool_choice,
+                    )
+                    result = parse_response_output(await self._client.responses.create(**body))
+                    self._record_responses_success(model, reasoning_effort)
+                    return result
+                except Exception as responses_error:
+                    if not self._should_fallback_from_responses_error(responses_error):
+                        raise
+                    self._record_responses_failure(model, reasoning_effort)
+
+            kwargs = self._build_kwargs(
+                messages, tools, model, max_tokens, temperature,
+                reasoning_effort, tool_choice,
+            )
             return self._parse(await self._client.chat.completions.create(**kwargs))
         except Exception as e:
-            return self._handle_error(e)
+            return self._handle_error(e, spec=self._spec, api_base=self.api_base)
 
     async def chat_stream(
         self,
@@ -736,14 +983,51 @@ class OpenAICompatProvider(LLMProvider):
         tool_choice: str | dict[str, Any] | None = None,
         on_content_delta: Callable[[str], Awaitable[None]] | None = None,
     ) -> LLMResponse:
-        kwargs = self._build_kwargs(
-            messages, tools, model, max_tokens, temperature,
-            reasoning_effort, tool_choice,
-        )
-        kwargs["stream"] = True
-        kwargs["stream_options"] = {"include_usage": True}
         idle_timeout_s = int(os.environ.get("NANOBOT_STREAM_IDLE_TIMEOUT_S", "90"))
         try:
+            if self._should_use_responses_api(model, reasoning_effort):
+                try:
+                    body = self._build_responses_body(
+                        messages, tools, model, max_tokens, temperature,
+                        reasoning_effort, tool_choice,
+                    )
+                    body["stream"] = True
+                    stream = await self._client.responses.create(**body)
+
+                    async def _timed_stream():
+                        stream_iter = stream.__aiter__()
+                        while True:
+                            try:
+                                yield await asyncio.wait_for(
+                                    stream_iter.__anext__(),
+                                    timeout=idle_timeout_s,
+                                )
+                            except StopAsyncIteration:
+                                break
+
+                    content, tool_calls, finish_reason, usage, reasoning_content = await consume_sdk_stream(
+                        _timed_stream(),
+                        on_content_delta,
+                    )
+                    self._record_responses_success(model, reasoning_effort)
+                    return LLMResponse(
+                        content=content or None,
+                        tool_calls=tool_calls,
+                        finish_reason=finish_reason,
+                        usage=usage,
+                        reasoning_content=reasoning_content,
+                    )
+                except Exception as responses_error:
+                    if not self._should_fallback_from_responses_error(responses_error):
+                        raise
+                    self._record_responses_failure(model, reasoning_effort)
+
+            kwargs = self._build_kwargs(
+                messages, tools, model, max_tokens, temperature,
+                reasoning_effort, tool_choice,
+            )
+            kwargs["stream"] = True
+            kwargs["stream_options"] = {"include_usage": True}
             stream = await self._client.chat.completions.create(**kwargs)
             chunks: list[Any] = []
             stream_iter = stream.__aiter__()
@@ -771,7 +1055,7 @@ class OpenAICompatProvider(LLMProvider):
                 error_kind="timeout",
             )
         except Exception as e:
-            return self._handle_error(e)
+            return self._handle_error(e, spec=self._spec, api_base=self.api_base)
 
     def get_default_model(self) -> str:
         return self.default_model

nanobot/providers/registry.py

@@ -280,6 +280,15 @@ PROVIDERS: tuple[ProviderSpec, ...] = (
         backend="openai_compat",
         default_api_base="https://api.minimax.io/v1",
     ),
+    # MiniMax Anthropic-compatible endpoint: supports thinking mode
+    ProviderSpec(
+        name="minimax_anthropic",
+        keywords=("minimax_anthropic",),
+        env_key="MINIMAX_API_KEY",
+        display_name="MiniMax (Anthropic)",
+        backend="anthropic",
+        default_api_base="https://api.minimax.io/anthropic",
+    ),
     # Mistral AI: OpenAI-compatible API
     ProviderSpec(
         name="mistral",
@@ -328,6 +337,17 @@ PROVIDERS: tuple[ProviderSpec, ...] = (
         detect_by_base_keyword="11434",
         default_api_base="http://localhost:11434/v1",
     ),
+    # LM Studio (local, OpenAI-compatible)
+    ProviderSpec(
+        name="lm_studio",
+        keywords=("lm-studio", "lmstudio", "lm_studio"),
+        env_key="LM_STUDIO_API_KEY",
+        display_name="LM Studio",
+        backend="openai_compat",
+        is_local=True,
+        detect_by_base_keyword="1234",
+        default_api_base="http://localhost:1234/v1",
+    ),
     # === OpenVINO Model Server (direct, local, OpenAI-compatible at /v3) ===
     ProviderSpec(
         name="ovms",

nanobot/providers/transcription.py

@@ -10,9 +10,13 @@ from loguru import logger
 class OpenAITranscriptionProvider:
     """Voice transcription provider using OpenAI's Whisper API."""
 
-    def __init__(self, api_key: str | None = None):
+    def __init__(self, api_key: str | None = None, api_base: str | None = None):
         self.api_key = api_key or os.environ.get("OPENAI_API_KEY")
-        self.api_url = "https://api.openai.com/v1/audio/transcriptions"
+        self.api_url = (
+            api_base
+            or os.environ.get("OPENAI_TRANSCRIPTION_BASE_URL")
+            or "https://api.openai.com/v1/audio/transcriptions"
+        )
 
     async def transcribe(self, file_path: str | Path) -> str:
         if not self.api_key:
@@ -44,9 +48,9 @@ class GroqTranscriptionProvider:
     Groq offers extremely fast transcription with a generous free tier.
     """
 
-    def __init__(self, api_key: str | None = None):
+    def __init__(self, api_key: str | None = None, api_base: str | None = None):
         self.api_key = api_key or os.environ.get("GROQ_API_KEY")
-        self.api_url = "https://api.groq.com/openai/v1/audio/transcriptions"
+        self.api_url = api_base or os.environ.get("GROQ_BASE_URL") or "https://api.groq.com/openai/v1/audio/transcriptions"
 
     async def transcribe(self, file_path: str | Path) -> str:
         """

nanobot/skills/my/SKILL.md

@@ -0,0 +1,72 @@
+---
+name: my
+description: Check and set the agent's own runtime state (model, iterations, context window, token usage, web config). Use when diagnosing why something doesn't work ("why can't you search the web?", "why did you stop?"), checking resource limits before complex tasks, adapting configuration for long or simple tasks, or remembering user preferences across turns. Also use when the user asks what model you are running, how many tokens you've used, or what your settings are.
+always: true
+---
+
+# Self-Awareness
+
+## How to use
+
+1. **Identify the situation** from the categories below
+2. **Call the my tool** with the appropriate action
+3. **If set**, warn the user before changing impactful settings (model, iterations)
+4. **For detailed examples**, read [references/examples.md](references/examples.md)
+
+## When to check
+
+<rule>
+**Diagnose before explaining.** When something doesn't work, check your state first.
+</rule>
+
+<rule>
+**Check budget before complex tasks.** Know your limits before committing.
+</rule>
+
+<rule>
+**Recall across turns.** Store preferences in your scratchpad, read them back later.
+</rule>
+
+## When to set
+
+<rule>
+**Only set when benefit is clear and user is informed.** Warn before changing model.
+</rule>
+
+| Situation | Command |
+|-----------|---------|
+| Large codebase analysis | `my(action="set", key="context_window_tokens", value=131072)` |
+| Repetitive simple tasks | `my(action="set", key="model", value="<fast-model>")` |
+| Long multi-step task | `my(action="set", key="max_iterations", value=80)` |
+
+**Tradeoff:** Bias toward stability. Only set when defaults are genuinely insufficient.
+
+## Anti-patterns
+
+<rule>
+**Don't check every turn.** Costs a tool call. Use when you need information, not reflexively.
+</rule>
+
+<rule>
+**Don't store sensitive data.** No API keys, passwords, or tokens in scratchpad.
+</rule>
+
+<rule>
+**Don't set workspace.** Does not update file tool boundaries — won't work.
+</rule>
+
+## Constraints
+
+- All modifications in-memory only — restart resets everything
+- Protected params have type/range validation: `max_iterations` (1–100), `context_window_tokens` (4096–1M), `model` (non-empty str)
+- If `tools.my.allow_set` is false, check only
+
+## Related tools
+
+| Need | Use | Persists? |
+|------|-----|-----------|
+| Per-session temp state | `my(action="set", key="...", value=...)` | No |
+| Long-term facts | Memory skill (`MEMORY.md`, `USER.md`) | Yes |
+| Permanent config change | Edit config file | Yes |
+
+**Rule of thumb:** Tomorrow? Memory. This turn only? My.

nanobot/skills/my/references/examples.md

@@ -0,0 +1,75 @@
+# My Tool — Practical Examples
+
+Concrete scenarios showing when and how to use the my tool effectively.
+
+## Diagnosis
+
+### "Why can't you search the web?"
+```
+→ my(action="check", key="web_config.enable")
+  → False
+→ "Web search is disabled. Add web.enable: true to your config to enable it."
+```
+
+### "Why did you stop?"
+```
+→ my(action="check", key="max_iterations")
+  → 40
+→ my(action="check", key="_last_usage")
+  → {"prompt_tokens": 62000, "completion_tokens": 3000}
+→ "I hit the iteration limit (40). The task was complex. I can ask the user if they want to increase it."
+```
+
+### "What model are you running?"
+```
+→ my(action="check", key="model")
+  → 'anthropic/claude-sonnet-4-20250514'
+```
+
+## Adaptive Behavior
+
+### Large codebase analysis
+```
+→ my(action="check")
+  → context_window_tokens: 65536
+→ my(action="set", key="context_window_tokens", value=131072)
+  → "Set context_window_tokens = 131072 (was 65536)"
+→ "I've expanded my context window to handle this large codebase."
+```
+
+### Switching to a faster model for repetitive tasks
+```
+→ my(action="set", key="model", value="anthropic/claude-haiku-4-5-20251001")
+  → "Set model = 'anthropic/claude-haiku-4-5-20251001' (was 'anthropic/claude-sonnet-4-20250514')"
+→ "Switched to a faster model for these batch tasks."
+```
+
+## Cross-Turn Memory
+
+### Remembering user preferences
+```
+# Turn 1: user says "keep it brief"
+→ my(action="set", key="user_style", value="concise")
+  → "Set scratchpad.user_style = 'concise'"
+
+# Turn 3: new topic
+→ my(action="check", key="user_style")
+  → 'concise'
+  (adjusts response style accordingly)
+```
+
+### Tracking project context
+```
+→ my(action="set", key="active_branch", value="feat/auth")
+→ my(action="set", key="test_framework", value="pytest")
+→ my(action="set", key="has_docker", value=true)
+```
+
+## Budget Awareness
+
+### Token-conscious behavior
+```
+→ my(action="check", key="_last_usage")
+  → {"prompt_tokens": 58000, "completion_tokens": 12000}
+→ "I've consumed ~70k tokens. I'll keep my remaining responses focused."
+```

nanobot/templates/AGENTS.md

@@ -1,7 +1,5 @@
 # Agent Instructions
 
-You are a helpful AI assistant. Be concise, accurate, and friendly.
-
 ## Scheduled Reminders
 
 Before scheduling reminders, check available skills and follow skill guidance first.

nanobot/templates/SOUL.md

@@ -2,20 +2,8 @@
 
 I am nanobot 🐈, a personal AI assistant.
 
-## Personality
-
-- Helpful and friendly
-- Concise and to the point
-- Curious and eager to learn
-
-## Values
-
-- Accuracy over speed
-- User privacy and safety
-- Transparency in actions
-
-## Communication Style
-
-- Be clear and direct
-- Explain reasoning when helpful
-- Ask clarifying questions when needed
+I solve problems by doing, not by describing what I would do.
+I keep responses short unless depth is asked for.
+I say what I know, flag what I don't, and never fake confidence.
+I stay friendly and curious — I'd rather ask a good question than guess wrong.
+I treat the user's time as the scarcest resource, and their trust as the most valuable.

nanobot/templates/agent/dream_phase1.md

@@ -1,8 +1,11 @@
-Compare conversation history against current memory files. Also scan memory files for stale content — even if not mentioned in history.
+You have TWO equally important tasks:
+1. Extract new facts from conversation history
+2. Deduplicate existing memory files — find and flag redundant, overlapping, or stale content even if NOT mentioned in history
 
 Output one line per finding:
 [FILE] atomic fact (not already in memory)
 [FILE-REMOVE] reason for removal
+[SKILL] kebab-case-name: one-line description of the reusable pattern
 
 Files: USER (identity, preferences), SOUL (bot behavior, tone), MEMORY (knowledge, project context)
 
@@ -11,12 +14,26 @@ Rules:
 - Corrections: [USER] location is Tokyo, not Osaka
 - Capture confirmed approaches the user validated
 
-Staleness — flag for [FILE-REMOVE]:
-- Time-sensitive data older than 14 days: weather, daily status, one-time meetings, passed events
-- Completed one-time tasks: triage, one-time reviews, finished research, resolved incidents
-- Resolved tracking: merged/closed PRs, fixed issues, completed migrations
-- Detailed incident info after 14 days — reduce to one-line summary
-- Superseded: approaches replaced by newer solutions, deprecated dependencies
+Deduplication — scan ALL memory files for these redundancy patterns:
+- Same fact stated in multiple places (e.g., "communicates in Chinese" in both USER.md and multiple MEMORY.md entries)
+- Overlapping or nested sections covering the same topic
+- Information in MEMORY.md that is already captured in USER.md or SOUL.md (MEMORY.md should not duplicate permanent-file content)
+- Verbose entries that can be condensed without losing information
+For each duplicate found, output [FILE-REMOVE] for the less authoritative copy (prefer keeping facts in their canonical location)
+
+Staleness — MEMORY.md lines may have a ``← Nd`` suffix showing days since last modification:
+- SOUL.md and USER.md have no age annotations — they are permanent, only update with corrections
+- Age only indicates when content was last touched, not whether it should be removed
+- Use content judgment: user habits/preferences/personality traits are permanent regardless of age
+- Only prune content that is objectively outdated: passed events, resolved tracking, superseded approaches
+- Lines with ``← Nd`` (N>{{ stale_threshold_days }}) deserve closer review but are NOT automatically removable
+- When removing: prefer deleting individual items over entire sections
+
+Skill discovery — flag [SKILL] when ALL of these are true:
+- A specific, repeatable workflow appeared 2+ times in the conversation history
+- It involves clear steps (not vague preferences like "likes concise answers")
+- It is substantial enough to warrant its own instruction set (not trivial like "read a file")
+- Do not worry about duplicates — the next phase will check against existing skills
 
 Do not add: current weather, transient status, temporary errors, conversational filler.
 

nanobot/templates/agent/dream_phase2.md

@@ -1,11 +1,13 @@
 Update memory files based on the analysis below.
 - [FILE] entries: add the described content to the appropriate file
 - [FILE-REMOVE] entries: delete the corresponding content from memory files
+- [SKILL] entries: create a new skill under skills/<name>/SKILL.md using write_file
 
 ## File paths (relative to workspace root)
 - SOUL.md
 - USER.md
 - memory/MEMORY.md
+- skills/<name>/SKILL.md (for [SKILL] entries only)
 
 Do NOT guess paths.
 
@@ -17,6 +19,17 @@ Do NOT guess paths.
 - Surgical edits only — never rewrite entire files
 - If nothing to update, stop without calling tools
 
+## Skill creation rules (for [SKILL] entries)
+- Use write_file to create skills/<name>/SKILL.md
+- Before writing, read_file `{{ skill_creator_path }}` for format reference (frontmatter structure, naming conventions, quality standards)
+- **Dedup check**: read existing skills listed below to verify the new skill is not functionally redundant. Skip creation if an existing skill already covers the same workflow.
+- Include YAML frontmatter with name and description fields
+- Keep SKILL.md under 2000 words — concise and actionable
+- Include: when to use, steps, output format, at least one example
+- Do NOT overwrite existing skills — skip if the skill directory already exists
+- Reference specific tools the agent has access to (read_file, write_file, exec, web_search, etc.)
+- Skills are instruction sets, not code — do not include implementation code
+
 ## Quality
 - Every line must carry standalone value
 - Concise bullets under clear headers

nanobot/templates/agent/identity.md

@@ -12,15 +12,32 @@ Your workspace is at: {{ workspace_path }}
 - Custom skills: {{ workspace_path }}/skills/{% raw %}{skill-name}{% endraw %}/SKILL.md
 
 {{ platform_policy }}
+{% if channel == 'telegram' or channel == 'qq' or channel == 'discord' %}
+## Format Hint
+This conversation is on a messaging app. Use short paragraphs. Avoid large headings (#, ##). Use **bold** sparingly. No tables — use plain lists.
+{% elif channel == 'whatsapp' or channel == 'sms' %}
+## Format Hint
+This conversation is on a text messaging platform that does not render markdown. Use plain text only.
+{% elif channel == 'email' %}
+## Format Hint
+This conversation is via email. Structure with clear sections. Markdown may not render — keep formatting simple.
+{% elif channel == 'cli' or channel == 'mochat' %}
+## Format Hint
+Output is rendered in a terminal. Avoid markdown headings and tables. Use plain text with minimal formatting.
+{% endif %}
 
-## nanobot Guidelines
-- State intent before tool calls, but NEVER predict or claim results before receiving them.
-- Before modifying a file, read it first. Do not assume files or directories exist.
-- After writing or editing a file, re-read it if accuracy matters.
-- If a tool call fails, analyze the error before retrying with a different approach.
-- Ask for clarification when the request is ambiguous.
-- Prefer built-in `grep` / `glob` tools for workspace search before falling back to `exec`.
-- On broad searches, use `grep(output_mode="count")` or `grep(output_mode="files_with_matches")` to scope the result set before requesting full content.
+## Execution Rules
+
+- Act, don't narrate. If you can do it with a tool, do it now — never end a turn with just a plan or promise.
+- Read before you write. Do not assume a file exists or contains what you expect.
+- If a tool call fails, diagnose the error and retry with a different approach before reporting failure.
+- When information is missing, look it up with tools first. Only ask the user when tools cannot answer.
+- After multi-step changes, verify the result (re-read the file, run the test, check the output).
+
+## Search & Discovery
+
+- Prefer built-in `grep` / `glob` over `exec` for workspace search.
+- On broad searches, use `grep(output_mode="count")` to scope before requesting full content.
 {% include 'agent/_snippets/untrusted_content.md' %}
 
 Reply directly with text for conversations. Only use the 'message' tool to send to a specific chat channel.

nanobot/templates/agent/skills_section.md

@@ -1,6 +1,6 @@
 # Skills
 
 The following skills extend your capabilities. To use a skill, read its SKILL.md file using the read_file tool.
-Skills with available="false" need dependencies installed first - you can try installing them with apt/brew.
+Unavailable skills need dependencies installed first — you can try installing them with apt/brew.
 
 {{ skills_summary }}

nanobot/utils/document.py

@@ -0,0 +1,267 @@
+"""Document text extraction utilities for nanobot."""
+
+import mimetypes
+from pathlib import Path
+
+from loguru import logger
+
+from nanobot.utils.helpers import detect_image_mime
+
+try:
+    from pypdf import PdfReader
+except ImportError:
+    PdfReader = None  # type: ignore
+
+try:
+    from docx import Document as DocxDocument
+except ImportError:
+    DocxDocument = None  # type: ignore
+
+try:
+    from openpyxl import load_workbook
+except ImportError:
+    load_workbook = None  # type: ignore
+
+try:
+    from pptx import Presentation as PptxPresentation
+except ImportError:
+    PptxPresentation = None  # type: ignore
+
+
+# Supported file extensions for text extraction
+SUPPORTED_EXTENSIONS: set[str] = {
+    # Document formats
+    ".pdf",
+    ".docx",
+    ".xlsx",
+    ".pptx",
+    # Text formats
+    ".txt",
+    ".md",
+    ".csv",
+    ".json",
+    ".xml",
+    ".html",
+    ".htm",
+    ".log",
+    ".yaml",
+    ".yml",
+    ".toml",
+    ".ini",
+    ".cfg",
+    # Image formats (for future OCR support)
+    ".png",
+    ".jpg",
+    ".jpeg",
+    ".gif",
+    ".webp",
+}
+
+_MAX_TEXT_LENGTH = 200_000
+
+
+def extract_text(path: Path) -> str | None:
+    """Extract text from a file.
+
+    Args:
+        path: Path to the file.
+
+    Returns:
+        Extracted text as string, None for unsupported types,
+        or error string for failures.
+    """
+    if not isinstance(path, Path):
+        path = Path(path)
+
+    if not path.exists():
+        return f"[error: file not found: {path}]"
+
+    ext = path.suffix.lower()
+
+    # Document formats
+    if ext == ".pdf":
+        if PdfReader is None:
+            return "[error: pypdf not installed]"
+        return _extract_pdf(path)
+    elif ext == ".docx":
+        if DocxDocument is None:
+            return "[error: python-docx not installed]"
+        return _extract_docx(path)
+    elif ext == ".xlsx":
+        if load_workbook is None:
+            return "[error: openpyxl not installed]"
+        return _extract_xlsx(path)
+    elif ext == ".pptx":
+        if PptxPresentation is None:
+            return "[error: python-pptx not installed]"
+        return _extract_pptx(path)
+    elif _is_text_extension(ext):
+        return _extract_text_file(path)
+    elif ext in {".png", ".jpg", ".jpeg", ".gif", ".webp"}:
+        # Image files - for future OCR support
+        return f"[image: {path.name}]"
+    else:
+        # Unsupported extension
+        return None
+
+
+def _extract_pdf(path: Path) -> str:
+    """Extract text from PDF using pypdf."""
+    try:
+        reader = PdfReader(path)
+        pages: list[str] = []
+        for i, page in enumerate(reader.pages, 1):
+            text = page.extract_text() or ""
+            pages.append(f"--- Page {i} ---\n{text}")
+        return _truncate("\n\n".join(pages), _MAX_TEXT_LENGTH)
+    except Exception as e:
+        logger.error("Failed to extract PDF {}: {}", path, e)
+        return f"[error: failed to extract PDF: {e!s}]"
+
+
+def _extract_docx(path: Path) -> str:
+    """Extract text from DOCX using python-docx."""
+    try:
+        doc = DocxDocument(path)
+        paragraphs: list[str] = [p.text for p in doc.paragraphs if p.text.strip()]
+        return _truncate("\n\n".join(paragraphs), _MAX_TEXT_LENGTH)
+    except Exception as e:
+        logger.error("Failed to extract DOCX {}: {}", path, e)
+        return f"[error: failed to extract DOCX: {e!s}]"
+
+
+def _extract_xlsx(path: Path) -> str:
+    """Extract text from XLSX using openpyxl."""
+    try:
+        wb = load_workbook(path, read_only=True, data_only=True)
+        sheets: list[str] = []
+        for sheet_name in wb.sheetnames:
+            ws = wb[sheet_name]
+            rows: list[str] = []
+            for row in ws.iter_rows(values_only=True):
+                row_text = "\t".join(str(cell) if cell is not None else "" for cell in row)
+                if row_text.strip():
+                    rows.append(row_text)
+            if rows:
+                sheets.append(f"--- Sheet: {sheet_name} ---\n" + "\n".join(rows))
+        wb.close()
+        return _truncate("\n\n".join(sheets), _MAX_TEXT_LENGTH)
+    except Exception as e:
+        logger.error("Failed to extract XLSX {}: {}", path, e)
+        return f"[error: failed to extract XLSX: {e!s}]"
+
+
+def _extract_pptx(path: Path) -> str:
+    """Extract text from PPTX using python-pptx."""
+    try:
+        prs = PptxPresentation(path)
+        slides: list[str] = []
+        for i, slide in enumerate(prs.slides, 1):
+            slide_text: list[str] = []
+            for shape in slide.shapes:
+                if hasattr(shape, "text") and shape.text:
+                    slide_text.append(shape.text)
+            if slide_text:
+                slides.append(f"--- Slide {i} ---\n" + "\n".join(slide_text))
+        return _truncate("\n\n".join(slides), _MAX_TEXT_LENGTH)
+    except Exception as e:
+        logger.error("Failed to extract PPTX {}: {}", path, e)
+        return f"[error: failed to extract PPTX: {e!s}]"
+
+
+def _extract_text_file(path: Path) -> str:
+    """Extract text from a plain text file."""
+    try:
+        # Try UTF-8 first, then latin-1 fallback
+        try:
+            content = path.read_text(encoding="utf-8")
+        except UnicodeDecodeError:
+            content = path.read_text(encoding="latin-1")
+        return _truncate(content, _MAX_TEXT_LENGTH)
+    except Exception as e:
+        logger.error("Failed to read text file {}: {}", path, e)
+        return f"[error: failed to read file: {e!s}]"
+
+
+def _truncate(text: str, max_length: int) -> str:
+    """Truncate text with a suffix indicating truncation."""
+    if len(text) <= max_length:
+        return text
+    return text[:max_length] + f"... (truncated, {len(text)} chars total)"
+
+
+def _is_text_extension(ext: str) -> bool:
+    """Check if extension is a text format."""
+    return ext in {
+        ".txt",
+        ".md",
+        ".csv",
+        ".json",
+        ".xml",
+        ".html",
+        ".htm",
+        ".log",
+        ".yaml",
+        ".yml",
+        ".toml",
+        ".ini",
+        ".cfg",
+    }
+
+
+# ---------------------------------------------------------------------------
+# High-level helper: split media into images + extracted document text
+# ---------------------------------------------------------------------------
+
+_MAX_EXTRACT_FILE_SIZE = 50 * 1024 * 1024  # 50 MB
+
+
+def extract_documents(
+    text: str,
+    media_paths: list[str],
+    *,
+    max_file_size: int = _MAX_EXTRACT_FILE_SIZE,
+) -> tuple[str, list[str]]:
+    """Separate images from documents in *media_paths*.
+
+    Documents (PDF, DOCX, XLSX, PPTX, plain-text, …) have their text
+    extracted and appended to *text*.  Only image paths are kept in the
+    returned list so that downstream layers only need to handle vision
+    blocks.
+
+    Files larger than *max_file_size* bytes are skipped with a warning
+    to avoid unbounded memory / CPU usage.
+    """
+    image_paths: list[str] = []
+    doc_texts: list[str] = []
+
+    for path_str in media_paths:
+        p = Path(path_str)
+        if not p.is_file():
+            continue
+
+        try:
+            size = p.stat().st_size
+        except OSError:
+            continue
+        if size > max_file_size:
+            logger.warning(
+                "Skipping oversized file for extraction: {} ({:.1f} MB > {} MB limit)",
+                p.name, size / (1024 * 1024), max_file_size // (1024 * 1024),
+            )
+            continue
+
+        with open(p, "rb") as f:
+            header = f.read(16)
+        mime = detect_image_mime(header) or mimetypes.guess_type(path_str)[0]
+        if mime and mime.startswith("image/"):
+            image_paths.append(path_str)
+        else:
+            extracted = extract_text(p)
+            if extracted and not extracted.startswith("[error:"):
+                doc_texts.append(f"[File: {p.name}]\n{extracted}")
+
+    if doc_texts:
+        text = text + "\n\n" + "\n\n".join(doc_texts)
+
+    return text, image_paths

nanobot/utils/gitstore.py

@@ -5,6 +5,7 @@ from __future__ import annotations
 import io
 import time
 from dataclasses import dataclass
+from datetime import datetime, timezone
 from pathlib import Path
 
 from loguru import logger
@@ -24,6 +25,23 @@ class CommitInfo:
         return f"{header}\n(no file changes)"
 
 
+@dataclass
+class LineAge:
+    """Age of a single line based on git blame."""
+
+    age_days: int  # days since last modification
+
+
+def _compute_line_ages(annotated) -> list[LineAge]:
+    """Convert annotate results to per-line ages."""
+    now = datetime.now(tz=timezone.utc).date()
+    ages: list[LineAge] = []
+    for (commit, _tree_entry), _line_bytes in annotated:
+        dt = datetime.fromtimestamp(commit.commit_time, tz=timezone.utc).date()
+        ages.append(LineAge(age_days=(now - dt).days))
+    return ages
+
+
 class GitStore:
     """Git-backed version control for memory files."""
 
@@ -191,6 +209,34 @@ class GitStore:
             logger.warning("Git log failed")
             return []
 
+    def line_ages(self, file_path: str) -> list[LineAge]:
+        """Compute the age of each line in a tracked file via git blame.
+
+        Returns one LineAge per line, in order.
+        Returns an empty list if the repo is not initialized, the file is
+        empty, or annotation fails.
+        """
+
+        if not self.is_initialized():
+            return []
+
+        target = self._workspace / file_path
+        if not target.exists() or target.stat().st_size == 0:
+            return []
+
+        try:
+            from dulwich import porcelain
+
+            annotated = porcelain.annotate(str(self._workspace), file_path)
+        except Exception:
+            logger.warning("Git line_ages annotate failed for {}", file_path)
+            return []
+
+        if not annotated:
+            return []
+
+        return _compute_line_ages(annotated)
+
     def diff_commits(self, sha1: str, sha2: str) -> str:
         """Show diff between two commits."""
         if not self.is_initialized():

nanobot/utils/helpers.py

@@ -17,10 +17,10 @@ from loguru import logger
 def strip_think(text: str) -> str:
     """Remove thinking blocks and any unclosed trailing tag."""
     text = re.sub(r"<think>[\s\S]*?</think>", "", text)
-    text = re.sub(r"<think>[\s\S]*$", "", text)
+    text = re.sub(r"^\s*<think>[\s\S]*$", "", text)
     # Gemma 4 and similar models use <thought>...</thought> blocks
     text = re.sub(r"<thought>[\s\S]*?</thought>", "", text)
-    text = re.sub(r"<thought>[\s\S]*$", "", text)
+    text = re.sub(r"^\s*<thought>[\s\S]*$", "", text)
     return text.strip()
 
 
@@ -275,7 +275,7 @@ def build_assistant_message(
     thinking_blocks: list[dict] | None = None,
 ) -> dict[str, Any]:
     """Build a provider-safe assistant message with optional reasoning fields."""
-    msg: dict[str, Any] = {"role": "assistant", "content": content}
+    msg: dict[str, Any] = {"role": "assistant", "content": content or ""}
     if tool_calls:
         msg["tool_calls"] = tool_calls
     if reasoning_content is not None or thinking_blocks:
@@ -400,6 +400,8 @@ def build_status_content(
     session_msg_count: int,
     context_tokens_estimate: int,
     search_usage_text: str | None = None,
+    active_task_count: int = 0,
+    max_completion_tokens: int = 8192,
 ) -> str:
     """Build a human-readable runtime status snapshot.
     
@@ -418,9 +420,11 @@ def build_status_content(
     last_out = last_usage.get("completion_tokens", 0)
     cached = last_usage.get("cached_tokens", 0)
     ctx_total = max(context_window_tokens, 0)
-    ctx_pct = int((context_tokens_estimate / ctx_total) * 100) if ctx_total > 0 else 0
+    # Budget mirrors Consolidator formula: ctx_window - max_completion - _SAFETY_BUFFER
+    ctx_budget = max(ctx_total - int(max_completion_tokens) - 1024, 1)
+    ctx_pct = min(int((context_tokens_estimate / ctx_budget) * 100), 999) if ctx_budget > 0 else 0
     ctx_used_str = f"{context_tokens_estimate // 1000}k" if context_tokens_estimate >= 1000 else str(context_tokens_estimate)
-    ctx_total_str = f"{ctx_total // 1024}k" if ctx_total > 0 else "n/a"
+    ctx_total_str = f"{ctx_total // 1000}k" if ctx_total > 0 else "n/a"
     token_line = f"\U0001f4ca Tokens: {last_in} in / {last_out} out"
     if cached and last_in:
         token_line += f" ({cached * 100 // last_in}% cached)"
@@ -428,9 +432,10 @@ def build_status_content(
         f"\U0001f408 nanobot v{version}",
         f"\U0001f9e0 Model: {model}",
         token_line,
-        f"\U0001f4da Context: {ctx_used_str}/{ctx_total_str} ({ctx_pct}%)",
+        f"\U0001f4da Context: {ctx_used_str}/{ctx_total_str} ({ctx_pct}% of input budget)",
         f"\U0001f4ac Session: {session_msg_count} messages",
         f"\u23f1 Uptime: {uptime}",
+        f"\u26a1 Tasks: {active_task_count} active",
     ]
     if search_usage_text:
         lines.append(search_usage_text)

nanobot/utils/runtime.py

@@ -19,6 +19,11 @@ FINALIZATION_RETRY_PROMPT = (
     "Please provide your response to the user based on the conversation above."
 )
 
+LENGTH_RECOVERY_PROMPT = (
+    "Output limit reached. Continue exactly where you left off "
+    "— no recap, no apology. Break remaining work into smaller steps if needed."
+)
+
 
 def empty_tool_result_message(tool_name: str) -> str:
     """Short prompt-safe marker for tools that completed without visible output."""
@@ -50,6 +55,11 @@ def build_finalization_retry_message() -> dict[str, str]:
     return {"role": "user", "content": FINALIZATION_RETRY_PROMPT}
 
 
+def build_length_recovery_message() -> dict[str, str]:
+    """Prompt the model to continue after hitting output token limit."""
+    return {"role": "user", "content": LENGTH_RECOVERY_PROMPT}
+
+
 def external_lookup_signature(tool_name: str, arguments: dict[str, Any]) -> str | None:
     """Stable signature for repeated external lookups we want to throttle."""
     if tool_name == "web_fetch":

nanobot/utils/tool_hints.py

@@ -19,9 +19,11 @@ _TOOL_FORMATS: dict[str, tuple[list[str], str, bool, bool]] = {
     "list_dir":   (["path"],                           "ls {}",       True,  False),
 }
 
-# Matches file paths embedded in shell commands (Windows drive, ~/, or absolute after space)
+# Matches file paths embedded in shell commands, including quoted paths with spaces.
 _PATH_IN_CMD_RE = re.compile(
-    r"(?:[A-Za-z]:[/\\]|~/|(?<=\s)/)[^\s;&|<>\"']+"
+    r'"(?P<double>(?:[A-Za-z]:[/\\]|~/|/)[^"]+)"'
+    r"|'(?P<single>(?:[A-Za-z]:[/\\]|~/|/)[^']+)'"
+    r"|(?P<bare>(?:[A-Za-z]:[/\\]|~/|(?<=\s)/)[^\s;&|<>\"']+)"
 )
 
 
@@ -92,9 +94,14 @@ def _fmt_known(tc, fmt: tuple) -> str:
 
 def _abbreviate_command(cmd: str, max_len: int = 40) -> str:
     """Abbreviate paths in a command string, then truncate."""
-    abbreviated = _PATH_IN_CMD_RE.sub(
-        lambda m: abbreviate_path(m.group(), max_len=25), cmd
-    )
+    def _replace_path(match: re.Match[str]) -> str:
+        if match.group("double") is not None:
+            return f'"{abbreviate_path(match.group("double"), max_len=25)}"'
+        if match.group("single") is not None:
+            return f"'{abbreviate_path(match.group('single'), max_len=25)}'"
+        return abbreviate_path(match.group("bare"), max_len=25)
+
+    abbreviated = _PATH_IN_CMD_RE.sub(_replace_path, cmd)
     if len(abbreviated) <= max_len:
         return abbreviated
     return abbreviated[:max_len - 1] + "\u2026"

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "nanobot-ai"
-version = "0.1.5"
+version = "0.1.5.post1"
 description = "A lightweight personal AI assistant framework"
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.11"
@@ -50,6 +50,12 @@ dependencies = [
     "tiktoken>=0.12.0,<1.0.0",
     "jinja2>=3.1.0,<4.0.0",
     "dulwich>=0.22.0,<1.0.0",
+    "pyyaml>=6.0,<7.0.0",
+    "pypdf>=5.0.0,<6.0.0",
+    "python-docx>=1.1.0,<2.0.0",
+    "openpyxl>=3.1.0,<4.0.0",
+    "python-pptx>=1.0.0,<2.0.0",
+    "filelock>=3.25.2",
 ]
 
 [project.optional-dependencies]
@@ -79,12 +85,16 @@ discord = [
 langsmith = [
     "langsmith>=0.1.0",
 ]
+pdf = [
+    "pymupdf>=1.25.0",
+]
 dev = [
     "pytest>=9.0.0,<10.0.0",
     "pytest-asyncio>=1.3.0,<2.0.0",
     "aiohttp>=3.9.0,<4.0.0",
     "pytest-cov>=6.0.0,<7.0.0",
     "ruff>=0.1.0",
+    "pymupdf>=1.25.0",
 ]
 
 [project.scripts]

tests/agent/test_auto_compact.py

@@ -35,6 +35,13 @@ def _make_loop(tmp_path: Path, session_ttl_minutes: int = 15) -> AgentLoop:
     return loop
 
 
+def _add_turns(session, turns: int, *, prefix: str = "msg") -> None:
+    """Append simple user/assistant turns to a session."""
+    for i in range(turns):
+        session.add_message("user", f"{prefix} user {i}")
+        session.add_message("assistant", f"{prefix} assistant {i}")
+
+
 class TestSessionTTLConfig:
     """Test session TTL configuration."""
 
@@ -48,6 +55,23 @@ class TestSessionTTLConfig:
         defaults = AgentDefaults(session_ttl_minutes=30)
         assert defaults.session_ttl_minutes == 30
 
+    def test_user_friendly_alias_is_supported(self):
+        """Config should accept idleCompactAfterMinutes as the preferred JSON key."""
+        defaults = AgentDefaults.model_validate({"idleCompactAfterMinutes": 30})
+        assert defaults.session_ttl_minutes == 30
+
+    def test_legacy_alias_is_still_supported(self):
+        """Config should still accept the old sessionTtlMinutes key for compatibility."""
+        defaults = AgentDefaults.model_validate({"sessionTtlMinutes": 30})
+        assert defaults.session_ttl_minutes == 30
+
+    def test_serializes_with_user_friendly_alias(self):
+        """Config dumps should use idleCompactAfterMinutes for JSON output."""
+        defaults = AgentDefaults(session_ttl_minutes=30)
+        data = defaults.model_dump(mode="json", by_alias=True)
+        assert data["idleCompactAfterMinutes"] == 30
+        assert "sessionTtlMinutes" not in data
+
 
 class TestAgentLoopTTLParam:
     """Test that AutoCompact receives and stores session_ttl_minutes."""
@@ -101,7 +125,7 @@ class TestAutoCompact:
         loop.sessions.save(s2)
 
         async def _fake_archive(messages):
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
         loop.auto_compact.check_expired(loop._schedule_background)
@@ -113,28 +137,28 @@ class TestAutoCompact:
         await loop.close_mcp()
 
     @pytest.mark.asyncio
-    async def test_auto_compact_archives_and_clears(self, tmp_path):
-        """_archive should archive un-consolidated messages and clear session."""
+    async def test_auto_compact_archives_prefix_and_keeps_recent_suffix(self, tmp_path):
+        """_archive should summarize the old prefix and keep a recent legal suffix."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        for i in range(4):
-            session.add_message("user", f"msg{i}")
-            session.add_message("assistant", f"resp{i}")
+        _add_turns(session, 6)
         loop.sessions.save(session)
 
         archived_messages = []
 
         async def _fake_archive(messages):
             archived_messages.extend(messages)
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
 
         await loop.auto_compact._archive("cli:test")
 
-        assert len(archived_messages) == 8
+        assert len(archived_messages) == 4
         session_after = loop.sessions.get_or_create("cli:test")
-        assert len(session_after.messages) == 0
+        assert len(session_after.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
+        assert session_after.messages[0]["content"] == "msg user 2"
+        assert session_after.messages[-1]["content"] == "msg assistant 5"
         await loop.close_mcp()
 
     @pytest.mark.asyncio
@@ -142,17 +166,13 @@ class TestAutoCompact:
         """_archive should store the summary in _summaries."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "hello")
-        session.add_message("assistant", "hi there")
+        _add_turns(session, 6, prefix="hello")
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "User said hello."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "User said hello.",
-        }
 
         await loop.auto_compact._archive("cli:test")
 
@@ -160,7 +180,7 @@ class TestAutoCompact:
         assert entry is not None
         assert entry[0] == "User said hello."
         session_after = loop.sessions.get_or_create("cli:test")
-        assert len(session_after.messages) == 0
+        assert len(session_after.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
         await loop.close_mcp()
 
     @pytest.mark.asyncio
@@ -174,7 +194,7 @@ class TestAutoCompact:
         async def _fake_archive(messages):
             nonlocal archive_called
             archive_called = True
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
 
@@ -190,9 +210,7 @@ class TestAutoCompact:
         """_archive should only archive un-consolidated messages."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        for i in range(10):
-            session.add_message("user", f"msg{i}")
-            session.add_message("assistant", f"resp{i}")
+        _add_turns(session, 14)
         session.last_consolidated = 18
         loop.sessions.save(session)
 
@@ -201,7 +219,7 @@ class TestAutoCompact:
         async def _fake_archive(messages):
             nonlocal archived_count
             archived_count = len(messages)
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
 
@@ -235,7 +253,7 @@ class TestAutoCompactIdleDetection:
         """Proactive auto-new archives expired session; _process_message reloads it."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "old message")
+        _add_turns(session, 6, prefix="old")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -243,12 +261,9 @@ class TestAutoCompactIdleDetection:
 
         async def _fake_archive(messages):
             archived_messages.extend(messages)
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         # Simulate proactive archive completing before message arrives
         await loop.auto_compact._archive("cli:test")
@@ -257,7 +272,8 @@ class TestAutoCompactIdleDetection:
         await loop._process_message(msg)
 
         session_after = loop.sessions.get_or_create("cli:test")
-        assert not any(m["content"] == "old message" for m in session_after.messages)
+        assert len(archived_messages) == 4
+        assert not any(m["content"] == "old user 0" for m in session_after.messages)
         assert any(m["content"] == "new msg" for m in session_after.messages)
         await loop.close_mcp()
 
@@ -311,7 +327,7 @@ class TestAutoCompactIdleDetection:
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
 
@@ -335,17 +351,14 @@ class TestAutoCompactSystemMessages:
         """Proactive auto-new archives expired session; system messages reload it."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "old message from subagent context")
+        _add_turns(session, 6, prefix="old")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         # Simulate proactive archive completing before system message arrives
         await loop.auto_compact._archive("cli:test")
@@ -358,7 +371,7 @@ class TestAutoCompactSystemMessages:
 
         session_after = loop.sessions.get_or_create("cli:test")
         assert not any(
-            m["content"] == "old message from subagent context"
+            m["content"] == "old user 0"
             for m in session_after.messages
         )
         await loop.close_mcp()
@@ -372,8 +385,7 @@ class TestAutoCompactEdgeCases:
         """Auto-new should not inject when archive produces '(nothing)'."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "thanks")
-        session.add_message("assistant", "you're welcome")
+        _add_turns(session, 6, prefix="thanks")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -384,18 +396,18 @@ class TestAutoCompactEdgeCases:
         await loop.auto_compact._archive("cli:test")
 
         session_after = loop.sessions.get_or_create("cli:test")
-        assert len(session_after.messages) == 0
+        assert len(session_after.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
         # "(nothing)" summary should not be stored
         assert "cli:test" not in loop.auto_compact._summaries
 
         await loop.close_mcp()
 
     @pytest.mark.asyncio
-    async def test_auto_compact_archive_failure_still_clears(self, tmp_path):
-        """Auto-new should clear session even if LLM archive fails (raw_archive fallback)."""
+    async def test_auto_compact_archive_failure_still_keeps_recent_suffix(self, tmp_path):
+        """Auto-new should keep the recent suffix even if LLM archive falls back to raw dump."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "important data")
+        _add_turns(session, 6, prefix="important")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -405,14 +417,13 @@ class TestAutoCompactEdgeCases:
         await loop.auto_compact._archive("cli:test")
 
         session_after = loop.sessions.get_or_create("cli:test")
-        # Session should be cleared (archive falls back to raw dump)
-        assert len(session_after.messages) == 0
+        assert len(session_after.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
 
         await loop.close_mcp()
 
     @pytest.mark.asyncio
     async def test_auto_compact_preserves_runtime_checkpoint_before_check(self, tmp_path):
-        """Runtime checkpoint is restored; proactive archive handles the expired session."""
+        """Short expired sessions keep recent messages; checkpoint restore still works on resume."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
         session.metadata[AgentLoop._RUNTIME_CHECKPOINT_KEY] = {
@@ -428,12 +439,9 @@ class TestAutoCompactEdgeCases:
 
         async def _fake_archive(messages):
             archived_messages.extend(messages)
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         # Simulate proactive archive completing before message arrives
         await loop.auto_compact._archive("cli:test")
@@ -441,8 +449,10 @@ class TestAutoCompactEdgeCases:
         msg = InboundMessage(channel="cli", sender_id="user", chat_id="test", content="continue")
         await loop._process_message(msg)
 
-        # The checkpoint-restored message should have been archived by proactive path
-        assert len(archived_messages) >= 1
+        session_after = loop.sessions.get_or_create("cli:test")
+        assert archived_messages == []
+        assert any(m["content"] == "previous message" for m in session_after.messages)
+        assert any(m["content"] == "interrupted response" for m in session_after.messages)
 
         await loop.close_mcp()
 
@@ -458,11 +468,17 @@ class TestAutoCompactIntegration:
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
 
-        # Phase 1: User has a conversation
+        # Phase 1: User has a conversation longer than the retained recent suffix
         session.add_message("user", "I'm learning English, teach me past tense")
         session.add_message("assistant", "Past tense is used for actions completed in the past...")
         session.add_message("user", "Give me an example")
         session.add_message("assistant", '"I walked to the store yesterday."')
+        session.add_message("user", "Give me another example")
+        session.add_message("assistant", '"She visited Paris last year."')
+        session.add_message("user", "Quiz me")
+        session.add_message("assistant", "What is the past tense of go?")
+        session.add_message("user", "I think it is went")
+        session.add_message("assistant", "Correct.")
         loop.sessions.save(session)
 
         # Phase 2: Time passes (simulate idle)
@@ -486,7 +502,7 @@ class TestAutoCompactIntegration:
         # Phase 4: Verify
         session_after = loop.sessions.get_or_create("cli:test")
 
-        # Old messages should be gone
+        # The oldest messages should be trimmed from live session history
         assert not any(
             "past tense is used" in str(m.get("content", "")) for m in session_after.messages
         )
@@ -509,8 +525,8 @@ class TestAutoCompactIntegration:
         await loop.close_mcp()
 
     @pytest.mark.asyncio
-    async def test_multi_paragraph_user_message_preserved(self, tmp_path):
-        """Multi-paragraph user messages must be fully preserved after auto-new."""
+    async def test_runtime_context_markers_not_persisted_for_multi_paragraph_turn(self, tmp_path):
+        """Auto-compact resume context must not leak runtime markers into persisted session history."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
         session.add_message("user", "old message")
@@ -518,12 +534,9 @@ class TestAutoCompactIntegration:
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         # Simulate proactive archive completing before message arrives
         await loop.auto_compact._archive("cli:test")
@@ -535,16 +548,11 @@ class TestAutoCompactIntegration:
         await loop._process_message(msg)
 
         session_after = loop.sessions.get_or_create("cli:test")
-        user_msgs = [m for m in session_after.messages if m.get("role") == "user"]
-        assert len(user_msgs) >= 1
-        # All three paragraphs must be preserved
-        persisted = user_msgs[-1]["content"]
-        assert "Paragraph one" in persisted
-        assert "Paragraph two" in persisted
-        assert "Paragraph three" in persisted
-        # No runtime context markers in persisted message
-        assert "[Runtime Context" not in persisted
-        assert "[/Runtime Context]" not in persisted
+        assert any(m.get("content") == "old message" for m in session_after.messages)
+        for persisted in session_after.messages:
+            content = str(persisted.get("content", ""))
+            assert "[Runtime Context" not in content
+            assert "[/Runtime Context]" not in content
         await loop.close_mcp()
 
 
@@ -552,9 +560,12 @@ class TestProactiveAutoCompact:
     """Test proactive auto-new on idle ticks (TimeoutError path in run loop)."""
 
     @staticmethod
-    async def _run_check_expired(loop):
+    async def _run_check_expired(loop, active_session_keys=()):
         """Helper: run check_expired via callback and wait for background tasks."""
-        loop.auto_compact.check_expired(loop._schedule_background)
+        loop.auto_compact.check_expired(
+            loop._schedule_background,
+            active_session_keys=active_session_keys,
+        )
         await asyncio.sleep(0.1)
 
     @pytest.mark.asyncio
@@ -577,8 +588,7 @@ class TestProactiveAutoCompact:
         """Expired session should be archived during idle tick."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "old message")
-        session.add_message("assistant", "old response")
+        _add_turns(session, 5, prefix="old")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -586,17 +596,14 @@ class TestProactiveAutoCompact:
 
         async def _fake_archive(messages):
             archived_messages.extend(messages)
-            return True
+            return "User chatted about old things."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "User chatted about old things.",
-        }
 
         await self._run_check_expired(loop)
 
         session_after = loop.sessions.get_or_create("cli:test")
-        assert len(session_after.messages) == 0
+        assert len(session_after.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
         assert len(archived_messages) == 2
         entry = loop.auto_compact._summaries.get("cli:test")
         assert entry is not None
@@ -622,7 +629,7 @@ class TestProactiveAutoCompact:
         """Should not archive the same session twice if already in progress."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "old message")
+        _add_turns(session, 6, prefix="old")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -635,7 +642,7 @@ class TestProactiveAutoCompact:
             archive_count += 1
             started.set()
             await block_forever.wait()
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _slow_archive
 
@@ -659,7 +666,7 @@ class TestProactiveAutoCompact:
         """Proactive archive failure should be caught and not block future ticks."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "old message")
+        _add_turns(session, 6, prefix="old")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -688,7 +695,7 @@ class TestProactiveAutoCompact:
         async def _fake_archive(messages):
             nonlocal archive_called
             archive_called = True
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
 
@@ -698,12 +705,11 @@ class TestProactiveAutoCompact:
         await loop.close_mcp()
 
     @pytest.mark.asyncio
-    async def test_no_reschedule_after_successful_archive(self, tmp_path):
-        """Already-archived session should NOT be re-scheduled on subsequent ticks."""
+    async def test_skip_expired_session_with_active_agent_task(self, tmp_path):
+        """Expired session with an active agent task should NOT be archived."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "old message")
-        session.add_message("assistant", "old response")
+        _add_turns(session, 6, prefix="old")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -712,12 +718,102 @@ class TestProactiveAutoCompact:
         async def _fake_archive(messages):
             nonlocal archive_count
             archive_count += 1
-            return True
+            return "Summary."
+
+        loop.consolidator.archive = _fake_archive
+
+        # Simulate an active agent task for this session
+        await self._run_check_expired(loop, active_session_keys={"cli:test"})
+        assert archive_count == 0
+
+        session_after = loop.sessions.get_or_create("cli:test")
+        assert len(session_after.messages) == 12  # All messages preserved
+
+        await loop.close_mcp()
+
+    @pytest.mark.asyncio
+    async def test_archive_after_active_task_completes(self, tmp_path):
+        """Session should be archived on next tick after active task completes."""
+        loop = _make_loop(tmp_path, session_ttl_minutes=15)
+        session = loop.sessions.get_or_create("cli:test")
+        _add_turns(session, 6, prefix="old")
+        session.updated_at = datetime.now() - timedelta(minutes=20)
+        loop.sessions.save(session)
+
+        archive_count = 0
+
+        async def _fake_archive(messages):
+            nonlocal archive_count
+            archive_count += 1
+            return "Summary."
+
+        loop.consolidator.archive = _fake_archive
+
+        # First tick: active task, skip
+        await self._run_check_expired(loop, active_session_keys={"cli:test"})
+        assert archive_count == 0
+
+        # Second tick: task completed, should archive
+        await self._run_check_expired(loop)
+        assert archive_count == 1
+        await loop.close_mcp()
+
+    @pytest.mark.asyncio
+    async def test_partial_active_set_only_archives_inactive_expired(self, tmp_path):
+        """With multiple sessions, only the expired+inactive one should be archived."""
+        loop = _make_loop(tmp_path, session_ttl_minutes=15)
+        # Session A: expired, no active task -> should be archived
+        s1 = loop.sessions.get_or_create("cli:expired_idle")
+        _add_turns(s1, 6, prefix="old_a")
+        s1.updated_at = datetime.now() - timedelta(minutes=20)
+        loop.sessions.save(s1)
+        # Session B: expired, has active task -> should be skipped
+        s2 = loop.sessions.get_or_create("cli:expired_active")
+        _add_turns(s2, 6, prefix="old_b")
+        s2.updated_at = datetime.now() - timedelta(minutes=20)
+        loop.sessions.save(s2)
+        # Session C: recent, no active task -> should be skipped
+        s3 = loop.sessions.get_or_create("cli:recent")
+        s3.add_message("user", "recent")
+        loop.sessions.save(s3)
+
+        archive_count = 0
+
+        async def _fake_archive(messages):
+            nonlocal archive_count
+            archive_count += 1
+            return "Summary."
+
+        loop.consolidator.archive = _fake_archive
+
+        await self._run_check_expired(loop, active_session_keys={"cli:expired_active"})
+
+        assert archive_count == 1
+        s1_after = loop.sessions.get_or_create("cli:expired_idle")
+        assert len(s1_after.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
+        s2_after = loop.sessions.get_or_create("cli:expired_active")
+        assert len(s2_after.messages) == 12  # Preserved
+        s3_after = loop.sessions.get_or_create("cli:recent")
+        assert len(s3_after.messages) == 1  # Preserved
+        await loop.close_mcp()
+
+    @pytest.mark.asyncio
+    async def test_no_reschedule_after_successful_archive(self, tmp_path):
+        """Already-archived session should NOT be re-scheduled on subsequent ticks."""
+        loop = _make_loop(tmp_path, session_ttl_minutes=15)
+        session = loop.sessions.get_or_create("cli:test")
+        _add_turns(session, 5, prefix="old")
+        session.updated_at = datetime.now() - timedelta(minutes=20)
+        loop.sessions.save(session)
+
+        archive_count = 0
+
+        async def _fake_archive(messages):
+            nonlocal archive_count
+            archive_count += 1
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         # First tick: archives the session
         await self._run_check_expired(loop)
@@ -741,7 +837,7 @@ class TestProactiveAutoCompact:
         async def _fake_archive(messages):
             nonlocal archive_count
             archive_count += 1
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
 
@@ -759,8 +855,7 @@ class TestProactiveAutoCompact:
         """After successful compact + user sends new messages + idle again, should compact again."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "first conversation")
-        session.add_message("assistant", "first response")
+        _add_turns(session, 5, prefix="first")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
@@ -769,12 +864,9 @@ class TestProactiveAutoCompact:
         async def _fake_archive(messages):
             nonlocal archive_count
             archive_count += 1
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         # First compact cycle
         await loop.auto_compact._archive("cli:test")
@@ -804,18 +896,14 @@ class TestSummaryPersistence:
         """After archive, _last_summary should be in session metadata."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "hello")
-        session.add_message("assistant", "hi there")
+        _add_turns(session, 6, prefix="hello")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "User said hello."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "User said hello.",
-        }
 
         await loop.auto_compact._archive("cli:test")
 
@@ -832,19 +920,15 @@ class TestSummaryPersistence:
         """Summary should be recovered from metadata when _summaries is empty (simulates restart)."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "hello")
-        session.add_message("assistant", "hi there")
+        _add_turns(session, 6, prefix="hello")
         last_active = datetime.now() - timedelta(minutes=20)
         session.updated_at = last_active
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "User said hello."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "User said hello.",
-        }
 
         # Archive
         await loop.auto_compact._archive("cli:test")
@@ -855,6 +939,7 @@ class TestSummaryPersistence:
 
         # prepare_session should recover summary from metadata
         reloaded = loop.sessions.get_or_create("cli:test")
+        assert len(reloaded.messages) == loop.auto_compact._RECENT_SUFFIX_MESSAGES
         _, summary = loop.auto_compact.prepare_session(reloaded, "cli:test")
 
         assert summary is not None
@@ -869,17 +954,14 @@ class TestSummaryPersistence:
         """_last_summary should be removed from metadata after being consumed."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "hello")
+        _add_turns(session, 6, prefix="hello")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         await loop.auto_compact._archive("cli:test")
 
@@ -903,17 +985,14 @@ class TestSummaryPersistence:
         """In-memory _summaries path should also clean up _last_summary from metadata."""
         loop = _make_loop(tmp_path, session_ttl_minutes=15)
         session = loop.sessions.get_or_create("cli:test")
-        session.add_message("user", "hello")
+        _add_turns(session, 6, prefix="hello")
         session.updated_at = datetime.now() - timedelta(minutes=20)
         loop.sessions.save(session)
 
         async def _fake_archive(messages):
-            return True
+            return "Summary."
 
         loop.consolidator.archive = _fake_archive
-        loop.consolidator.get_last_history_entry = lambda: {
-            "cursor": 1, "timestamp": "2026-01-01 00:00", "content": "Summary.",
-        }
 
         await loop.auto_compact._archive("cli:test")
 

tests/agent/test_consolidation_ratio.py

@@ -0,0 +1,179 @@
+"""Tests for the configurable consolidation_ratio feature."""
+
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+
+from nanobot.agent.loop import AgentLoop
+import nanobot.agent.memory as memory_module
+from nanobot.bus.queue import MessageBus
+from nanobot.providers.base import GenerationSettings, LLMResponse
+
+
+def _make_loop(
+    tmp_path,
+    *,
+    estimated_tokens: int = 0,
+    context_window_tokens: int = 200,
+    consolidation_ratio: float = 0.5,
+) -> AgentLoop:
+    provider = MagicMock()
+    provider.get_default_model.return_value = "test-model"
+    provider.generation = GenerationSettings(max_tokens=0)
+    provider.estimate_prompt_tokens.return_value = (estimated_tokens, "test-counter")
+    _response = LLMResponse(content="ok", tool_calls=[])
+    provider.chat_with_retry = AsyncMock(return_value=_response)
+    provider.chat_stream_with_retry = AsyncMock(return_value=_response)
+
+    loop = AgentLoop(
+        bus=MessageBus(),
+        provider=provider,
+        workspace=tmp_path,
+        model="test-model",
+        context_window_tokens=context_window_tokens,
+        consolidation_ratio=consolidation_ratio,
+    )
+    loop.tools.get_definitions = MagicMock(return_value=[])
+    loop.consolidator._SAFETY_BUFFER = 0
+    return loop
+
+
+@pytest.mark.asyncio
+async def test_default_ratio_uses_half_budget(tmp_path, monkeypatch) -> None:
+    """With ratio=0.5 (default), target should be half of budget."""
+    loop = _make_loop(tmp_path, context_window_tokens=200, consolidation_ratio=0.5)
+    loop.consolidator.archive = AsyncMock(return_value=True)  # type: ignore[method-assign]
+
+    session = loop.sessions.get_or_create("cli:test")
+    session.messages = [
+        {"role": "user", "content": "u1", "timestamp": "2026-01-01T00:00:00"},
+        {"role": "assistant", "content": "a1", "timestamp": "2026-01-01T00:00:01"},
+        {"role": "user", "content": "u2", "timestamp": "2026-01-01T00:00:02"},
+        {"role": "assistant", "content": "a2", "timestamp": "2026-01-01T00:00:03"},
+        {"role": "user", "content": "u3", "timestamp": "2026-01-01T00:00:04"},
+        {"role": "assistant", "content": "a3", "timestamp": "2026-01-01T00:00:05"},
+        {"role": "user", "content": "u4", "timestamp": "2026-01-01T00:00:06"},
+    ]
+    loop.sessions.save(session)
+
+    # budget = 200 - 0 (max_tokens) - 0 (safety_buffer) = 200
+    # target = int(200 * 0.5) = 100
+    # estimated must be >= budget to trigger consolidation
+    call_count = [0]
+
+    def mock_estimate(_session):
+        call_count[0] += 1
+        if call_count[0] == 1:
+            return (250, "test")
+        return (90, "test")
+
+    loop.consolidator.estimate_session_prompt_tokens = mock_estimate  # type: ignore[method-assign]
+    monkeypatch.setattr(memory_module, "estimate_message_tokens", lambda _m: 100)
+
+    await loop.consolidator.maybe_consolidate_by_tokens(session)
+
+    # 250 >= 200 (budget, triggers) → 250 > 100 (target) → archive → 90 < 100, stops.
+    assert loop.consolidator.archive.await_count == 1
+
+
+@pytest.mark.asyncio
+async def test_low_ratio_aggressively_consolidates(tmp_path, monkeypatch) -> None:
+    """With ratio=0.1, target is only 10% of budget — more rounds of archiving."""
+    loop = _make_loop(tmp_path, context_window_tokens=1000, consolidation_ratio=0.1)
+    loop.consolidator.archive = AsyncMock(return_value=True)  # type: ignore[method-assign]
+
+    session = loop.sessions.get_or_create("cli:test")
+    # Interleave user/assistant so pick_consolidation_boundary can find boundaries
+    session.messages = []
+    for i in range(10):
+        session.messages.append({"role": "user", "content": f"u{i}", "timestamp": f"2026-01-01T00:00:{i:02d}"})
+        session.messages.append({"role": "assistant", "content": f"a{i}", "timestamp": f"2026-01-01T00:00:{i:02d}"})
+    loop.sessions.save(session)
+
+    # budget = 1000, target = int(1000 * 0.1) = 100
+    call_count = [0]
+
+    def mock_estimate(_session):
+        call_count[0] += 1
+        if call_count[0] == 1:
+            return (1200, "test")
+        if call_count[0] == 2:
+            return (800, "test")
+        if call_count[0] == 3:
+            return (400, "test")
+        return (50, "test")
+
+    loop.consolidator.estimate_session_prompt_tokens = mock_estimate  # type: ignore[method-assign]
+    monkeypatch.setattr(memory_module, "estimate_message_tokens", lambda _m: 100)
+
+    await loop.consolidator.maybe_consolidate_by_tokens(session)
+
+    # With low ratio, more rounds needed to reach target; at least 2 rounds
+    assert loop.consolidator.archive.await_count >= 2
+
+
+@pytest.mark.asyncio
+async def test_high_ratio_preserves_more_history(tmp_path, monkeypatch) -> None:
+    """With ratio=0.9, target is 90% of budget — consolidation stops sooner."""
+    loop = _make_loop(tmp_path, context_window_tokens=200, consolidation_ratio=0.9)
+    loop.consolidator.archive = AsyncMock(return_value=True)  # type: ignore[method-assign]
+
+    session = loop.sessions.get_or_create("cli:test")
+    session.messages = [
+        {"role": "user", "content": "u1", "timestamp": "2026-01-01T00:00:00"},
+        {"role": "assistant", "content": "a1", "timestamp": "2026-01-01T00:00:01"},
+        {"role": "user", "content": "u2", "timestamp": "2026-01-01T00:00:02"},
+        {"role": "assistant", "content": "a2", "timestamp": "2026-01-01T00:00:03"},
+        {"role": "user", "content": "u3", "timestamp": "2026-01-01T00:00:04"},
+        {"role": "assistant", "content": "a3", "timestamp": "2026-01-01T00:00:05"},
+        {"role": "user", "content": "u4", "timestamp": "2026-01-01T00:00:06"},
+    ]
+    loop.sessions.save(session)
+
+    # budget = 200, target = int(200 * 0.9) = 180
+    call_count = [0]
+
+    def mock_estimate(_session):
+        call_count[0] += 1
+        if call_count[0] == 1:
+            return (300, "test")
+        return (175, "test")
+
+    loop.consolidator.estimate_session_prompt_tokens = mock_estimate  # type: ignore[method-assign]
+    monkeypatch.setattr(memory_module, "estimate_message_tokens", lambda _m: 100)
+
+    await loop.consolidator.maybe_consolidate_by_tokens(session)
+
+    # 300 >= 200 (triggers) → 300 > 180 → archive → 175 < 180 → stop
+    assert loop.consolidator.archive.await_count == 1
+
+
+@pytest.mark.asyncio
+async def test_ratio_propagated_from_config_schema() -> None:
+    """Verify consolidation_ratio is parsed from config with camelCase alias."""
+    from nanobot.config.schema import AgentDefaults
+
+    # Default
+    defaults = AgentDefaults()
+    assert defaults.consolidation_ratio == 0.5
+
+    # camelCase alias
+    defaults = AgentDefaults.model_validate({"consolidationRatio": 0.3})
+    assert defaults.consolidation_ratio == 0.3
+
+    # Serialization uses alias
+    dumped = defaults.model_dump(by_alias=True)
+    assert dumped["consolidationRatio"] == 0.3
+
+
+@pytest.mark.asyncio
+async def test_ratio_validation_rejects_out_of_range() -> None:
+    """Invalid ratio values should be rejected by validation."""
+    from pydantic import ValidationError
+    from nanobot.config.schema import AgentDefaults
+
+    with pytest.raises(ValidationError):
+        AgentDefaults(consolidation_ratio=0.05)
+
+    with pytest.raises(ValidationError):
+        AgentDefaults(consolidation_ratio=1.0)

tests/agent/test_consolidator.py

@@ -46,7 +46,7 @@ class TestConsolidatorSummarize:
             {"role": "assistant", "content": "Done, fixed the race condition."},
         ]
         result = await consolidator.archive(messages)
-        assert result is True
+        assert result == "User fixed a bug in the auth module."
         entries = store.read_unprocessed_history(since_cursor=0)
         assert len(entries) == 1
 
@@ -55,14 +55,54 @@ class TestConsolidatorSummarize:
         mock_provider.chat_with_retry.side_effect = Exception("API error")
         messages = [{"role": "user", "content": "hello"}]
         result = await consolidator.archive(messages)
-        assert result is True  # always succeeds
+        assert result is None  # no summary on raw dump fallback
         entries = store.read_unprocessed_history(since_cursor=0)
         assert len(entries) == 1
         assert "[RAW]" in entries[0]["content"]
 
     async def test_summarize_skips_empty_messages(self, consolidator):
         result = await consolidator.archive([])
-        assert result is False
+        assert result is None
+
+
+class TestConsolidatorArchiveErrorHandling:
+    """archive() must fall back to raw_archive when the LLM returns an error
+    response (finish_reason == 'error'), e.g. overloaded / quota exceeded.
+    See https://github.com/HKUDS/nanobot/issues/3244
+    """
+
+    async def test_archive_falls_back_on_error_finish_reason(self, consolidator, mock_provider, store):
+        """LLM returning finish_reason='error' should trigger raw_archive, not write error text."""
+        mock_provider.chat_with_retry.return_value = MagicMock(
+            content="Error: {'type': 'error', 'error': {'type': 'overloaded_error', 'message': 'overloaded_error (529)'}}",
+            finish_reason="error",
+        )
+        messages = [
+            {"role": "user", "content": "fix the auth bug"},
+            {"role": "assistant", "content": "Done, fixed the race condition."},
+        ]
+        result = await consolidator.archive(messages)
+        assert result is None
+        entries = store.read_unprocessed_history(since_cursor=0)
+        assert len(entries) == 1
+        assert "[RAW]" in entries[0]["content"]
+        assert "Error:" not in entries[0]["content"]
+
+    async def test_archive_preserves_summary_on_success(self, consolidator, mock_provider, store):
+        """Normal LLM response should still produce a proper summary entry."""
+        mock_provider.chat_with_retry.return_value = MagicMock(
+            content="User fixed a bug in the auth module.",
+            finish_reason="stop",
+        )
+        messages = [
+            {"role": "user", "content": "fix the auth bug"},
+            {"role": "assistant", "content": "Done."},
+        ]
+        result = await consolidator.archive(messages)
+        assert result == "User fixed a bug in the auth module."
+        entries = store.read_unprocessed_history(since_cursor=0)
+        assert len(entries) == 1
+        assert "[RAW]" not in entries[0]["content"]
 
 
 class TestConsolidatorTokenBudget:
@@ -76,3 +116,52 @@ class TestConsolidatorTokenBudget:
         consolidator.archive = AsyncMock(return_value=True)
         await consolidator.maybe_consolidate_by_tokens(session)
         consolidator.archive.assert_not_called()
+
+    async def test_chunk_cap_preserves_user_turn_boundary(self, consolidator):
+        """Chunk cap should rewind to the last user boundary within the cap."""
+        consolidator._SAFETY_BUFFER = 0
+        session = MagicMock()
+        session.last_consolidated = 0
+        session.key = "test:key"
+        session.messages = [
+            {
+                "role": "user" if i in {0, 50, 61} else "assistant",
+                "content": f"m{i}",
+            }
+            for i in range(70)
+        ]
+        consolidator.estimate_session_prompt_tokens = MagicMock(
+            side_effect=[(1200, "tiktoken"), (400, "tiktoken")]
+        )
+        consolidator.pick_consolidation_boundary = MagicMock(return_value=(61, 999))
+        consolidator.archive = AsyncMock(return_value=True)
+
+        await consolidator.maybe_consolidate_by_tokens(session)
+
+        archived_chunk = consolidator.archive.await_args.args[0]
+        assert len(archived_chunk) == 50
+        assert archived_chunk[0]["content"] == "m0"
+        assert archived_chunk[-1]["content"] == "m49"
+        assert session.last_consolidated == 50
+
+    async def test_chunk_cap_skips_when_no_user_boundary_within_cap(self, consolidator):
+        """If the cap would cut mid-turn, consolidation should skip that round."""
+        consolidator._SAFETY_BUFFER = 0
+        session = MagicMock()
+        session.last_consolidated = 0
+        session.key = "test:key"
+        session.messages = [
+            {
+                "role": "user" if i in {0, 61} else "assistant",
+                "content": f"m{i}",
+            }
+            for i in range(70)
+        ]
+        consolidator.estimate_session_prompt_tokens = MagicMock(return_value=(1200, "tiktoken"))
+        consolidator.pick_consolidation_boundary = MagicMock(return_value=(61, 999))
+        consolidator.archive = AsyncMock(return_value=True)
+
+        await consolidator.maybe_consolidate_by_tokens(session)
+
+        consolidator.archive.assert_not_awaited()
+        assert session.last_consolidated == 0

tests/agent/test_context_prompt_cache.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import re
 from datetime import datetime as real_datetime
 from importlib.resources import files as pkg_files
 from pathlib import Path
@@ -86,6 +87,124 @@ def test_runtime_context_is_separate_untrusted_user_message(tmp_path) -> None:
     assert "Return exactly: OK" in user_content
 
 
+def test_unprocessed_history_injected_into_system_prompt(tmp_path) -> None:
+    """Entries in history.jsonl not yet consumed by Dream appear with timestamps."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    builder.memory.append_history("User asked about weather in Tokyo")
+    builder.memory.append_history("Agent fetched forecast via web_search")
+
+    prompt = builder.build_system_prompt()
+    assert "# Recent History" in prompt
+    assert "User asked about weather in Tokyo" in prompt
+    assert "Agent fetched forecast via web_search" in prompt
+    assert re.search(r"\[\d{4}-\d{2}-\d{2} \d{2}:\d{2}\]", prompt)
+
+
+def test_recent_history_capped_at_max(tmp_path) -> None:
+    """Only the most recent _MAX_RECENT_HISTORY entries are injected."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    for i in range(builder._MAX_RECENT_HISTORY + 20):
+        builder.memory.append_history(f"entry-{i}")
+
+    prompt = builder.build_system_prompt()
+    assert "entry-0" not in prompt
+    assert "entry-19" not in prompt
+    assert f"entry-{builder._MAX_RECENT_HISTORY + 19}" in prompt
+
+
+def test_no_recent_history_when_dream_has_processed_all(tmp_path) -> None:
+    """If Dream has consumed everything, no Recent History section should appear."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    cursor = builder.memory.append_history("already processed entry")
+    builder.memory.set_last_dream_cursor(cursor)
+
+    prompt = builder.build_system_prompt()
+    assert "# Recent History" not in prompt
+
+
+def test_partial_dream_processing_shows_only_remainder(tmp_path) -> None:
+    """When Dream has processed some entries, only the unprocessed ones appear."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    c1 = builder.memory.append_history("old conversation about Python")
+    c2 = builder.memory.append_history("old conversation about Rust")
+    builder.memory.append_history("recent question about Docker")
+    builder.memory.append_history("recent question about K8s")
+
+    builder.memory.set_last_dream_cursor(c2)
+
+    prompt = builder.build_system_prompt()
+    assert "# Recent History" in prompt
+    assert "old conversation about Python" not in prompt
+    assert "old conversation about Rust" not in prompt
+    assert "recent question about Docker" in prompt
+    assert "recent question about K8s" in prompt
+
+
+def test_execution_rules_in_system_prompt(tmp_path) -> None:
+    """New execution rules should appear in the system prompt."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    prompt = builder.build_system_prompt()
+    assert "Act, don't narrate" in prompt
+    assert "Read before you write" in prompt
+    assert "verify the result" in prompt
+
+
+def test_channel_format_hint_telegram(tmp_path) -> None:
+    """Telegram channel should get messaging-app format hint."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    prompt = builder.build_system_prompt(channel="telegram")
+    assert "Format Hint" in prompt
+    assert "messaging app" in prompt
+
+
+def test_channel_format_hint_whatsapp(tmp_path) -> None:
+    """WhatsApp should get plain-text format hint."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    prompt = builder.build_system_prompt(channel="whatsapp")
+    assert "Format Hint" in prompt
+    assert "plain text only" in prompt
+
+
+def test_channel_format_hint_absent_for_unknown(tmp_path) -> None:
+    """Unknown or None channel should not inject a format hint."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    prompt = builder.build_system_prompt(channel=None)
+    assert "Format Hint" not in prompt
+
+    prompt2 = builder.build_system_prompt(channel="feishu")
+    assert "Format Hint" not in prompt2
+
+
+def test_build_messages_passes_channel_to_system_prompt(tmp_path) -> None:
+    """build_messages should pass channel through to build_system_prompt."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    messages = builder.build_messages(
+        history=[], current_message="hi",
+        channel="telegram", chat_id="123",
+    )
+    system = messages[0]["content"]
+    assert "Format Hint" in system
+    assert "messaging app" in system
+
+
 def test_subagent_result_does_not_create_consecutive_assistant_messages(tmp_path) -> None:
     workspace = _make_workspace(tmp_path)
     builder = ContextBuilder(workspace)
@@ -100,3 +219,55 @@ def test_subagent_result_does_not_create_consecutive_assistant_messages(tmp_path
 
     for left, right in zip(messages, messages[1:]):
         assert not (left.get("role") == right.get("role") == "assistant")
+
+
+def test_always_skills_excluded_from_skills_index(tmp_path) -> None:
+    """Always skills should appear in Active Skills but NOT in the skills index."""
+    workspace = _make_workspace(tmp_path)
+    builder = ContextBuilder(workspace)
+
+    prompt = builder.build_system_prompt()
+
+    # memory skill should be in Active Skills section
+    assert "# Active Skills" in prompt
+    assert "### Skill: memory" in prompt
+
+    # memory skill should NOT appear in the skills index
+    skills_section = prompt.split("# Skills\n", 1)
+    if len(skills_section) > 1:
+        index_text = skills_section[1].split("\n\n---")[0]
+        assert "**memory**" not in index_text
+
+
+def test_template_memory_md_is_skipped(tmp_path) -> None:
+    """MEMORY.md matching the bundled template should not inject the Memory section."""
+    workspace = _make_workspace(tmp_path)
+    from nanobot.utils.helpers import sync_workspace_templates
+    sync_workspace_templates(workspace, silent=True)
+
+    builder = ContextBuilder(workspace)
+    prompt = builder.build_system_prompt()
+
+    # The "# Memory\n\n## Long-term Memory" block is produced only by
+    # build_system_prompt() when MEMORY.md is injected.  The memory skill
+    # also contains "# Memory" but is followed by "## Structure", not
+    # "## Long-term Memory".
+    assert "# Memory\n\n## Long-term Memory" not in prompt
+    assert "This file is automatically updated by nanobot" not in prompt
+
+
+def test_customized_memory_md_is_injected(tmp_path) -> None:
+    """A Dream-populated MEMORY.md should be injected normally."""
+    workspace = _make_workspace(tmp_path)
+    from nanobot.utils.helpers import sync_workspace_templates
+    sync_workspace_templates(workspace, silent=True)
+
+    (workspace / "memory" / "MEMORY.md").write_text(
+        "# Long-term Memory\n\nUser prefers dark mode.\n", encoding="utf-8"
+    )
+
+    builder = ContextBuilder(workspace)
+    prompt = builder.build_system_prompt()
+
+    assert "# Memory\n\n## Long-term Memory" in prompt
+    assert "User prefers dark mode" in prompt

tests/agent/test_dream.py

@@ -2,10 +2,12 @@
 
 import pytest
 
-from unittest.mock import AsyncMock, MagicMock
+from unittest.mock import AsyncMock, MagicMock, patch
 
 from nanobot.agent.memory import Dream, MemoryStore
 from nanobot.agent.runner import AgentRunResult
+from nanobot.agent.skills import BUILTIN_SKILLS_DIR
+from nanobot.utils.gitstore import LineAge
 
 
 @pytest.fixture
@@ -95,3 +97,162 @@ class TestDreamRun:
         entries = store.read_unprocessed_history(since_cursor=0)
         assert all(e["cursor"] > 0 for e in entries)
 
+    async def test_skill_phase_uses_builtin_skill_creator_path(self, dream, mock_provider, mock_runner, store):
+        """Dream should point skill creation guidance at the builtin skill-creator template."""
+        store.append_history("Repeated workflow one")
+        store.append_history("Repeated workflow two")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKILL] test-skill: test description")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        await dream.run()
+
+        spec = mock_runner.run.call_args[0][0]
+        system_prompt = spec.initial_messages[0]["content"]
+        expected = str(BUILTIN_SKILLS_DIR / "skill-creator" / "SKILL.md")
+        assert expected in system_prompt
+
+    async def test_skill_write_tool_accepts_workspace_relative_skill_path(self, dream, store):
+        """Dream skill creation should allow skills/<name>/SKILL.md relative to workspace root."""
+        write_tool = dream._tools.get("write_file")
+        assert write_tool is not None
+
+        result = await write_tool.execute(
+            path="skills/test-skill/SKILL.md",
+            content="---\nname: test-skill\ndescription: Test\n---\n",
+        )
+
+        assert "Successfully wrote" in result
+        assert (store.workspace / "skills" / "test-skill" / "SKILL.md").exists()
+
+    async def test_phase1_prompt_includes_line_age_annotations(self, dream, mock_provider, mock_runner, store):
+        """Phase 1 prompt should have per-line age suffixes in MEMORY.md when git is available."""
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        # Init git so line_ages works
+        store.git.init()
+        store.git.auto_commit("initial memory state")
+
+        await dream.run()
+
+        # The MEMORY.md section should not crash and should contain the memory content
+        call_args = mock_provider.chat_with_retry.call_args
+        user_msg = call_args.kwargs.get("messages", call_args[1].get("messages"))[1]["content"]
+        assert "## Current MEMORY.md" in user_msg
+
+    async def test_phase1_annotates_only_memory_not_soul_or_user(self, dream, mock_provider, mock_runner, store):
+        """SOUL.md and USER.md should never have age annotations — they are permanent."""
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        store.git.init()
+        store.git.auto_commit("initial state")
+
+        await dream.run()
+
+        call_args = mock_provider.chat_with_retry.call_args
+        user_msg = call_args.kwargs.get("messages", call_args[1].get("messages"))[1]["content"]
+        # The ← suffix should only appear in MEMORY.md section
+        memory_section = user_msg.split("## Current MEMORY.md")[1].split("## Current SOUL.md")[0]
+        soul_section = user_msg.split("## Current SOUL.md")[1].split("## Current USER.md")[0]
+        user_section = user_msg.split("## Current USER.md")[1]
+        # SOUL and USER should not contain age arrows
+        assert "\u2190" not in soul_section
+        assert "\u2190" not in user_section
+
+    async def test_phase1_prompt_works_without_git(self, dream, mock_provider, mock_runner, store):
+        """Phase 1 should work fine even if git is not initialized (no age annotations)."""
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        await dream.run()
+
+        # Should still succeed — just without age annotations
+        mock_provider.chat_with_retry.assert_called_once()
+        call_args = mock_provider.chat_with_retry.call_args
+        user_msg = call_args.kwargs.get("messages", call_args[1].get("messages"))[1]["content"]
+        assert "## Current MEMORY.md" in user_msg
+
+    async def test_phase1_prompt_carries_age_suffix_for_stale_lines(
+        self, dream, mock_provider, mock_runner, store,
+    ):
+        """End-to-end: ages >14d must appear verbatim in the LLM prompt, ages ≤14d must not."""
+        # MEMORY.md fixture has 2 non-blank lines ("# Memory" and "- Project X active").
+        # Inject four ages to cover threshold boundaries: >14 suffix, ==14 no suffix, <14 no suffix.
+        store.write_memory("# Memory\n- Project X active\n- fresh item\n- edge case line")
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        fake_ages = [
+            LineAge(age_days=30),   # "# Memory"        → should get ← 30d
+            LineAge(age_days=20),   # "- Project X..."  → should get ← 20d
+            LineAge(age_days=14),   # "- fresh item"    → ==14, threshold is strictly >14, no suffix
+            LineAge(age_days=5),    # "- edge case..."  → no suffix
+        ]
+        with patch.object(store.git, "line_ages", return_value=fake_ages):
+            await dream.run()
+
+        call_args = mock_provider.chat_with_retry.call_args
+        user_msg = call_args.kwargs.get("messages", call_args[1].get("messages"))[1]["content"]
+        memory_section = user_msg.split("## Current MEMORY.md")[1].split("## Current SOUL.md")[0]
+        assert "\u2190 30d" in memory_section
+        assert "\u2190 20d" in memory_section
+        assert "\u2190 14d" not in memory_section
+        assert "\u2190 5d" not in memory_section
+
+    async def test_phase1_skips_annotation_when_disabled(
+        self, dream, mock_provider, mock_runner, store,
+    ):
+        """`annotate_line_ages=False` must bypass the git lookup entirely and keep MEMORY.md raw."""
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        dream.annotate_line_ages = False
+        # line_ages must be bypassed entirely — verify with a spy rather than a
+        # raising side_effect, because _annotate_with_ages catches Exception
+        # (which swallows AssertionError) and would hide an accidental call.
+        with patch.object(store.git, "line_ages") as mock_line_ages:
+            await dream.run()
+            mock_line_ages.assert_not_called()
+
+        call_args = mock_provider.chat_with_retry.call_args
+        user_msg = call_args.kwargs.get("messages", call_args[1].get("messages"))[1]["content"]
+        assert "\u2190" not in user_msg
+
+    async def test_phase1_skips_annotation_on_line_ages_length_mismatch(
+        self, dream, mock_provider, mock_runner, store,
+    ):
+        """If ages length != lines length (dirty working tree), skip annotation instead of mis-tagging."""
+        # MEMORY.md has 2 non-blank lines but we hand back only 1 age → mismatch.
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        with patch.object(store.git, "line_ages", return_value=[LineAge(age_days=999)]):
+            await dream.run()
+
+        call_args = mock_provider.chat_with_retry.call_args
+        user_msg = call_args.kwargs.get("messages", call_args[1].get("messages"))[1]["content"]
+        memory_section = user_msg.split("## Current MEMORY.md")[1].split("## Current SOUL.md")[0]
+        # No age arrow at all — we refused to annotate rather than tag the wrong line.
+        assert "\u2190" not in memory_section
+
+    async def test_phase1_prompt_uses_threshold_from_template_var(
+        self, dream, mock_provider, mock_runner, store,
+    ):
+        """System prompt should reference the stale-threshold constant, not a hardcoded 14."""
+        store.append_history("some event")
+        mock_provider.chat_with_retry.return_value = MagicMock(content="[SKIP]")
+        mock_runner.run = AsyncMock(return_value=_make_run_result())
+
+        await dream.run()
+
+        system_msg = mock_provider.chat_with_retry.call_args.kwargs["messages"][0]["content"]
+        # The template renders with stale_threshold_days=14 → LLM must see "N>14"
+        assert "N>14" in system_msg
+

tests/agent/test_gemini_thought_signature.py

@@ -184,17 +184,22 @@ def test_stale_extra_content_in_tool_calls_survives_sanitize() -> None:
     with patch("nanobot.providers.openai_compat_provider.AsyncOpenAI"):
         provider = OpenAICompatProvider()
 
-    messages = [{
-        "role": "assistant",
-        "content": None,
-        "tool_calls": [{
-            "id": "call_1",
-            "type": "function",
-            "function": {"name": "fn", "arguments": "{}"},
-            "extra_content": GEMINI_EXTRA,
-        }],
-    }]
+    messages = [
+        {"role": "user", "content": "hi"},
+        {
+            "role": "assistant",
+            "content": None,
+            "tool_calls": [{
+                "id": "call_1",
+                "type": "function",
+                "function": {"name": "fn", "arguments": "{}"},
+                "extra_content": GEMINI_EXTRA,
+            }],
+        },
+        {"role": "tool", "content": "ok", "tool_call_id": "call_1"},
+        {"role": "user", "content": "thanks"},
+    ]
 
     sanitized = provider._sanitize_messages(messages)
 
-    assert sanitized[0]["tool_calls"][0]["extra_content"] == GEMINI_EXTRA
+    assert sanitized[1]["tool_calls"][0]["extra_content"] == GEMINI_EXTRA

tests/agent/test_hook_composite.py

@@ -232,6 +232,35 @@ async def test_composite_empty_hooks_no_ops():
     assert hook.finalize_content(ctx, "test") == "test"
 
 
+@pytest.mark.asyncio
+async def test_composite_supports_legacy_hook_init_without_super():
+    calls: list[str] = []
+
+    class LegacyHook(AgentHook):
+        def __init__(self, label: str) -> None:
+            self.label = label
+
+        async def before_iteration(self, context: AgentHookContext) -> None:
+            calls.append(self.label)
+
+    hook = CompositeHook([LegacyHook("legacy")])
+    await hook.before_iteration(_ctx())
+    assert calls == ["legacy"]
+
+
+@pytest.mark.asyncio
+async def test_composite_can_wrap_another_composite():
+    calls: list[str] = []
+
+    class Inner(AgentHook):
+        async def before_iteration(self, context: AgentHookContext) -> None:
+            calls.append("inner")
+
+    hook = CompositeHook([CompositeHook([Inner()])])
+    await hook.before_iteration(_ctx())
+    assert calls == ["inner"]
+
+
 # ---------------------------------------------------------------------------
 # Integration: AgentLoop with extra hooks
 # ---------------------------------------------------------------------------

tests/agent/test_loop_save_turn.py

@@ -1,5 +1,13 @@
+import asyncio
+from pathlib import Path
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+
 from nanobot.agent.context import ContextBuilder
 from nanobot.agent.loop import AgentLoop
+from nanobot.bus.events import InboundMessage
+from nanobot.bus.queue import MessageBus
 from nanobot.session.manager import Session
 
 
@@ -11,6 +19,12 @@ def _mk_loop() -> AgentLoop:
     return loop
 
 
+def _make_full_loop(tmp_path: Path) -> AgentLoop:
+    provider = MagicMock()
+    provider.get_default_model.return_value = "test-model"
+    return AgentLoop(bus=MessageBus(), provider=provider, workspace=tmp_path, model="test-model")
+
+
 def test_save_turn_skips_multimodal_user_when_only_runtime_context() -> None:
     loop = _mk_loop()
     session = Session(key="test:runtime-only")
@@ -200,3 +214,206 @@ def test_restore_runtime_checkpoint_dedupes_overlapping_tail() -> None:
     assert session.messages[0]["role"] == "assistant"
     assert session.messages[1]["tool_call_id"] == "call_done"
     assert session.messages[2]["tool_call_id"] == "call_pending"
+
+
+@pytest.mark.asyncio
+async def test_process_message_persists_user_message_before_turn_completes(tmp_path: Path) -> None:
+    loop = _make_full_loop(tmp_path)
+    loop.consolidator.maybe_consolidate_by_tokens = AsyncMock(return_value=False)  # type: ignore[method-assign]
+    loop._run_agent_loop = AsyncMock(side_effect=RuntimeError("boom"))  # type: ignore[method-assign]
+
+    msg = InboundMessage(channel="feishu", sender_id="u1", chat_id="c1", content="persist me")
+    with pytest.raises(RuntimeError, match="boom"):
+        await loop._process_message(msg)
+
+    loop.sessions.invalidate("feishu:c1")
+    persisted = loop.sessions.get_or_create("feishu:c1")
+    assert [m["role"] for m in persisted.messages] == ["user"]
+    assert persisted.messages[0]["content"] == "persist me"
+    assert persisted.metadata.get(AgentLoop._PENDING_USER_TURN_KEY) is True
+    assert persisted.updated_at >= persisted.created_at
+
+
+@pytest.mark.asyncio
+async def test_process_message_does_not_duplicate_early_persisted_user_message(tmp_path: Path) -> None:
+    loop = _make_full_loop(tmp_path)
+    loop.consolidator.maybe_consolidate_by_tokens = AsyncMock(return_value=False)  # type: ignore[method-assign]
+    loop._run_agent_loop = AsyncMock(return_value=(
+        "done",
+        None,
+        [
+            {"role": "system", "content": "system"},
+            {"role": "user", "content": "hello"},
+            {"role": "assistant", "content": "done"},
+        ],
+        "stop",
+        False,
+    ))  # type: ignore[method-assign]
+
+    result = await loop._process_message(
+        InboundMessage(channel="feishu", sender_id="u1", chat_id="c2", content="hello")
+    )
+
+    assert result is not None
+    assert result.content == "done"
+    session = loop.sessions.get_or_create("feishu:c2")
+    assert [
+        {k: v for k, v in m.items() if k in {"role", "content"}}
+        for m in session.messages
+    ] == [
+        {"role": "user", "content": "hello"},
+        {"role": "assistant", "content": "done"},
+    ]
+    assert AgentLoop._PENDING_USER_TURN_KEY not in session.metadata
+
+
+@pytest.mark.asyncio
+async def test_next_turn_after_crash_closes_pending_user_turn_before_new_input(tmp_path: Path) -> None:
+    loop = _make_full_loop(tmp_path)
+    loop.consolidator.maybe_consolidate_by_tokens = AsyncMock(return_value=False)  # type: ignore[method-assign]
+    loop.provider.chat_with_retry = AsyncMock(return_value=MagicMock())  # unused because _run_agent_loop is stubbed
+
+    session = loop.sessions.get_or_create("feishu:c3")
+    session.add_message("user", "old question")
+    session.metadata[AgentLoop._PENDING_USER_TURN_KEY] = True
+    loop.sessions.save(session)
+
+    loop._run_agent_loop = AsyncMock(return_value=(
+        "new answer",
+        None,
+        [
+            {"role": "system", "content": "system"},
+            {"role": "user", "content": "old question"},
+            {"role": "assistant", "content": "Error: Task interrupted before a response was generated."},
+            {"role": "user", "content": "new question"},
+            {"role": "assistant", "content": "new answer"},
+        ],
+        "stop",
+        False,
+    ))  # type: ignore[method-assign]
+
+    result = await loop._process_message(
+        InboundMessage(channel="feishu", sender_id="u1", chat_id="c3", content="new question")
+    )
+
+    assert result is not None
+    assert result.content == "new answer"
+    session = loop.sessions.get_or_create("feishu:c3")
+    assert [
+        {k: v for k, v in m.items() if k in {"role", "content"}}
+        for m in session.messages
+    ] == [
+        {"role": "user", "content": "old question"},
+        {"role": "assistant", "content": "Error: Task interrupted before a response was generated."},
+        {"role": "user", "content": "new question"},
+        {"role": "assistant", "content": "new answer"},
+    ]
+    assert AgentLoop._PENDING_USER_TURN_KEY not in session.metadata
+
+
+@pytest.mark.asyncio
+async def test_stop_preserves_runtime_checkpoint_for_next_turn(tmp_path: Path) -> None:
+    from nanobot.command.builtin import cmd_stop
+    from nanobot.command.router import CommandContext
+
+    loop = _make_full_loop(tmp_path)
+    loop.consolidator.maybe_consolidate_by_tokens = AsyncMock(return_value=False)  # type: ignore[method-assign]
+
+    checkpoint_saved = asyncio.Event()
+
+    async def interrupted_run_agent_loop(_initial_messages, *, session=None, **_kwargs):
+        assert session is not None
+        loop._set_runtime_checkpoint(
+            session,
+            {
+                "assistant_message": {
+                    "role": "assistant",
+                    "content": "working",
+                    "tool_calls": [
+                        {
+                            "id": "call_done",
+                            "type": "function",
+                            "function": {"name": "read_file", "arguments": "{}"},
+                        },
+                        {
+                            "id": "call_pending",
+                            "type": "function",
+                            "function": {"name": "exec", "arguments": "{}"},
+                        },
+                    ],
+                },
+                "completed_tool_results": [
+                    {
+                        "role": "tool",
+                        "tool_call_id": "call_done",
+                        "name": "read_file",
+                        "content": "ok",
+                    }
+                ],
+                "pending_tool_calls": [
+                    {
+                        "id": "call_pending",
+                        "type": "function",
+                        "function": {"name": "exec", "arguments": "{}"},
+                    }
+                ],
+            },
+        )
+        checkpoint_saved.set()
+        await asyncio.Event().wait()
+
+    loop._run_agent_loop = interrupted_run_agent_loop  # type: ignore[method-assign]
+
+    first_msg = InboundMessage(channel="feishu", sender_id="u1", chat_id="c4", content="keep progress")
+    task = asyncio.create_task(loop._process_message(first_msg))
+    loop._active_tasks[first_msg.session_key] = [task]
+    await asyncio.wait_for(checkpoint_saved.wait(), timeout=1.0)
+
+    stop_msg = InboundMessage(channel="feishu", sender_id="u1", chat_id="c4", content="/stop")
+    stop_ctx = CommandContext(msg=stop_msg, session=None, key=stop_msg.session_key, raw="/stop", loop=loop)
+    stop_result = await cmd_stop(stop_ctx)
+
+    assert "Stopped 1 task" in stop_result.content
+    assert task.done()
+
+    loop.sessions.invalidate("feishu:c4")
+    interrupted = loop.sessions.get_or_create("feishu:c4")
+    assert interrupted.metadata.get(AgentLoop._PENDING_USER_TURN_KEY) is True
+    assert interrupted.metadata.get(AgentLoop._RUNTIME_CHECKPOINT_KEY) is not None
+
+    async def resumed_run_agent_loop(initial_messages, **_kwargs):
+        return (
+            "next answer",
+            None,
+            [*initial_messages, {"role": "assistant", "content": "next answer"}],
+            "stop",
+            False,
+        )
+
+    loop._run_agent_loop = resumed_run_agent_loop  # type: ignore[method-assign]
+    result = await loop._process_message(
+        InboundMessage(channel="feishu", sender_id="u1", chat_id="c4", content="continue here")
+    )
+
+    assert result is not None
+    assert result.content == "next answer"
+
+    session = loop.sessions.get_or_create("feishu:c4")
+    assert [
+        {k: v for k, v in m.items() if k in {"role", "content", "tool_call_id", "name"}}
+        for m in session.messages
+    ] == [
+        {"role": "user", "content": "keep progress"},
+        {"role": "assistant", "content": "working"},
+        {"role": "tool", "tool_call_id": "call_done", "name": "read_file", "content": "ok"},
+        {
+            "role": "tool",
+            "tool_call_id": "call_pending",
+            "name": "exec",
+            "content": "Error: Task interrupted before this tool finished.",
+        },
+        {"role": "user", "content": "continue here"},
+        {"role": "assistant", "content": "next answer"},
+    ]
+    assert AgentLoop._PENDING_USER_TURN_KEY not in session.metadata
+    assert AgentLoop._RUNTIME_CHECKPOINT_KEY not in session.metadata

tests/agent/test_mcp_connection.py

@@ -0,0 +1,44 @@
+"""Tests for MCP connection lifecycle in AgentLoop."""
+
+from __future__ import annotations
+
+from unittest.mock import MagicMock
+
+import pytest
+
+from nanobot.agent.loop import AgentLoop
+from nanobot.bus.queue import MessageBus
+
+
+def _make_loop(tmp_path, *, mcp_servers: dict | None = None) -> AgentLoop:
+    bus = MessageBus()
+    provider = MagicMock()
+    provider.get_default_model.return_value = "test-model"
+    provider.generation.max_tokens = 4096
+    return AgentLoop(
+        bus=bus,
+        provider=provider,
+        workspace=tmp_path,
+        model="test-model",
+        mcp_servers=mcp_servers or {"test": object()},
+    )
+
+
+@pytest.mark.asyncio
+async def test_connect_mcp_retries_when_no_servers_connect(tmp_path, monkeypatch: pytest.MonkeyPatch):
+    loop = _make_loop(tmp_path)
+    attempts = 0
+
+    async def _fake_connect(_servers, _registry):
+        nonlocal attempts
+        attempts += 1
+        return {}
+
+    monkeypatch.setattr("nanobot.agent.tools.mcp.connect_mcp_servers", _fake_connect)
+
+    await loop._connect_mcp()
+    await loop._connect_mcp()
+
+    assert attempts == 2
+    assert loop._mcp_connected is False
+    assert loop._mcp_stacks == {}

tests/agent/test_memory_store.py

@@ -79,6 +79,29 @@ class TestHistoryWithCursor:
         entries = store.read_unprocessed_history(since_cursor=0)
         assert len(entries) == 2
 
+    def test_read_unprocessed_skips_entries_without_cursor(self, store):
+        """Regression: entries missing the cursor key should be silently skipped."""
+        store.history_file.write_text(
+            '{"timestamp": "2026-04-01 10:00", "content": "no cursor"}\n'
+            '{"cursor": 2, "timestamp": "2026-04-01 10:01", "content": "valid"}\n'
+            '{"cursor": 3, "timestamp": "2026-04-01 10:02", "content": "also valid"}\n',
+            encoding="utf-8",
+        )
+        entries = store.read_unprocessed_history(since_cursor=0)
+        assert [e["cursor"] for e in entries] == [2, 3]
+
+    def test_next_cursor_falls_back_when_last_entry_has_no_cursor(self, store):
+        """Regression: _next_cursor should not KeyError on entries without cursor."""
+        store.history_file.write_text(
+            '{"timestamp": "2026-04-01 10:01", "content": "no cursor"}\n',
+            encoding="utf-8",
+        )
+        # Delete .cursor file so _next_cursor falls back to reading JSONL
+        store._cursor_file.unlink(missing_ok=True)
+        # Last entry has no cursor — should safely return 1, not KeyError
+        cursor = store.append_history("new event")
+        assert cursor == 1
+
     def test_compact_history_drops_oldest(self, tmp_path):
         store = MemoryStore(tmp_path, max_history_entries=2)
         store.append_history("event 1")

tests/agent/test_skills_loader.py

@@ -250,3 +250,150 @@ def test_list_skills_openclaw_metadata_parsed_for_requirements(
     assert entries == [
         {"name": "openclaw_skill", "path": str(skill_path), "source": "workspace"},
     ]
+
+
+def test_disabled_skills_excluded_from_list(tmp_path: Path) -> None:
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    _write_skill(ws_skills, "alpha", body="# Alpha")
+    beta_path = _write_skill(ws_skills, "beta", body="# Beta")
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin, disabled_skills={"alpha"})
+    entries = loader.list_skills(filter_unavailable=False)
+    assert len(entries) == 1
+    assert entries[0]["name"] == "beta"
+    assert entries[0]["path"] == str(beta_path)
+
+
+def test_disabled_skills_empty_set_no_effect(tmp_path: Path) -> None:
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    _write_skill(ws_skills, "alpha", body="# Alpha")
+    _write_skill(ws_skills, "beta", body="# Beta")
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin, disabled_skills=set())
+    entries = loader.list_skills(filter_unavailable=False)
+    assert len(entries) == 2
+
+
+def test_disabled_skills_excluded_from_build_skills_summary(tmp_path: Path) -> None:
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    _write_skill(ws_skills, "alpha", body="# Alpha")
+    _write_skill(ws_skills, "beta", body="# Beta")
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin, disabled_skills={"alpha"})
+    summary = loader.build_skills_summary()
+    assert "alpha" not in summary
+    assert "beta" in summary
+
+
+def test_disabled_skills_excluded_from_get_always_skills(tmp_path: Path) -> None:
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    _write_skill(ws_skills, "alpha", metadata_json={"always": True}, body="# Alpha")
+    _write_skill(ws_skills, "beta", metadata_json={"always": True}, body="# Beta")
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin, disabled_skills={"alpha"})
+    always = loader.get_always_skills()
+    assert "alpha" not in always
+    assert "beta" in always
+
+
+# -- multiline description tests (YAML folded > and literal |) -----------------
+
+
+def test_build_skills_summary_folded_description(tmp_path: Path) -> None:
+    """description: > (YAML folded scalar) should be parsed correctly."""
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    skill_dir = ws_skills / "pdf"
+    skill_dir.mkdir(parents=True)
+    skill_path = skill_dir / "SKILL.md"
+    skill_path.write_text(
+        "---\n"
+        "name: pdf\n"
+        "description: >\n"
+        "  Use this skill when visual quality and design identity matter for a PDF.\n"
+        "  CREATE (generate from scratch): \"make a PDF\".\n"
+        "---\n\n# PDF Skill\n",
+        encoding="utf-8",
+    )
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin)
+    summary = loader.build_skills_summary()
+    assert "pdf" in summary
+    assert "visual quality" in summary
+
+
+def test_build_skills_summary_literal_description(tmp_path: Path) -> None:
+    """description: | (YAML literal scalar) should be parsed correctly."""
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    skill_dir = ws_skills / "multi"
+    skill_dir.mkdir(parents=True)
+    skill_path = skill_dir / "SKILL.md"
+    skill_path.write_text(
+        "---\n"
+        "name: multi\n"
+        "description: |\n"
+        "  Line one of description.\n"
+        "  Line two of description.\n"
+        "---\n\n# Multi\n",
+        encoding="utf-8",
+    )
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin)
+    meta = loader.get_skill_metadata("multi")
+    assert meta is not None
+    desc = meta.get("description")
+    assert isinstance(desc, str)
+    assert "Line one" in desc
+    assert "Line two" in desc
+
+
+def test_get_skill_metadata_handles_yaml_types(tmp_path: Path) -> None:
+    """yaml.safe_load returns native types; always should be True, not 'true'."""
+    workspace = tmp_path / "ws"
+    ws_skills = workspace / "skills"
+    ws_skills.mkdir(parents=True)
+    skill_dir = ws_skills / "typed"
+    skill_dir.mkdir(parents=True)
+    payload = json.dumps({"nanobot": {"requires": {"bins": ["gh"]}, "always": True}}, separators=(",", ":"))
+    skill_path = skill_dir / "SKILL.md"
+    skill_path.write_text(
+        "---\n"
+        "name: typed\n"
+        f"metadata: {payload}\n"
+        "always: true\n"
+        "---\n\n# Typed\n",
+        encoding="utf-8",
+    )
+    builtin = tmp_path / "builtin"
+    builtin.mkdir()
+
+    loader = SkillsLoader(workspace, builtin_skills_dir=builtin)
+    meta = loader.get_skill_metadata("typed")
+    assert meta is not None
+    # YAML parsed 'true' to Python True
+    assert meta.get("always") is True
+    # metadata is a parsed dict, not a JSON string
+    assert isinstance(meta.get("metadata"), dict)

tests/agent/test_task_cancel.py

@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import asyncio
+import time
 from types import SimpleNamespace
 from unittest.mock import AsyncMock, MagicMock, patch
 
@@ -269,7 +270,9 @@ class TestSubagentCancellation:
 
         monkeypatch.setattr("nanobot.agent.tools.filesystem.ListDirTool.execute", fake_execute)
 
-        await mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"})
+        from nanobot.agent.subagent import SubagentStatus
+        status = SubagentStatus(task_id="sub-1", label="label", task_description="do task", started_at=time.monotonic())
+        await mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"}, status)
 
         assistant_messages = [
             msg for msg in captured_second_call
@@ -308,7 +311,9 @@ class TestSubagentCancellation:
 
         mgr.runner.run = AsyncMock(side_effect=fake_run)
 
-        await mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"})
+        from nanobot.agent.subagent import SubagentStatus
+        status = SubagentStatus(task_id="sub-1", label="label", task_description="do task", started_at=time.monotonic())
+        await mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"}, status)
 
         mgr.runner.run.assert_awaited_once()
         mgr._announce_result.assert_awaited_once()
@@ -344,7 +349,9 @@ class TestSubagentCancellation:
 
         monkeypatch.setattr("nanobot.agent.tools.filesystem.ListDirTool.execute", fake_execute)
 
-        await mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"})
+        from nanobot.agent.subagent import SubagentStatus
+        status = SubagentStatus(task_id="sub-1", label="label", task_description="do task", started_at=time.monotonic())
+        await mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"}, status)
 
         mgr._announce_result.assert_awaited_once()
         args = mgr._announce_result.await_args.args
@@ -356,7 +363,7 @@ class TestSubagentCancellation:
 
     @pytest.mark.asyncio
     async def test_cancel_by_session_cancels_running_subagent_tool(self, monkeypatch, tmp_path):
-        from nanobot.agent.subagent import SubagentManager
+        from nanobot.agent.subagent import SubagentManager, SubagentStatus
         from nanobot.bus.queue import MessageBus
         from nanobot.providers.base import LLMResponse, ToolCallRequest
 
@@ -389,7 +396,10 @@ class TestSubagentCancellation:
         monkeypatch.setattr("nanobot.agent.tools.filesystem.ListDirTool.execute", fake_execute)
 
         task = asyncio.create_task(
-            mgr._run_subagent("sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"})
+            mgr._run_subagent(
+                "sub-1", "do task", "label", {"channel": "test", "chat_id": "c1"},
+                SubagentStatus(task_id="sub-1", label="label", task_description="do task", started_at=time.monotonic()),
+            )
         )
         mgr._running_tasks["sub-1"] = task
         mgr._session_tasks["test:c1"] = {"sub-1"}

tests/agent/test_tool_hint.py

@@ -72,6 +72,22 @@ class TestToolHintKnownTools:
         result = _hint([_tc("exec", {"command": cmd})])
         assert "\u2026/" in result
 
+    def test_exec_abbreviates_quoted_linux_paths_with_spaces(self):
+        """Quoted Unix paths with spaces should still be folded."""
+        cmd = 'cd "/home/user/My Documents/project" && pytest tests/'
+        result = _hint([_tc("exec", {"command": cmd})])
+        assert "\u2026/" in result
+        assert '"/home/user/My Documents/project"' not in result
+        assert '"' in result
+
+    def test_exec_abbreviates_quoted_windows_paths_with_spaces(self):
+        """Quoted Windows paths with spaces should still be folded."""
+        cmd = 'cd "C:/Program Files/Git/project" && git status'
+        result = _hint([_tc("exec", {"command": cmd})])
+        assert "\u2026/" in result
+        assert '"C:/Program Files/Git/project"' not in result
+        assert '"' in result
+
     def test_exec_short_command_unchanged(self):
         result = _hint([_tc("exec", {"command": "npm install typescript"})])
         assert result == "$ npm install typescript"

tests/agent/test_unified_session.py

@@ -0,0 +1,502 @@
+"""Tests for unified_session feature.
+
+Covers:
+- AgentLoop._dispatch() rewrites session_key to "unified:default" when enabled
+- Existing session_key_override is respected (not overwritten)
+- Feature is off by default (no behavior change for existing users)
+- Config schema serialises unified_session as camelCase "unifiedSession"
+- onboard-generated config.json contains "unifiedSession" key
+- /new command correctly clears the shared session in unified mode
+- /new is NOT a priority command (goes through _dispatch, key rewrite applies)
+- Context window consolidation is unaffected by unified_session
+"""
+
+import asyncio
+import json
+from pathlib import Path
+from types import SimpleNamespace
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from nanobot.agent.loop import AgentLoop
+from nanobot.bus.events import InboundMessage
+from nanobot.bus.queue import MessageBus
+from nanobot.command.builtin import cmd_new, register_builtin_commands
+from nanobot.command.router import CommandContext, CommandRouter
+from nanobot.config.schema import AgentDefaults, Config
+from nanobot.session.manager import Session, SessionManager
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _make_loop(tmp_path: Path, unified_session: bool = False) -> AgentLoop:
+    """Create a minimal AgentLoop for dispatch-level tests."""
+    bus = MessageBus()
+    provider = MagicMock()
+    provider.get_default_model.return_value = "test-model"
+
+    with patch("nanobot.agent.loop.SessionManager"), \
+         patch("nanobot.agent.loop.SubagentManager") as MockSubMgr, \
+         patch("nanobot.agent.loop.Dream"):
+        MockSubMgr.return_value.cancel_by_session = AsyncMock(return_value=0)
+        loop = AgentLoop(
+            bus=bus,
+            provider=provider,
+            workspace=tmp_path,
+            unified_session=unified_session,
+        )
+    return loop
+
+
+def _make_msg(channel: str = "telegram", chat_id: str = "111",
+              session_key_override: str | None = None) -> InboundMessage:
+    return InboundMessage(
+        channel=channel,
+        chat_id=chat_id,
+        sender_id="user1",
+        content="hello",
+        session_key_override=session_key_override,
+    )
+
+
+# ---------------------------------------------------------------------------
+# TestUnifiedSessionDispatch — core behaviour
+# ---------------------------------------------------------------------------
+
+class TestUnifiedSessionDispatch:
+    """AgentLoop._dispatch() session key rewriting logic."""
+
+    @pytest.mark.asyncio
+    async def test_unified_session_rewrites_key_to_unified_default(self, tmp_path: Path):
+        """When unified_session=True, all messages use 'unified:default' as session key."""
+        loop = _make_loop(tmp_path, unified_session=True)
+
+        captured: list[str] = []
+
+        async def fake_process(msg, **kwargs):
+            captured.append(msg.session_key)
+            return None
+
+        loop._process_message = fake_process  # type: ignore[method-assign]
+
+        msg = _make_msg(channel="telegram", chat_id="111")
+        await loop._dispatch(msg)
+
+        assert captured == ["unified:default"]
+
+    @pytest.mark.asyncio
+    async def test_unified_session_different_channels_share_same_key(self, tmp_path: Path):
+        """Messages from different channels all resolve to the same session key."""
+        loop = _make_loop(tmp_path, unified_session=True)
+
+        captured: list[str] = []
+
+        async def fake_process(msg, **kwargs):
+            captured.append(msg.session_key)
+            return None
+
+        loop._process_message = fake_process  # type: ignore[method-assign]
+
+        await loop._dispatch(_make_msg(channel="telegram", chat_id="111"))
+        await loop._dispatch(_make_msg(channel="discord", chat_id="222"))
+        await loop._dispatch(_make_msg(channel="cli", chat_id="direct"))
+
+        assert captured == ["unified:default", "unified:default", "unified:default"]
+
+    @pytest.mark.asyncio
+    async def test_unified_session_disabled_preserves_original_key(self, tmp_path: Path):
+        """When unified_session=False (default), session key is channel:chat_id as usual."""
+        loop = _make_loop(tmp_path, unified_session=False)
+
+        captured: list[str] = []
+
+        async def fake_process(msg, **kwargs):
+            captured.append(msg.session_key)
+            return None
+
+        loop._process_message = fake_process  # type: ignore[method-assign]
+
+        msg = _make_msg(channel="telegram", chat_id="999")
+        await loop._dispatch(msg)
+
+        assert captured == ["telegram:999"]
+
+    @pytest.mark.asyncio
+    async def test_unified_session_respects_existing_override(self, tmp_path: Path):
+        """If session_key_override is already set (e.g. Telegram thread), it is NOT overwritten."""
+        loop = _make_loop(tmp_path, unified_session=True)
+
+        captured: list[str] = []
+
+        async def fake_process(msg, **kwargs):
+            captured.append(msg.session_key)
+            return None
+
+        loop._process_message = fake_process  # type: ignore[method-assign]
+
+        msg = _make_msg(channel="telegram", chat_id="111", session_key_override="telegram:thread:42")
+        await loop._dispatch(msg)
+
+        assert captured == ["telegram:thread:42"]
+
+    def test_unified_session_default_is_false(self, tmp_path: Path):
+        """unified_session defaults to False — no behavior change for existing users."""
+        loop = _make_loop(tmp_path)
+        assert loop._unified_session is False
+
+
+# ---------------------------------------------------------------------------
+# TestUnifiedSessionConfig — schema & serialisation
+# ---------------------------------------------------------------------------
+
+class TestUnifiedSessionConfig:
+    """Config schema and onboard serialisation for unified_session."""
+
+    def test_agent_defaults_unified_session_default_is_false(self):
+        """AgentDefaults.unified_session defaults to False."""
+        defaults = AgentDefaults()
+        assert defaults.unified_session is False
+
+    def test_agent_defaults_unified_session_can_be_enabled(self):
+        """AgentDefaults.unified_session can be set to True."""
+        defaults = AgentDefaults(unified_session=True)
+        assert defaults.unified_session is True
+
+    def test_config_serialises_unified_session_as_camel_case(self):
+        """model_dump(by_alias=True) outputs 'unifiedSession' (camelCase) for JSON."""
+        config = Config()
+        data = config.model_dump(mode="json", by_alias=True)
+        agents_defaults = data["agents"]["defaults"]
+        assert "unifiedSession" in agents_defaults
+        assert agents_defaults["unifiedSession"] is False
+
+    def test_config_parses_unified_session_from_camel_case(self):
+        """Config can be loaded from JSON with camelCase 'unifiedSession'."""
+        raw = {"agents": {"defaults": {"unifiedSession": True}}}
+        config = Config.model_validate(raw)
+        assert config.agents.defaults.unified_session is True
+
+    def test_config_parses_unified_session_from_snake_case(self):
+        """Config also accepts snake_case 'unified_session' (populate_by_name=True)."""
+        raw = {"agents": {"defaults": {"unified_session": True}}}
+        config = Config.model_validate(raw)
+        assert config.agents.defaults.unified_session is True
+
+    def test_onboard_generated_config_contains_unified_session(self, tmp_path: Path):
+        """save_config() writes 'unifiedSession' into config.json (simulates nanobot onboard)."""
+        from nanobot.config.loader import save_config
+
+        config = Config()
+        config_path = tmp_path / "config.json"
+        save_config(config, config_path)
+
+        with open(config_path, encoding="utf-8") as f:
+            data = json.load(f)
+
+        agents_defaults = data["agents"]["defaults"]
+        assert "unifiedSession" in agents_defaults, (
+            "onboard-generated config.json must contain 'unifiedSession' key"
+        )
+        assert agents_defaults["unifiedSession"] is False
+
+
+# ---------------------------------------------------------------------------
+# TestCmdNewUnifiedSession — /new command behaviour in unified mode
+# ---------------------------------------------------------------------------
+
+class TestCmdNewUnifiedSession:
+    """/new command routing and session-clear behaviour in unified mode."""
+
+    def test_new_is_not_a_priority_command(self):
+        """/new must NOT be in the priority table — it must go through _dispatch()
+        so the unified session key rewrite applies before cmd_new runs."""
+        router = CommandRouter()
+        register_builtin_commands(router)
+        assert router.is_priority("/new") is False
+
+    def test_new_is_an_exact_command(self):
+        """/new must be registered as an exact command."""
+        router = CommandRouter()
+        register_builtin_commands(router)
+        assert "/new" in router._exact
+
+    @pytest.mark.asyncio
+    async def test_cmd_new_clears_unified_session(self, tmp_path: Path):
+        """cmd_new called with key='unified:default' clears the shared session."""
+        sessions = SessionManager(tmp_path)
+
+        # Pre-populate the shared session with some messages
+        shared = sessions.get_or_create("unified:default")
+        shared.add_message("user", "hello from telegram")
+        shared.add_message("assistant", "hi there")
+        sessions.save(shared)
+        assert len(sessions.get_or_create("unified:default").messages) == 2
+
+        # _schedule_background is a *sync* method that schedules a coroutine via
+        # asyncio.create_task().  Mirror that exactly so the coroutine is consumed
+        # and no RuntimeWarning is emitted.
+        loop = SimpleNamespace(
+            sessions=sessions,
+            consolidator=SimpleNamespace(archive=AsyncMock(return_value=True)),
+        )
+        loop._schedule_background = lambda coro: asyncio.ensure_future(coro)
+
+        msg = InboundMessage(
+            channel="telegram", sender_id="user1", chat_id="111", content="/new",
+            session_key_override="unified:default",  # as _dispatch() would set it
+        )
+        ctx = CommandContext(msg=msg, session=None, key="unified:default", raw="/new", loop=loop)
+
+        result = await cmd_new(ctx)
+
+        assert "New session started" in result.content
+        # Invalidate cache and reload from disk to confirm persistence
+        sessions.invalidate("unified:default")
+        reloaded = sessions.get_or_create("unified:default")
+        assert reloaded.messages == []
+
+    @pytest.mark.asyncio
+    async def test_cmd_new_in_unified_mode_does_not_affect_other_sessions(self, tmp_path: Path):
+        """Clearing unified:default must not touch other sessions on disk."""
+        sessions = SessionManager(tmp_path)
+
+        other = sessions.get_or_create("discord:999")
+        other.add_message("user", "discord message")
+        sessions.save(other)
+
+        shared = sessions.get_or_create("unified:default")
+        shared.add_message("user", "shared message")
+        sessions.save(shared)
+
+        loop = SimpleNamespace(
+            sessions=sessions,
+            consolidator=SimpleNamespace(archive=AsyncMock(return_value=True)),
+        )
+        loop._schedule_background = lambda coro: asyncio.ensure_future(coro)
+
+        msg = InboundMessage(
+            channel="telegram", sender_id="user1", chat_id="111", content="/new",
+            session_key_override="unified:default",
+        )
+        ctx = CommandContext(msg=msg, session=None, key="unified:default", raw="/new", loop=loop)
+        await cmd_new(ctx)
+
+        sessions.invalidate("unified:default")
+        sessions.invalidate("discord:999")
+        assert sessions.get_or_create("unified:default").messages == []
+        assert len(sessions.get_or_create("discord:999").messages) == 1
+
+
+# ---------------------------------------------------------------------------
+# TestConsolidationUnaffectedByUnifiedSession — consolidation is key-agnostic
+# ---------------------------------------------------------------------------
+
+class TestConsolidationUnaffectedByUnifiedSession:
+    """maybe_consolidate_by_tokens() behaviour is identical regardless of session key."""
+
+    @pytest.mark.asyncio
+    async def test_consolidation_skips_empty_session_for_unified_key(self):
+        """Empty unified:default session → consolidation exits immediately, archive not called."""
+        from nanobot.agent.memory import Consolidator, MemoryStore
+
+        store = MagicMock(spec=MemoryStore)
+        mock_provider = MagicMock()
+        mock_provider.chat_with_retry = AsyncMock(return_value=MagicMock(content="summary"))
+        # Use spec= so MagicMock doesn't auto-generate AsyncMock for non-async methods,
+        # which would leave unawaited coroutines and trigger RuntimeWarning.
+        sessions = MagicMock(spec=SessionManager)
+
+        consolidator = Consolidator(
+            store=store,
+            provider=mock_provider,
+            model="test-model",
+            sessions=sessions,
+            context_window_tokens=1000,
+            build_messages=MagicMock(return_value=[]),
+            get_tool_definitions=MagicMock(return_value=[]),
+            max_completion_tokens=100,
+        )
+        consolidator.archive = AsyncMock()
+
+        session = Session(key="unified:default")
+        session.messages = []
+
+        await consolidator.maybe_consolidate_by_tokens(session)
+
+        consolidator.archive.assert_not_called()
+
+    @pytest.mark.asyncio
+    async def test_consolidation_behaviour_identical_for_any_key(self):
+        """archive call count is the same for 'telegram:123' and 'unified:default'
+        under identical token conditions."""
+        from nanobot.agent.memory import Consolidator, MemoryStore
+
+        archive_calls: dict[str, int] = {}
+
+        for key in ("telegram:123", "unified:default"):
+            store = MagicMock(spec=MemoryStore)
+            mock_provider = MagicMock()
+            mock_provider.chat_with_retry = AsyncMock(return_value=MagicMock(content="summary"))
+            sessions = MagicMock(spec=SessionManager)
+
+            consolidator = Consolidator(
+                store=store,
+                provider=mock_provider,
+                model="test-model",
+                sessions=sessions,
+                context_window_tokens=1000,
+                build_messages=MagicMock(return_value=[]),
+                get_tool_definitions=MagicMock(return_value=[]),
+                max_completion_tokens=100,
+            )
+
+            session = Session(key=key)
+            session.messages = []  # empty → exits immediately for both keys
+
+            consolidator.archive = AsyncMock()
+            await consolidator.maybe_consolidate_by_tokens(session)
+            archive_calls[key] = consolidator.archive.call_count
+
+        assert archive_calls["telegram:123"] == archive_calls["unified:default"] == 0
+
+    @pytest.mark.asyncio
+    async def test_consolidation_triggers_when_over_budget_unified_key(self):
+        """When tokens exceed budget, consolidation attempts to find a boundary —
+        behaviour is identical to any other session key."""
+        from nanobot.agent.memory import Consolidator, MemoryStore
+
+        store = MagicMock(spec=MemoryStore)
+        mock_provider = MagicMock()
+        sessions = MagicMock(spec=SessionManager)
+
+        consolidator = Consolidator(
+            store=store,
+            provider=mock_provider,
+            model="test-model",
+            sessions=sessions,
+            context_window_tokens=1000,
+            build_messages=MagicMock(return_value=[]),
+            get_tool_definitions=MagicMock(return_value=[]),
+            max_completion_tokens=100,
+        )
+
+        session = Session(key="unified:default")
+        session.messages = [{"role": "user", "content": "msg"}]
+
+        # Simulate over-budget: estimated > budget
+        consolidator.estimate_session_prompt_tokens = MagicMock(return_value=(950, "tiktoken"))
+        # No valid boundary found → returns gracefully without archiving
+        consolidator.pick_consolidation_boundary = MagicMock(return_value=None)
+        consolidator.archive = AsyncMock()
+
+        await consolidator.maybe_consolidate_by_tokens(session)
+
+        # estimate was called (consolidation was attempted)
+        consolidator.estimate_session_prompt_tokens.assert_called_once_with(session)
+        # but archive was not called (no valid boundary)
+        consolidator.archive.assert_not_called()
+
+
+# ---------------------------------------------------------------------------
+# TestStopCommandWithUnifiedSession — /stop command integration
+# ---------------------------------------------------------------------------
+
+
+class TestStopCommandWithUnifiedSession:
+    """Verify /stop command works correctly with unified session enabled."""
+
+    @pytest.mark.asyncio
+    async def test_active_tasks_use_effective_key_in_unified_mode(self, tmp_path: Path):
+        """When unified_session=True, tasks are stored under UNIFIED_SESSION_KEY."""
+        from nanobot.agent.loop import UNIFIED_SESSION_KEY
+
+        loop = _make_loop(tmp_path, unified_session=True)
+        
+        # Create a message from telegram channel
+        msg = _make_msg(channel="telegram", chat_id="123456")
+
+        # Mock _dispatch to complete immediately
+        async def fake_dispatch(m):
+            pass
+
+        loop._dispatch = fake_dispatch  # type: ignore[method-assign]
+
+        # Simulate the task creation flow (from _run loop)
+        effective_key = UNIFIED_SESSION_KEY if loop._unified_session and not msg.session_key_override else msg.session_key
+        task = asyncio.create_task(loop._dispatch(msg))
+        loop._active_tasks.setdefault(effective_key, []).append(task)
+
+        # Wait for task to complete
+        await task
+
+        # Verify the task is stored under UNIFIED_SESSION_KEY, not the original channel:chat_id
+        assert UNIFIED_SESSION_KEY in loop._active_tasks
+        assert "telegram:123456" not in loop._active_tasks
+
+    @pytest.mark.asyncio
+    async def test_stop_command_finds_task_in_unified_mode(self, tmp_path: Path):
+        """cmd_stop can cancel tasks when unified_session=True."""
+        from nanobot.agent.loop import UNIFIED_SESSION_KEY
+        from nanobot.command.builtin import cmd_stop
+
+        loop = _make_loop(tmp_path, unified_session=True)
+
+        # Create a long-running task stored under UNIFIED_SESSION_KEY
+        async def long_running():
+            await asyncio.sleep(10)  # Will be cancelled
+
+        task = asyncio.create_task(long_running())
+        loop._active_tasks[UNIFIED_SESSION_KEY] = [task]
+
+        # Create a message that would have session_key=UNIFIED_SESSION_KEY after dispatch
+        msg = InboundMessage(
+            channel="telegram",
+            chat_id="123456",
+            sender_id="user1",
+            content="/stop",
+            session_key_override=UNIFIED_SESSION_KEY,  # Simulate post-dispatch state
+        )
+
+        ctx = CommandContext(msg=msg, session=None, key=UNIFIED_SESSION_KEY, raw="/stop", loop=loop)
+
+        # Execute /stop
+        result = await cmd_stop(ctx)
+
+        # Verify task was cancelled
+        assert task.cancelled() or task.done()
+        assert "Stopped 1 task" in result.content
+
+    @pytest.mark.asyncio
+    async def test_stop_command_cross_channel_in_unified_mode(self, tmp_path: Path):
+        """In unified mode, /stop from one channel cancels tasks from another channel."""
+        from nanobot.agent.loop import UNIFIED_SESSION_KEY
+        from nanobot.command.builtin import cmd_stop
+
+        loop = _make_loop(tmp_path, unified_session=True)
+
+        # Create tasks from different channels, all stored under UNIFIED_SESSION_KEY
+        async def long_running():
+            await asyncio.sleep(10)
+
+        task1 = asyncio.create_task(long_running())
+        task2 = asyncio.create_task(long_running())
+        loop._active_tasks[UNIFIED_SESSION_KEY] = [task1, task2]
+
+        # /stop from discord should cancel tasks started from telegram
+        msg = InboundMessage(
+            channel="discord",
+            chat_id="789012",
+            sender_id="user2",
+            content="/stop",
+            session_key_override=UNIFIED_SESSION_KEY,
+        )
+
+        ctx = CommandContext(msg=msg, session=None, key=UNIFIED_SESSION_KEY, raw="/stop", loop=loop)
+
+        result = await cmd_stop(ctx)
+
+        # Both tasks should be cancelled
+        assert "Stopped 2 task" in result.content

tests/channels/test_base_channel.py

@@ -23,3 +23,15 @@ def test_is_allowed_requires_exact_match() -> None:
 
     assert channel.is_allowed("allow@email.com") is True
     assert channel.is_allowed("attacker|allow@email.com") is False
+
+
+def test_is_allowed_supports_dict_allow_from_alias() -> None:
+    channel = _DummyChannel({"allowFrom": ["alice"]}, MessageBus())
+
+    assert channel.is_allowed("alice") is True
+
+
+def test_is_allowed_denies_empty_dict_allow_from() -> None:
+    channel = _DummyChannel({"allow_from": []}, MessageBus())
+
+    assert channel.is_allowed("alice") is False

tests/channels/test_channel_plugins.py

@@ -175,7 +175,7 @@ async def test_manager_loads_plugin_from_dict_config():
         channels=ChannelsConfig.model_validate({
             "fakeplugin": {"enabled": True, "allowFrom": ["*"]},
         }),
-        providers=SimpleNamespace(groq=SimpleNamespace(api_key="")),
+        providers=SimpleNamespace(groq=SimpleNamespace(api_key="", api_base="")),
     )
 
     with patch(
@@ -193,6 +193,113 @@ async def test_manager_loads_plugin_from_dict_config():
     assert isinstance(mgr.channels["fakeplugin"], _FakePlugin)
 
 
+@pytest.mark.asyncio
+async def test_manager_propagates_groq_transcription_api_base_to_channels():
+    from nanobot.channels.manager import ChannelManager
+
+    fake_config = SimpleNamespace(
+        channels=ChannelsConfig.model_validate({
+            "fakeplugin": {"enabled": True, "allowFrom": ["*"]},
+        }),
+        transcription_provider="groq",
+        providers=SimpleNamespace(
+            groq=SimpleNamespace(api_key="groq-key", api_base="http://proxy.local/v1/audio/transcriptions"),
+            openai=SimpleNamespace(api_key="openai-key", api_base="https://api.openai.com/v1/audio/transcriptions"),
+        ),
+    )
+
+    with patch(
+        "nanobot.channels.registry.discover_all",
+        return_value={"fakeplugin": _FakePlugin},
+    ):
+        mgr = ChannelManager.__new__(ChannelManager)
+        mgr.config = fake_config
+        mgr.bus = MessageBus()
+        mgr.channels = {}
+        mgr._dispatch_task = None
+        mgr._init_channels()
+
+    channel = mgr.channels["fakeplugin"]
+    assert channel.transcription_provider == "groq"
+    assert channel.transcription_api_key == "groq-key"
+    assert channel.transcription_api_base == "http://proxy.local/v1/audio/transcriptions"
+
+
+@pytest.mark.asyncio
+async def test_manager_propagates_openai_transcription_api_base_to_channels():
+    from nanobot.channels.manager import ChannelManager
+
+    fake_config = SimpleNamespace(
+        channels=ChannelsConfig.model_validate({
+            "fakeplugin": {"enabled": True, "allowFrom": ["*"]},
+            "transcriptionProvider": "openai",
+        }),
+        providers=SimpleNamespace(
+            openai=SimpleNamespace(
+                api_key="openai-key",
+                api_base="http://proxy.local/v1/audio/transcriptions",
+            ),
+            groq=SimpleNamespace(api_key="groq-key", api_base=""),
+        ),
+    )
+
+    with patch(
+        "nanobot.channels.registry.discover_all",
+        return_value={"fakeplugin": _FakePlugin},
+    ):
+        mgr = ChannelManager.__new__(ChannelManager)
+        mgr.config = fake_config
+        mgr.bus = MessageBus()
+        mgr.channels = {}
+        mgr._dispatch_task = None
+        mgr._init_channels()
+
+    channel = mgr.channels["fakeplugin"]
+    assert channel.transcription_provider == "openai"
+    assert channel.transcription_api_key == "openai-key"
+    assert channel.transcription_api_base == "http://proxy.local/v1/audio/transcriptions"
+
+
+@pytest.mark.asyncio
+async def test_base_channel_passes_api_base_to_openai_transcription_provider():
+    """BaseChannel.transcribe_audio must forward transcription_api_base to OpenAI."""
+    from nanobot.providers import transcription as transcription_mod
+
+    channel = _FakePlugin({"enabled": True, "allowFrom": ["*"]}, MessageBus())
+    channel.transcription_provider = "openai"
+    channel.transcription_api_key = "k"
+    channel.transcription_api_base = "http://override/v1/audio/transcriptions"
+
+    captured: dict[str, object] = {}
+
+    class _StubOpenAI:
+        def __init__(self, api_key=None, api_base=None):
+            captured["api_key"] = api_key
+            captured["api_base"] = api_base
+
+        async def transcribe(self, file_path):
+            return "ok"
+
+    with patch.object(transcription_mod, "OpenAITranscriptionProvider", _StubOpenAI):
+        result = await channel.transcribe_audio("/tmp/does-not-matter.wav")
+
+    assert result == "ok"
+    assert captured["api_key"] == "k"
+    assert captured["api_base"] == "http://override/v1/audio/transcriptions"
+
+
+def test_openai_transcription_provider_honors_api_base_argument():
+    from nanobot.providers.transcription import OpenAITranscriptionProvider
+
+    default = OpenAITranscriptionProvider(api_key="k")
+    assert default.api_url == "https://api.openai.com/v1/audio/transcriptions"
+
+    custom = OpenAITranscriptionProvider(
+        api_key="k", api_base="http://override/v1/audio/transcriptions"
+    )
+    assert custom.api_url == "http://override/v1/audio/transcriptions"
+
+
 def test_channels_login_uses_discovered_plugin_class(monkeypatch):
     from nanobot.cli.commands import app
     from nanobot.config.schema import Config
@@ -646,7 +753,10 @@ class _ChannelWithAllowFrom(BaseChannel):
 
     def __init__(self, config, bus, allow_from):
         super().__init__(config, bus)
-        self.config.allow_from = allow_from
+        if isinstance(self.config, dict):
+            self.config["allow_from"] = allow_from
+        else:
+            self.config.allow_from = allow_from
 
     async def start(self) -> None:
         pass
@@ -714,6 +824,25 @@ async def test_validate_allow_from_passes_with_asterisk():
     mgr._validate_allow_from()
 
 
+@pytest.mark.asyncio
+async def test_validate_allow_from_raises_on_empty_dict_allow_from():
+    """_validate_allow_from should reject empty dict-backed allow_from lists."""
+    fake_config = SimpleNamespace(
+        channels=ChannelsConfig(),
+        providers=SimpleNamespace(groq=SimpleNamespace(api_key="")),
+    )
+
+    mgr = ChannelManager.__new__(ChannelManager)
+    mgr.config = fake_config
+    mgr.channels = {"test": _ChannelWithAllowFrom({"enabled": True}, None, [])}
+    mgr._dispatch_task = None
+
+    with pytest.raises(SystemExit) as exc_info:
+        mgr._validate_allow_from()
+
+    assert "empty allowFrom" in str(exc_info.value)
+
+
 @pytest.mark.asyncio
 async def test_get_channel_returns_channel_if_exists():
     """get_channel should return the channel if it exists."""

tests/channels/test_dingtalk_channel.py

@@ -1,6 +1,10 @@
 import asyncio
+import zipfile
+from io import BytesIO
 from types import SimpleNamespace
+from unittest.mock import AsyncMock
 
+import httpx
 import pytest
 
 # Check optional dingtalk dependencies before running tests
@@ -50,6 +54,21 @@ class _FakeHttp:
         return self._next_response()
 
 
+class _NetworkErrorHttp:
+    """HTTP client stub that raises httpx.TransportError on every request."""
+
+    def __init__(self) -> None:
+        self.calls: list[dict] = []
+
+    async def post(self, url: str, json=None, headers=None, **kwargs):
+        self.calls.append({"method": "POST", "url": url, "json": json, "headers": headers})
+        raise httpx.ConnectError("Connection refused")
+
+    async def get(self, url: str, **kwargs):
+        self.calls.append({"method": "GET", "url": url})
+        raise httpx.ConnectError("Connection refused")
+
+
 @pytest.mark.asyncio
 async def test_group_message_keeps_sender_id_and_routes_chat_id() -> None:
     config = DingTalkConfig(client_id="app", client_secret="secret", allow_from=["user1"])
@@ -221,3 +240,216 @@ async def test_download_dingtalk_file(tmp_path, monkeypatch) -> None:
     assert "messageFiles/download" in channel._http.calls[0]["url"]
     assert channel._http.calls[0]["json"]["downloadCode"] == "code123"
     assert channel._http.calls[1]["method"] == "GET"
+
+
+def test_normalize_upload_payload_zips_html_attachment() -> None:
+    channel = DingTalkChannel(
+        DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"]),
+        MessageBus(),
+    )
+
+    data, filename, content_type = channel._normalize_upload_payload(
+        "report.html",
+        b"<html><body>Hello</body></html>",
+        "text/html",
+    )
+
+    assert filename == "report.zip"
+    assert content_type == "application/zip"
+
+    archive = zipfile.ZipFile(BytesIO(data))
+    assert archive.namelist() == ["report.html"]
+    assert archive.read("report.html") == b"<html><body>Hello</body></html>"
+
+
+@pytest.mark.asyncio
+async def test_send_media_ref_zips_html_before_upload(tmp_path, monkeypatch) -> None:
+    channel = DingTalkChannel(
+        DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"]),
+        MessageBus(),
+    )
+
+    html_path = tmp_path / "report.html"
+    html_path.write_text("<html><body>Hello</body></html>", encoding="utf-8")
+
+    captured: dict[str, object] = {}
+
+    async def fake_upload_media(*, token, data, media_type, filename, content_type):
+        captured.update(
+            {
+                "token": token,
+                "data": data,
+                "media_type": media_type,
+                "filename": filename,
+                "content_type": content_type,
+            }
+        )
+        return "media-123"
+
+    async def fake_send_batch_message(token, chat_id, msg_key, msg_param):
+        captured.update(
+            {
+                "sent_token": token,
+                "chat_id": chat_id,
+                "msg_key": msg_key,
+                "msg_param": msg_param,
+            }
+        )
+        return True
+
+    monkeypatch.setattr(channel, "_upload_media", fake_upload_media)
+    monkeypatch.setattr(channel, "_send_batch_message", fake_send_batch_message)
+
+    ok = await channel._send_media_ref("token-123", "user-1", str(html_path))
+
+    assert ok is True
+    assert captured["media_type"] == "file"
+    assert captured["filename"] == "report.zip"
+    assert captured["content_type"] == "application/zip"
+    assert captured["msg_key"] == "sampleFile"
+    assert captured["msg_param"] == {
+        "mediaId": "media-123",
+        "fileName": "report.zip",
+        "fileType": "zip",
+    }
+
+    archive = zipfile.ZipFile(BytesIO(captured["data"]))
+    assert archive.namelist() == ["report.html"]
+
+
+# ── Exception handling tests ──────────────────────────────────────────
+
+
+@pytest.mark.asyncio
+async def test_send_batch_message_propagates_transport_error() -> None:
+    """Network/transport errors must re-raise so callers can retry."""
+    config = DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"])
+    channel = DingTalkChannel(config, MessageBus())
+    channel._http = _NetworkErrorHttp()
+
+    with pytest.raises(httpx.ConnectError, match="Connection refused"):
+        await channel._send_batch_message(
+            "token",
+            "user123",
+            "sampleMarkdown",
+            {"text": "hello", "title": "Nanobot Reply"},
+        )
+
+    # The POST was attempted exactly once
+    assert len(channel._http.calls) == 1
+    assert channel._http.calls[0]["method"] == "POST"
+
+
+@pytest.mark.asyncio
+async def test_send_batch_message_returns_false_on_api_error() -> None:
+    """DingTalk API-level errors (non-200 status, errcode != 0) should return False."""
+    config = DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"])
+    channel = DingTalkChannel(config, MessageBus())
+
+    # Non-200 status code → API error → return False
+    channel._http = _FakeHttp(responses=[_FakeResponse(400, {"errcode": 400})])
+    result = await channel._send_batch_message(
+        "token", "user123", "sampleMarkdown", {"text": "hello"}
+    )
+    assert result is False
+
+    # 200 with non-zero errcode → API error → return False
+    channel._http = _FakeHttp(responses=[_FakeResponse(200, {"errcode": 100})])
+    result = await channel._send_batch_message(
+        "token", "user123", "sampleMarkdown", {"text": "hello"}
+    )
+    assert result is False
+
+    # 200 with errcode=0 → success → return True
+    channel._http = _FakeHttp(responses=[_FakeResponse(200, {"errcode": 0})])
+    result = await channel._send_batch_message(
+        "token", "user123", "sampleMarkdown", {"text": "hello"}
+    )
+    assert result is True
+
+
+@pytest.mark.asyncio
+async def test_send_media_ref_short_circuits_on_transport_error() -> None:
+    """When the first send fails with a transport error, _send_media_ref must
+    re-raise immediately instead of trying download+upload+fallback."""
+    config = DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"])
+    channel = DingTalkChannel(config, MessageBus())
+    channel._http = _NetworkErrorHttp()
+
+    # An image URL triggers the sampleImageMsg path first
+    with pytest.raises(httpx.ConnectError, match="Connection refused"):
+        await channel._send_media_ref("token", "user123", "https://example.com/photo.jpg")
+
+    # Only one POST should have been attempted — no download/upload/fallback
+    assert len(channel._http.calls) == 1
+    assert channel._http.calls[0]["method"] == "POST"
+
+
+@pytest.mark.asyncio
+async def test_send_media_ref_short_circuits_on_download_transport_error() -> None:
+    """When the image URL send returns an API error (False) but the download
+    for the fallback hits a transport error, it must re-raise rather than
+    silently returning False."""
+    config = DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"])
+    channel = DingTalkChannel(config, MessageBus())
+
+    # First POST (sampleImageMsg) returns API error → False, then GET (download) raises transport error
+    class _MixedHttp:
+        def __init__(self) -> None:
+            self.calls: list[dict] = []
+
+        async def post(self, url, json=None, headers=None, **kwargs):
+            self.calls.append({"method": "POST", "url": url})
+            # API-level failure: 200 with errcode != 0
+            return _FakeResponse(200, {"errcode": 100})
+
+        async def get(self, url, **kwargs):
+            self.calls.append({"method": "GET", "url": url})
+            raise httpx.ConnectError("Connection refused")
+
+    channel._http = _MixedHttp()
+
+    with pytest.raises(httpx.ConnectError, match="Connection refused"):
+        await channel._send_media_ref("token", "user123", "https://example.com/photo.jpg")
+
+    # Should have attempted POST (image URL) and GET (download), but NOT upload
+    assert len(channel._http.calls) == 2
+    assert channel._http.calls[0]["method"] == "POST"
+    assert channel._http.calls[1]["method"] == "GET"
+
+
+@pytest.mark.asyncio
+async def test_send_media_ref_short_circuits_on_upload_transport_error() -> None:
+    """When download succeeds but upload hits a transport error, must re-raise."""
+    config = DingTalkConfig(client_id="app", client_secret="secret", allow_from=["*"])
+    channel = DingTalkChannel(config, MessageBus())
+
+    image_bytes = b"\xff\xd8\xff\xe0" + b"\x00" * 100  # minimal JPEG-ish data
+
+    class _UploadFailsHttp:
+        def __init__(self) -> None:
+            self.calls: list[dict] = []
+
+        async def post(self, url, json=None, headers=None, files=None, **kwargs):
+            self.calls.append({"method": "POST", "url": url})
+            # If it's the upload endpoint, raise transport error
+            if "media/upload" in url:
+                raise httpx.ConnectError("Connection refused")
+            # Otherwise (sampleImageMsg), return API error to trigger fallback
+            return _FakeResponse(200, {"errcode": 100})
+
+        async def get(self, url, **kwargs):
+            self.calls.append({"method": "GET", "url": url})
+            resp = _FakeResponse(200)
+            resp.content = image_bytes
+            resp.headers = {"content-type": "image/jpeg"}
+            return resp
+
+    channel._http = _UploadFailsHttp()
+
+    with pytest.raises(httpx.ConnectError, match="Connection refused"):
+        await channel._send_media_ref("token", "user123", "https://example.com/photo.jpg")
+
+    # POST (image URL), GET (download), POST (upload) attempted — no further sends
+    methods = [c["method"] for c in channel._http.calls]
+    assert methods == ["POST", "GET", "POST"]

tests/channels/test_discord_channel.py

@@ -5,11 +5,17 @@ from pathlib import Path
 from types import SimpleNamespace
 
 import pytest
+
 discord = pytest.importorskip("discord")
 
 from nanobot.bus.events import OutboundMessage
 from nanobot.bus.queue import MessageBus
-from nanobot.channels.discord import DiscordBotClient, DiscordChannel, DiscordConfig
+from nanobot.channels.discord import (
+    MAX_MESSAGE_LEN,
+    DiscordBotClient,
+    DiscordChannel,
+    DiscordConfig,
+)
 from nanobot.command.builtin import build_help_text
 
 
@@ -18,9 +24,11 @@ class _FakeDiscordClient:
     instances: list["_FakeDiscordClient"] = []
     start_error: Exception | None = None
 
-    def __init__(self, owner, *, intents) -> None:
+    def __init__(self, owner, *, intents, proxy=None, proxy_auth=None) -> None:
         self.owner = owner
         self.intents = intents
+        self.proxy = proxy
+        self.proxy_auth = proxy_auth
         self.closed = False
         self.ready = True
         self.channels: dict[int, object] = {}
@@ -53,7 +61,9 @@ class _FakeDiscordClient:
 
 class _FakeAttachment:
     # Attachment double that can simulate successful or failing save() calls.
-    def __init__(self, attachment_id: int, filename: str, *, size: int = 1, fail: bool = False) -> None:
+    def __init__(
+        self, attachment_id: int, filename: str, *, size: int = 1, fail: bool = False
+    ) -> None:
         self.id = attachment_id
         self.filename = filename
         self.size = size
@@ -71,11 +81,25 @@ class _FakePartialMessage:
         self.id = message_id
 
 
+class _FakeSentMessage:
+    # Sent-message double supporting edit() for streaming tests.
+    def __init__(self, channel, content: str) -> None:
+        self.channel = channel
+        self.content = content
+        self.edits: list[dict] = []
+
+    async def edit(self, **kwargs) -> None:
+        self.edits.append(dict(kwargs))
+        if "content" in kwargs:
+            self.content = kwargs["content"]
+
+
 class _FakeChannel:
     # Channel double that records outbound payloads and typing activity.
     def __init__(self, channel_id: int = 123) -> None:
         self.id = channel_id
         self.sent_payloads: list[dict] = []
+        self.sent_messages: list[_FakeSentMessage] = []
         self.trigger_typing_calls = 0
         self.typing_enter_hook = None
 
@@ -85,6 +109,9 @@ class _FakeChannel:
             payload["file_name"] = payload["file"].filename
             del payload["file"]
         self.sent_payloads.append(payload)
+        message = _FakeSentMessage(self, payload.get("content", ""))
+        self.sent_messages.append(message)
+        return message
 
     def get_partial_message(self, message_id: int) -> _FakePartialMessage:
         return _FakePartialMessage(message_id)
@@ -194,7 +221,7 @@ async def test_start_handles_client_construction_failure(monkeypatch) -> None:
         MessageBus(),
     )
 
-    def _boom(owner, *, intents):
+    def _boom(owner, *, intents, proxy=None, proxy_auth=None):
         raise RuntimeError("bad client")
 
     monkeypatch.setattr("nanobot.channels.discord.DiscordBotClient", _boom)
@@ -286,6 +313,45 @@ async def test_on_message_accepts_allowlisted_dm() -> None:
     assert handled[0]["metadata"] == {"message_id": "789", "guild_id": None, "reply_to": None}
 
 
+@pytest.mark.asyncio
+async def test_on_message_accepts_when_channel_in_allow_channels() -> None:
+    # When allow_channels is set, messages from listed channels should be forwarded.
+    channel = DiscordChannel(
+        DiscordConfig(enabled=True, allow_from=["*"], allow_channels=["456"]),
+        MessageBus(),
+    )
+    handled: list[dict] = []
+
+    async def capture_handle(**kwargs) -> None:
+        handled.append(kwargs)
+
+    channel._handle_message = capture_handle  # type: ignore[method-assign]
+
+    await channel._on_message(_make_message(author_id=123, channel_id=456))
+
+    assert len(handled) == 1
+    assert handled[0]["chat_id"] == "456"
+
+
+@pytest.mark.asyncio
+async def test_on_message_drops_when_channel_not_in_allow_channels() -> None:
+    # When allow_channels is set and incoming channel is not listed, drop silently.
+    channel = DiscordChannel(
+        DiscordConfig(enabled=True, allow_from=["*"], allow_channels=["999"]),
+        MessageBus(),
+    )
+    handled: list[dict] = []
+
+    async def capture_handle(**kwargs) -> None:
+        handled.append(kwargs)
+
+    channel._handle_message = capture_handle  # type: ignore[method-assign]
+
+    await channel._on_message(_make_message(author_id=123, channel_id=456))
+
+    assert handled == []
+
+
 @pytest.mark.asyncio
 async def test_on_message_ignores_unmentioned_guild_message() -> None:
     # With mention-only group policy, guild messages without a bot mention are dropped.
@@ -427,6 +493,60 @@ async def test_send_fetches_channel_when_not_cached() -> None:
     assert target.sent_payloads == [{"content": "hello"}]
 
 
+def test_supports_streaming_enabled_by_default() -> None:
+    channel = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+
+    assert channel.supports_streaming is True
+
+
+@pytest.mark.asyncio
+async def test_send_delta_streams_by_editing_message(monkeypatch) -> None:
+    owner = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+    client = _FakeDiscordClient(owner, intents=None)
+    owner._client = client
+    owner._running = True
+    target = _FakeChannel(channel_id=123)
+    client.channels[123] = target
+
+    times = iter([1.0, 3.0, 5.0])
+    monkeypatch.setattr("nanobot.channels.discord.time.monotonic", lambda: next(times, 5.0))
+
+    await owner.send_delta("123", "hel", {"_stream_delta": True, "_stream_id": "s1"})
+    await owner.send_delta("123", "lo", {"_stream_delta": True, "_stream_id": "s1"})
+    await owner.send_delta("123", "", {"_stream_end": True, "_stream_id": "s1"})
+
+    assert target.sent_payloads[0] == {"content": "hel"}
+    assert target.sent_messages[0].edits == [{"content": "hello"}, {"content": "hello"}]
+    assert owner._stream_bufs == {}
+
+
+@pytest.mark.asyncio
+async def test_send_delta_stream_end_splits_oversized_reply(monkeypatch) -> None:
+    owner = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+    client = _FakeDiscordClient(owner, intents=None)
+    owner._client = client
+    owner._running = True
+    target = _FakeChannel(channel_id=123)
+    client.channels[123] = target
+
+    prefix = "a" * (MAX_MESSAGE_LEN - 100)
+    suffix = "b" * 150
+    full_text = prefix + suffix
+    chunks = DiscordBotClient._build_chunks(full_text, [], False)
+    assert len(chunks) == 2
+
+    times = iter([1.0, 3.0])
+    monkeypatch.setattr("nanobot.channels.discord.time.monotonic", lambda: next(times, 3.0))
+
+    await owner.send_delta("123", prefix, {"_stream_delta": True, "_stream_id": "s1"})
+    await owner.send_delta("123", suffix, {"_stream_delta": True, "_stream_id": "s1"})
+    await owner.send_delta("123", "", {"_stream_end": True, "_stream_id": "s1"})
+
+    assert target.sent_payloads == [{"content": prefix}, {"content": chunks[1]}]
+    assert target.sent_messages[0].edits == [{"content": chunks[0]}, {"content": chunks[0]}]
+    assert owner._stream_bufs == {}
+
+
 @pytest.mark.asyncio
 async def test_slash_new_forwards_when_user_is_allowlisted() -> None:
     channel = DiscordChannel(DiscordConfig(enabled=True, allow_from=["123"]), MessageBus())
@@ -443,9 +563,7 @@ async def test_slash_new_forwards_when_user_is_allowlisted() -> None:
     assert new_cmd is not None
     await new_cmd.callback(interaction)
 
-    assert interaction.response.messages == [
-        {"content": "Processing /new...", "ephemeral": True}
-    ]
+    assert interaction.response.messages == [{"content": "Processing /new...", "ephemeral": True}]
     assert len(handled) == 1
     assert handled[0]["content"] == "/new"
     assert handled[0]["sender_id"] == "123"
@@ -519,9 +637,7 @@ async def test_slash_help_returns_ephemeral_help_text() -> None:
     assert help_cmd is not None
     await help_cmd.callback(interaction)
 
-    assert interaction.response.messages == [
-        {"content": build_help_text(), "ephemeral": True}
-    ]
+    assert interaction.response.messages == [{"content": build_help_text(), "ephemeral": True}]
     assert handled == []
 
 
@@ -656,11 +772,13 @@ async def test_start_typing_uses_typing_context_when_trigger_typing_missing() ->
         def typing(self):
             async def _waiter():
                 await release.wait()
+
             # Hold the loop so task remains active until explicitly stopped.
             class _Ctx(_TypingCtx):
                 async def __aenter__(self):
                     await super().__aenter__()
                     await _waiter()
+
             return _Ctx()
 
     typing_channel = _NoTriggerChannel(channel_id=123)
@@ -674,3 +792,214 @@ async def test_start_typing_uses_typing_context_when_trigger_typing_missing() ->
     await asyncio.sleep(0)
 
     assert channel._typing_tasks == {}
+
+
+def test_config_accepts_proxy_fields() -> None:
+    config = DiscordConfig(
+        enabled=True,
+        token="token",
+        allow_from=["*"],
+        proxy="http://127.0.0.1:7890",
+        proxy_username="user",
+        proxy_password="pass",
+    )
+    assert config.proxy == "http://127.0.0.1:7890"
+    assert config.proxy_username == "user"
+    assert config.proxy_password == "pass"
+
+
+def test_config_proxy_defaults_to_none() -> None:
+    config = DiscordConfig(enabled=True, token="token", allow_from=["*"])
+    assert config.proxy is None
+    assert config.proxy_username is None
+    assert config.proxy_password is None
+
+
+@pytest.mark.asyncio
+async def test_start_passes_proxy_to_client(monkeypatch) -> None:
+    _FakeDiscordClient.instances.clear()
+    channel = DiscordChannel(
+        DiscordConfig(
+            enabled=True,
+            token="token",
+            allow_from=["*"],
+            proxy="http://127.0.0.1:7890",
+        ),
+        MessageBus(),
+    )
+    monkeypatch.setattr("nanobot.channels.discord.DiscordBotClient", _FakeDiscordClient)
+
+    await channel.start()
+
+    assert channel.is_running is False
+    assert len(_FakeDiscordClient.instances) == 1
+    assert _FakeDiscordClient.instances[0].proxy == "http://127.0.0.1:7890"
+    assert _FakeDiscordClient.instances[0].proxy_auth is None
+
+
+@pytest.mark.asyncio
+async def test_start_passes_proxy_auth_when_credentials_provided(monkeypatch) -> None:
+    aiohttp = pytest.importorskip("aiohttp")
+    _FakeDiscordClient.instances.clear()
+    channel = DiscordChannel(
+        DiscordConfig(
+            enabled=True,
+            token="token",
+            allow_from=["*"],
+            proxy="http://127.0.0.1:7890",
+            proxy_username="user",
+            proxy_password="pass",
+        ),
+        MessageBus(),
+    )
+    monkeypatch.setattr("nanobot.channels.discord.DiscordBotClient", _FakeDiscordClient)
+
+    await channel.start()
+
+    assert channel.is_running is False
+    assert len(_FakeDiscordClient.instances) == 1
+    assert _FakeDiscordClient.instances[0].proxy == "http://127.0.0.1:7890"
+    assert _FakeDiscordClient.instances[0].proxy_auth is not None
+    assert isinstance(_FakeDiscordClient.instances[0].proxy_auth, aiohttp.BasicAuth)
+    assert _FakeDiscordClient.instances[0].proxy_auth.login == "user"
+    assert _FakeDiscordClient.instances[0].proxy_auth.password == "pass"
+
+
+@pytest.mark.asyncio
+async def test_start_no_proxy_auth_when_only_username(monkeypatch) -> None:
+    _FakeDiscordClient.instances.clear()
+    channel = DiscordChannel(
+        DiscordConfig(
+            enabled=True,
+            token="token",
+            allow_from=["*"],
+            proxy="http://127.0.0.1:7890",
+            proxy_username="user",
+        ),
+        MessageBus(),
+    )
+    monkeypatch.setattr("nanobot.channels.discord.DiscordBotClient", _FakeDiscordClient)
+
+    await channel.start()
+
+    assert channel.is_running is False
+    assert _FakeDiscordClient.instances[0].proxy_auth is None
+
+
+@pytest.mark.asyncio
+async def test_start_no_proxy_auth_when_only_password(monkeypatch) -> None:
+    _FakeDiscordClient.instances.clear()
+    channel = DiscordChannel(
+        DiscordConfig(
+            enabled=True,
+            token="token",
+            allow_from=["*"],
+            proxy="http://127.0.0.1:7890",
+            proxy_password="pass",
+        ),
+        MessageBus(),
+    )
+    monkeypatch.setattr("nanobot.channels.discord.DiscordBotClient", _FakeDiscordClient)
+
+    await channel.start()
+
+    assert channel.is_running is False
+    assert _FakeDiscordClient.instances[0].proxy == "http://127.0.0.1:7890"
+    assert _FakeDiscordClient.instances[0].proxy_auth is None
+
+
+# ---------------------------------------------------------------------------
+# Tests for the send() exception propagation fix
+# ---------------------------------------------------------------------------
+
+@pytest.mark.asyncio
+async def test_send_re_raises_network_error() -> None:
+    """Network errors during send must propagate so ChannelManager can retry."""
+    channel = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+    client = _FakeDiscordClient(channel, intents=None)
+    channel._client = client
+    channel._running = True
+
+    async def _failing_send_outbound(msg: OutboundMessage) -> None:
+        raise ConnectionError("network unreachable")
+
+    client.send_outbound = _failing_send_outbound  # type: ignore[method-assign]
+
+    with pytest.raises(ConnectionError, match="network unreachable"):
+        await channel.send(OutboundMessage(channel="discord", chat_id="123", content="hello"))
+
+
+@pytest.mark.asyncio
+async def test_send_re_raises_generic_exception() -> None:
+    """Any exception from send_outbound must propagate, not be swallowed."""
+    channel = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+    client = _FakeDiscordClient(channel, intents=None)
+    channel._client = client
+    channel._running = True
+
+    async def _failing_send_outbound(msg: OutboundMessage) -> None:
+        raise RuntimeError("discord API failure")
+
+    client.send_outbound = _failing_send_outbound  # type: ignore[method-assign]
+
+    with pytest.raises(RuntimeError, match="discord API failure"):
+        await channel.send(OutboundMessage(channel="discord", chat_id="123", content="hello"))
+
+
+@pytest.mark.asyncio
+async def test_send_still_stops_typing_on_error() -> None:
+    """Typing cleanup must still run in the finally block even when send raises."""
+    channel = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+    client = _FakeDiscordClient(channel, intents=None)
+    channel._client = client
+    channel._running = True
+
+    # Start a typing task so we can verify it gets cleaned up
+    start = asyncio.Event()
+    release = asyncio.Event()
+
+    async def slow_typing() -> None:
+        start.set()
+        await release.wait()
+
+    typing_channel = _FakeChannel(channel_id=123)
+    typing_channel.typing_enter_hook = slow_typing
+    await channel._start_typing(typing_channel)
+    await asyncio.wait_for(start.wait(), timeout=1.0)
+
+    async def _failing_send_outbound(msg: OutboundMessage) -> None:
+        raise ConnectionError("timeout")
+
+    client.send_outbound = _failing_send_outbound  # type: ignore[method-assign]
+
+    with pytest.raises(ConnectionError, match="timeout"):
+        await channel.send(OutboundMessage(channel="discord", chat_id="123", content="hello"))
+
+    release.set()
+    await asyncio.sleep(0)
+
+    # Typing should have been cleaned up by the finally block
+    assert channel._typing_tasks == {}
+
+
+@pytest.mark.asyncio
+async def test_send_succeeds_normally() -> None:
+    """Successful sends should work without raising."""
+    channel = DiscordChannel(DiscordConfig(enabled=True, allow_from=["*"]), MessageBus())
+    client = _FakeDiscordClient(channel, intents=None)
+    channel._client = client
+    channel._running = True
+
+    sent_messages: list[OutboundMessage] = []
+
+    async def _capture_send_outbound(msg: OutboundMessage) -> None:
+        sent_messages.append(msg)
+
+    client.send_outbound = _capture_send_outbound  # type: ignore[method-assign]
+
+    msg = OutboundMessage(channel="discord", chat_id="123", content="hello world")
+    await channel.send(msg)
+
+    assert len(sent_messages) == 1
+    assert sent_messages[0].content == "hello world"
+    assert sent_messages[0].chat_id == "123"

tests/channels/test_feishu_streaming.py

@@ -205,53 +205,22 @@ class TestSendDelta:
         ch._client.im.v1.message.create.assert_called_once()
 
     @pytest.mark.asyncio
-    async def test_stream_end_resuming_keeps_buffer(self):
-        """_resuming=True flushes text to card but keeps the buffer for the next segment."""
+    async def test_stream_end_fallback_when_final_update_fails(self):
+        """If streaming mode was closed (e.g. Feishu timeout), fall back to a regular card."""
         ch = _make_channel()
         ch._stream_bufs["oc_chat1"] = _FeishuStreamBuf(
-            text="Partial answer", card_id="card_1", sequence=2, last_edit=0.0,
+            text="Lost content", card_id="card_1", sequence=3, last_edit=0.0,
         )
-        ch._client.cardkit.v1.card_element.content.return_value = _mock_content_response()
+        ch._client.cardkit.v1.card_element.content.return_value = _mock_content_response(success=False)
+        ch._client.im.v1.message.create.return_value = _mock_send_response("om_fb")
 
-        await ch.send_delta("oc_chat1", "", metadata={"_stream_end": True, "_resuming": True})
-
-        assert "oc_chat1" in ch._stream_bufs
-        buf = ch._stream_bufs["oc_chat1"]
-        assert buf.card_id == "card_1"
-        assert buf.sequence == 3
-        ch._client.cardkit.v1.card_element.content.assert_called_once()
-        ch._client.cardkit.v1.card.settings.assert_not_called()
-
-    @pytest.mark.asyncio
-    async def test_stream_end_resuming_then_final_end(self):
-        """Full multi-segment flow: resuming mid-turn, then final end closes the card."""
-        ch = _make_channel()
-        ch._stream_bufs["oc_chat1"] = _FeishuStreamBuf(
-            text="Seg1", card_id="card_1", sequence=1, last_edit=0.0,
-        )
-        ch._client.cardkit.v1.card_element.content.return_value = _mock_content_response()
-        ch._client.cardkit.v1.card.settings.return_value = _mock_content_response()
-
-        await ch.send_delta("oc_chat1", "", metadata={"_stream_end": True, "_resuming": True})
-        assert "oc_chat1" in ch._stream_bufs
-
-        ch._stream_bufs["oc_chat1"].text += " Seg2"
         await ch.send_delta("oc_chat1", "", metadata={"_stream_end": True})
 
         assert "oc_chat1" not in ch._stream_bufs
-        ch._client.cardkit.v1.card.settings.assert_called_once()
-
-    @pytest.mark.asyncio
-    async def test_stream_end_resuming_no_card_is_noop(self):
-        """_resuming with no card_id (card creation failed) is a safe no-op."""
-        ch = _make_channel()
-        ch._stream_bufs["oc_chat1"] = _FeishuStreamBuf(
-            text="text", card_id=None, sequence=0, last_edit=0.0,
-        )
-        await ch.send_delta("oc_chat1", "", metadata={"_stream_end": True, "_resuming": True})
-
-        assert "oc_chat1" in ch._stream_bufs
-        ch._client.cardkit.v1.card_element.content.assert_not_called()
+        # Should NOT attempt to close streaming mode since update failed
+        ch._client.cardkit.v1.card.settings.assert_not_called()
+        # Should fall back to sending a regular interactive card
+        ch._client.im.v1.message.create.assert_called_once()
 
     @pytest.mark.asyncio
     async def test_stream_end_without_buf_is_noop(self):
@@ -375,22 +344,6 @@ class TestToolHintInlineStreaming:
         assert "🔧 $ cd /project" in buf.text
         assert "🔧 $ git status" in buf.text
 
-    @pytest.mark.asyncio
-    async def test_tool_hint_preserved_on_resuming_flush(self):
-        """When _resuming flushes the buffer, tool hint is kept as permanent content."""
-        ch = _make_channel()
-        ch._stream_bufs["oc_chat1"] = _FeishuStreamBuf(
-            text="Partial answer\n\n🔧 $ cd /project\n\n",
-            card_id="card_1", sequence=2, last_edit=0.0,
-        )
-        ch._client.cardkit.v1.card_element.content.return_value = _mock_content_response()
-
-        await ch.send_delta("oc_chat1", "", metadata={"_stream_end": True, "_resuming": True})
-
-        buf = ch._stream_bufs["oc_chat1"]
-        assert "Partial answer" in buf.text
-        assert "🔧 $ cd /project" in buf.text
-
     @pytest.mark.asyncio
     async def test_tool_hint_preserved_on_final_stream_end(self):
         """When final _stream_end closes the card, tool hint is kept in the final text."""

tests/channels/test_feishu_tool_hint_code_block.py

@@ -1,6 +1,7 @@
-"""Tests for FeishuChannel tool hint code block formatting."""
+"""Tests for FeishuChannel tool hint formatting."""
 
 import json
+from types import SimpleNamespace
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -28,15 +29,24 @@ def mock_feishu_channel():
     config.app_secret = "test_app_secret"
     config.encrypt_key = None
     config.verification_token = None
+    config.tool_hint_prefix = "\U0001f527"  # 🔧
     bus = MagicMock()
     channel = FeishuChannel(config, bus)
-    channel._client = MagicMock()  # Simulate initialized client
+    channel._client = MagicMock()
     return channel
 
 
+def _get_tool_hint_card(mock_send):
+    """Extract the interactive card from _send_message_sync calls."""
+    call_args = mock_send.call_args[0]
+    _, _, msg_type, content = call_args
+    assert msg_type == "interactive"
+    return json.loads(content)
+
+
 @mark.asyncio
-async def test_tool_hint_sends_code_message(mock_feishu_channel):
-    """Tool hint messages should be sent as interactive cards with code blocks."""
+async def test_tool_hint_sends_interactive_card(mock_feishu_channel):
+    """Tool hint without active buffer sends an interactive card with 🔧 style."""
     msg = OutboundMessage(
         channel="feishu",
         chat_id="oc_123456",
@@ -47,23 +57,12 @@ async def test_tool_hint_sends_code_message(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        # Verify interactive message with card was sent
         assert mock_send.call_count == 1
-        call_args = mock_send.call_args[0]
-        receive_id_type, receive_id, msg_type, content = call_args
-
-        assert receive_id_type == "chat_id"
-        assert receive_id == "oc_123456"
-        assert msg_type == "interactive"
-
-        # Parse content to verify card structure
-        card = json.loads(content)
+        card = _get_tool_hint_card(mock_send)
         assert card["config"]["wide_screen_mode"] is True
-        assert len(card["elements"]) == 1
-        assert card["elements"][0]["tag"] == "markdown"
-        # Check that code block is properly formatted with language hint
-        expected_md = "**Tool Calls**\n\n```text\nweb_search(\"test query\")\n```"
-        assert card["elements"][0]["content"] == expected_md
+        md = card["elements"][0]["content"]
+        assert "\U0001f527" in md
+        assert "web_search" in md
 
 
 @mark.asyncio
@@ -78,8 +77,6 @@ async def test_tool_hint_empty_content_does_not_send(mock_feishu_channel):
 
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
-
-        # Should not send any message
         mock_send.assert_not_called()
 
 
@@ -96,7 +93,6 @@ async def test_tool_hint_without_metadata_sends_as_normal(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        # Should send as text message (detected format)
         assert mock_send.call_count == 1
         call_args = mock_send.call_args[0]
         _, _, msg_type, content = call_args
@@ -106,7 +102,7 @@ async def test_tool_hint_without_metadata_sends_as_normal(mock_feishu_channel):
 
 @mark.asyncio
 async def test_tool_hint_multiple_tools_in_one_message(mock_feishu_channel):
-    """Multiple tool calls should be displayed each on its own line in a code block."""
+    """Multiple tool calls should each get the 🔧 prefix."""
     msg = OutboundMessage(
         channel="feishu",
         chat_id="oc_123456",
@@ -117,13 +113,11 @@ async def test_tool_hint_multiple_tools_in_one_message(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        call_args = mock_send.call_args[0]
-        msg_type = call_args[2]
-        content = json.loads(call_args[3])
-        assert msg_type == "interactive"
-        # Each tool call should be on its own line
-        expected_md = "**Tool Calls**\n\n```text\nweb_search(\"query\"),\nread_file(\"/path/to/file\")\n```"
-        assert content["elements"][0]["content"] == expected_md
+        card = _get_tool_hint_card(mock_send)
+        md = card["elements"][0]["content"]
+        assert "web_search" in md
+        assert "read_file" in md
+        assert "\U0001f527" in md
 
 
 @mark.asyncio
@@ -139,8 +133,8 @@ async def test_tool_hint_new_format_basic(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        content = json.loads(mock_send.call_args[0][3])
-        md = content["elements"][0]["content"]
+        card = _get_tool_hint_card(mock_send)
+        md = card["elements"][0]["content"]
         assert "read src/main.py" in md
         assert 'grep "TODO"' in md
 
@@ -158,16 +152,15 @@ async def test_tool_hint_new_format_with_comma_in_quotes(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        content = json.loads(mock_send.call_args[0][3])
-        md = content["elements"][0]["content"]
-        # The comma inside quotes should NOT cause a line break
+        card = _get_tool_hint_card(mock_send)
+        md = card["elements"][0]["content"]
         assert 'grep "hello, world"' in md
         assert "$ echo test" in md
 
 
 @mark.asyncio
 async def test_tool_hint_new_format_with_folding(mock_feishu_channel):
-    """Folded calls (× N) should display on separate lines."""
+    """Folded calls (× N) should display correctly."""
     msg = OutboundMessage(
         channel="feishu",
         chat_id="oc_123456",
@@ -178,8 +171,8 @@ async def test_tool_hint_new_format_with_folding(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        content = json.loads(mock_send.call_args[0][3])
-        md = content["elements"][0]["content"]
+        card = _get_tool_hint_card(mock_send)
+        md = card["elements"][0]["content"]
         assert "\u00d7 3" in md
         assert 'grep "pattern"' in md
 
@@ -197,9 +190,12 @@ async def test_tool_hint_new_format_mcp(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        content = json.loads(mock_send.call_args[0][3])
-        md = content["elements"][0]["content"]
+        card = _get_tool_hint_card(mock_send)
+        md = card["elements"][0]["content"]
         assert "4_5v::analyze_image" in md
+
+
+@mark.asyncio
 async def test_tool_hint_keeps_commas_inside_arguments(mock_feishu_channel):
     """Commas inside a single tool argument must not be split onto a new line."""
     msg = OutboundMessage(
@@ -212,10 +208,7 @@ async def test_tool_hint_keeps_commas_inside_arguments(mock_feishu_channel):
     with patch.object(mock_feishu_channel, '_send_message_sync') as mock_send:
         await mock_feishu_channel.send(msg)
 
-        content = json.loads(mock_send.call_args[0][3])
-        expected_md = (
-            "**Tool Calls**\n\n```text\n"
-            "web_search(\"foo, bar\"),\n"
-            "read_file(\"/path/to/file\")\n```"
-        )
-        assert content["elements"][0]["content"] == expected_md
+        card = _get_tool_hint_card(mock_send)
+        md = card["elements"][0]["content"]
+        assert 'web_search("foo, bar")' in md
+        assert 'read_file("/path/to/file")' in md

tests/channels/test_qq_channel.py

@@ -1,6 +1,7 @@
 import tempfile
 from pathlib import Path
 from types import SimpleNamespace
+from unittest.mock import AsyncMock, patch
 
 import pytest
 
@@ -14,6 +15,8 @@ except ImportError:
 if not QQ_AVAILABLE:
     pytest.skip("QQ dependencies not installed (qq-botpy)", allow_module_level=True)
 
+import aiohttp
+
 from nanobot.bus.events import OutboundMessage
 from nanobot.bus.queue import MessageBus
 from nanobot.channels.qq import QQChannel, QQConfig
@@ -170,3 +173,221 @@ async def test_read_media_bytes_missing_file() -> None:
     data, filename = await channel._read_media_bytes("/nonexistent/path/image.png")
     assert data is None
     assert filename is None
+
+
+# -------------------------------------------------------
+# Tests for _send_media exception handling
+# -------------------------------------------------------
+
+def _make_channel_with_local_file(suffix: str = ".png", content: bytes = b"\x89PNG\r\n"):
+    """Create a QQChannel with a fake client and a temp file for media."""
+    channel = QQChannel(
+        QQConfig(app_id="app", secret="secret", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._client = _FakeClient()
+    channel._chat_type_cache["user1"] = "c2c"
+
+    tmp = tempfile.NamedTemporaryFile(suffix=suffix, delete=False)
+    tmp.write(content)
+    tmp.close()
+    return channel, tmp.name
+
+
+@pytest.mark.asyncio
+async def test_send_media_network_error_propagates() -> None:
+    """aiohttp.ClientError (network/transport) should re-raise, not return False."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    # Make the base64 upload raise a network error
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=aiohttp.ServerDisconnectedError("connection lost"),
+    )
+
+    with pytest.raises(aiohttp.ServerDisconnectedError):
+        await channel._send_media(
+            chat_id="user1",
+            media_ref=tmp_path,
+            msg_id="msg1",
+            is_group=False,
+        )
+
+
+@pytest.mark.asyncio
+async def test_send_media_client_connector_error_propagates() -> None:
+    """aiohttp.ClientConnectorError (DNS/connection refused) should re-raise."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    from aiohttp.client_reqrep import ConnectionKey
+    conn_key = ConnectionKey("api.qq.com", 443, True, None, None, None, None)
+    connector_error = aiohttp.ClientConnectorError(
+        connection_key=conn_key,
+        os_error=OSError("Connection refused"),
+    )
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=connector_error,
+    )
+
+    with pytest.raises(aiohttp.ClientConnectorError):
+        await channel._send_media(
+            chat_id="user1",
+            media_ref=tmp_path,
+            msg_id="msg1",
+            is_group=False,
+        )
+
+
+@pytest.mark.asyncio
+async def test_send_media_oserror_propagates() -> None:
+    """OSError (low-level I/O) should re-raise for retry."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=OSError("Network is unreachable"),
+    )
+
+    with pytest.raises(OSError):
+        await channel._send_media(
+            chat_id="user1",
+            media_ref=tmp_path,
+            msg_id="msg1",
+            is_group=False,
+        )
+
+
+@pytest.mark.asyncio
+async def test_send_media_api_error_returns_false() -> None:
+    """API-level errors (botpy RuntimeError subclasses) should return False, not raise."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    # Simulate a botpy API error (e.g. ServerError is a RuntimeError subclass)
+    from botpy.errors import ServerError
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=ServerError("internal server error"),
+    )
+
+    result = await channel._send_media(
+        chat_id="user1",
+        media_ref=tmp_path,
+        msg_id="msg1",
+        is_group=False,
+    )
+    assert result is False
+
+
+@pytest.mark.asyncio
+async def test_send_media_generic_runtime_error_returns_false() -> None:
+    """Generic RuntimeError (not network) should return False."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=RuntimeError("some API error"),
+    )
+
+    result = await channel._send_media(
+        chat_id="user1",
+        media_ref=tmp_path,
+        msg_id="msg1",
+        is_group=False,
+    )
+    assert result is False
+
+
+@pytest.mark.asyncio
+async def test_send_media_value_error_returns_false() -> None:
+    """ValueError (bad API response data) should return False."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=ValueError("bad response data"),
+    )
+
+    result = await channel._send_media(
+        chat_id="user1",
+        media_ref=tmp_path,
+        msg_id="msg1",
+        is_group=False,
+    )
+    assert result is False
+
+
+@pytest.mark.asyncio
+async def test_send_media_timeout_error_propagates() -> None:
+    """asyncio.TimeoutError inherits from Exception but not ClientError/OSError.
+    However, aiohttp.ServerTimeoutError IS a ClientError subclass, so that propagates.
+    For a plain TimeoutError (which is also OSError in Python 3.11+), it should propagate."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=aiohttp.ServerTimeoutError("request timed out"),
+    )
+
+    with pytest.raises(aiohttp.ServerTimeoutError):
+        await channel._send_media(
+            chat_id="user1",
+            media_ref=tmp_path,
+            msg_id="msg1",
+            is_group=False,
+        )
+
+
+@pytest.mark.asyncio
+async def test_send_fallback_text_on_api_error() -> None:
+    """When _send_media returns False (API error), send() should emit fallback text."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    from botpy.errors import ServerError
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=ServerError("internal server error"),
+    )
+
+    await channel.send(
+        OutboundMessage(
+            channel="qq",
+            chat_id="user1",
+            content="",
+            media=[tmp_path],
+            metadata={"message_id": "msg1"},
+        )
+    )
+
+    # Should have sent a fallback text message
+    assert len(channel._client.api.c2c_calls) == 1
+    fallback_content = channel._client.api.c2c_calls[0]["content"]
+    assert "Attachment send failed" in fallback_content
+
+
+@pytest.mark.asyncio
+async def test_send_propagates_network_error_no_fallback() -> None:
+    """When _send_media raises a network error, send() should NOT silently fallback."""
+    channel, tmp_path = _make_channel_with_local_file()
+
+    channel._client.api._http = SimpleNamespace()
+    channel._client.api._http.request = AsyncMock(
+        side_effect=aiohttp.ServerDisconnectedError("connection lost"),
+    )
+
+    with pytest.raises(aiohttp.ServerDisconnectedError):
+        await channel.send(
+            OutboundMessage(
+                channel="qq",
+                chat_id="user1",
+                content="hello",
+                media=[tmp_path],
+                metadata={"message_id": "msg1"},
+            )
+        )
+
+    # No fallback text should have been sent
+    assert len(channel._client.api.c2c_calls) == 0

tests/channels/test_slack_channel.py

@@ -10,8 +10,7 @@ except ImportError:
 
 from nanobot.bus.events import OutboundMessage
 from nanobot.bus.queue import MessageBus
-from nanobot.channels.slack import SlackChannel
-from nanobot.channels.slack import SlackConfig
+from nanobot.channels.slack import SlackChannel, SlackConfig
 
 
 class _FakeAsyncWebClient:
@@ -20,6 +19,12 @@ class _FakeAsyncWebClient:
         self.file_upload_calls: list[dict[str, object | None]] = []
         self.reactions_add_calls: list[dict[str, object | None]] = []
         self.reactions_remove_calls: list[dict[str, object | None]] = []
+        self.conversations_list_calls: list[dict[str, object | None]] = []
+        self.users_list_calls: list[dict[str, object | None]] = []
+        self.conversations_open_calls: list[dict[str, object | None]] = []
+        self._conversations_pages: list[dict[str, object]] = []
+        self._users_pages: list[dict[str, object]] = []
+        self._open_dm_response: dict[str, object] = {"channel": {"id": "D_OPENED"}}
 
     async def chat_postMessage(
         self,
@@ -81,6 +86,22 @@ class _FakeAsyncWebClient:
             }
         )
 
+    async def conversations_list(self, **kwargs):
+        self.conversations_list_calls.append(kwargs)
+        if self._conversations_pages:
+            return self._conversations_pages.pop(0)
+        return {"channels": [], "response_metadata": {"next_cursor": ""}}
+
+    async def users_list(self, **kwargs):
+        self.users_list_calls.append(kwargs)
+        if self._users_pages:
+            return self._users_pages.pop(0)
+        return {"members": [], "response_metadata": {"next_cursor": ""}}
+
+    async def conversations_open(self, **kwargs):
+        self.conversations_open_calls.append(kwargs)
+        return self._open_dm_response
+
 
 @pytest.mark.asyncio
 async def test_send_uses_thread_for_channel_messages() -> None:
@@ -151,3 +172,147 @@ async def test_send_updates_reaction_when_final_response_sent() -> None:
     assert fake_web.reactions_add_calls == [
         {"channel": "C123", "name": "white_check_mark", "timestamp": "1700000000.000100"}
     ]
+
+
+@pytest.mark.asyncio
+async def test_send_resolves_channel_name_to_channel_id() -> None:
+    channel = SlackChannel(SlackConfig(enabled=True), MessageBus())
+    fake_web = _FakeAsyncWebClient()
+    fake_web._conversations_pages = [
+        {
+            "channels": [{"id": "C999", "name": "channel_x"}],
+            "response_metadata": {"next_cursor": ""},
+        }
+    ]
+    channel._web_client = fake_web
+
+    await channel.send(
+        OutboundMessage(
+            channel="slack",
+            chat_id="#channel_x",
+            content="hello",
+        )
+    )
+
+    assert fake_web.chat_post_calls == [
+        {"channel": "C999", "text": "hello\n", "thread_ts": None}
+    ]
+    assert len(fake_web.conversations_list_calls) == 1
+
+
+@pytest.mark.asyncio
+async def test_send_resolves_user_handle_to_dm_channel() -> None:
+    channel = SlackChannel(SlackConfig(enabled=True), MessageBus())
+    fake_web = _FakeAsyncWebClient()
+    fake_web._users_pages = [
+        {
+            "members": [
+                {
+                    "id": "U234",
+                    "name": "alice",
+                    "profile": {"display_name": "Alice"},
+                }
+            ],
+            "response_metadata": {"next_cursor": ""},
+        }
+    ]
+    fake_web._open_dm_response = {"channel": {"id": "D234"}}
+    channel._web_client = fake_web
+
+    await channel.send(
+        OutboundMessage(
+            channel="slack",
+            chat_id="@alice",
+            content="hello",
+        )
+    )
+
+    assert fake_web.conversations_open_calls == [{"users": "U234"}]
+    assert fake_web.chat_post_calls == [
+        {"channel": "D234", "text": "hello\n", "thread_ts": None}
+    ]
+
+
+@pytest.mark.asyncio
+async def test_send_updates_reaction_on_origin_channel_for_cross_channel_send() -> None:
+    channel = SlackChannel(SlackConfig(enabled=True, react_emoji="eyes"), MessageBus())
+    fake_web = _FakeAsyncWebClient()
+    fake_web._conversations_pages = [
+        {
+            "channels": [{"id": "C999", "name": "channel_x"}],
+            "response_metadata": {"next_cursor": ""},
+        }
+    ]
+    channel._web_client = fake_web
+
+    await channel.send(
+        OutboundMessage(
+            channel="slack",
+            chat_id="channel_x",
+            content="done",
+            metadata={
+                "slack": {
+                    "event": {"ts": "1700000000.000100", "channel": "D_ORIGIN"},
+                    "channel_type": "im",
+                },
+            },
+        )
+    )
+
+    assert fake_web.chat_post_calls == [
+        {"channel": "C999", "text": "done\n", "thread_ts": None}
+    ]
+    assert fake_web.reactions_remove_calls == [
+        {"channel": "D_ORIGIN", "name": "eyes", "timestamp": "1700000000.000100"}
+    ]
+    assert fake_web.reactions_add_calls == [
+        {"channel": "D_ORIGIN", "name": "white_check_mark", "timestamp": "1700000000.000100"}
+    ]
+
+
+@pytest.mark.asyncio
+async def test_send_does_not_reuse_origin_thread_ts_for_cross_channel_send() -> None:
+    channel = SlackChannel(SlackConfig(enabled=True), MessageBus())
+    fake_web = _FakeAsyncWebClient()
+    fake_web._conversations_pages = [
+        {
+            "channels": [{"id": "C999", "name": "channel_x"}],
+            "response_metadata": {"next_cursor": ""},
+        }
+    ]
+    channel._web_client = fake_web
+
+    await channel.send(
+        OutboundMessage(
+            channel="slack",
+            chat_id="channel_x",
+            content="done",
+            metadata={
+                "slack": {
+                    "event": {"ts": "1700000000.000100", "channel": "C_ORIGIN"},
+                    "thread_ts": "1700000000.000200",
+                    "channel_type": "channel",
+                },
+            },
+        )
+    )
+
+    assert fake_web.chat_post_calls == [
+        {"channel": "C999", "text": "done\n", "thread_ts": None}
+    ]
+
+
+@pytest.mark.asyncio
+async def test_send_raises_when_named_target_cannot_be_resolved() -> None:
+    channel = SlackChannel(SlackConfig(enabled=True), MessageBus())
+    fake_web = _FakeAsyncWebClient()
+    channel._web_client = fake_web
+
+    with pytest.raises(ValueError, match="was not found"):
+        await channel.send(
+            OutboundMessage(
+                channel="slack",
+                chat_id="#missing-channel",
+                content="hello",
+            )
+        )

tests/channels/test_telegram_channel.py

@@ -135,6 +135,7 @@ def _make_telegram_update(
     entities=None,
     caption_entities=None,
     reply_to_message=None,
+    location=None,
 ):
     user = SimpleNamespace(id=12345, username="alice", first_name="Alice")
     message = SimpleNamespace(
@@ -149,6 +150,7 @@ def _make_telegram_update(
         voice=None,
         audio=None,
         document=None,
+        location=location,
         media_group_id=None,
         message_thread_id=None,
         message_id=1,
@@ -385,6 +387,84 @@ async def test_send_delta_stream_end_treats_not_modified_as_success() -> None:
     assert "123" not in channel._stream_bufs
 
 
+@pytest.mark.asyncio
+async def test_send_delta_stream_end_does_not_fallback_on_network_timeout() -> None:
+    """TimedOut during HTML edit should propagate, never fall back to plain text."""
+    from telegram.error import TimedOut
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+    # _call_with_retry retries TimedOut up to 3 times, so the mock will be called
+    # multiple times – but all calls must be with parse_mode="HTML" (no plain fallback).
+    channel._app.bot.edit_message_text = AsyncMock(side_effect=TimedOut("network timeout"))
+    channel._stream_bufs["123"] = _StreamBuf(text="hello", message_id=7, last_edit=0.0)
+
+    with pytest.raises(TimedOut, match="network timeout"):
+        await channel.send_delta("123", "", {"_stream_end": True})
+
+    # Every call to edit_message_text must have used parse_mode="HTML" —
+    # no plain-text fallback call should have been made.
+    for call in channel._app.bot.edit_message_text.call_args_list:
+        assert call.kwargs.get("parse_mode") == "HTML"
+    # Buffer should still be present (not cleaned up on error)
+    assert "123" in channel._stream_bufs
+
+
+@pytest.mark.asyncio
+async def test_send_delta_stream_end_does_not_fallback_on_network_error() -> None:
+    """NetworkError during HTML edit should propagate, never fall back to plain text."""
+    from telegram.error import NetworkError
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+    channel._app.bot.edit_message_text = AsyncMock(side_effect=NetworkError("connection reset"))
+    channel._stream_bufs["123"] = _StreamBuf(text="hello", message_id=7, last_edit=0.0)
+
+    with pytest.raises(NetworkError, match="connection reset"):
+        await channel.send_delta("123", "", {"_stream_end": True})
+
+    # Every call to edit_message_text must have used parse_mode="HTML" —
+    # no plain-text fallback call should have been made.
+    for call in channel._app.bot.edit_message_text.call_args_list:
+        assert call.kwargs.get("parse_mode") == "HTML"
+    # Buffer should still be present (not cleaned up on error)
+    assert "123" in channel._stream_bufs
+
+
+@pytest.mark.asyncio
+async def test_send_delta_stream_end_falls_back_on_bad_request() -> None:
+    """BadRequest (HTML parse error) should still trigger plain-text fallback."""
+    from telegram.error import BadRequest
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+
+    # First call (HTML) raises BadRequest, second call (plain) succeeds
+    channel._app.bot.edit_message_text = AsyncMock(
+        side_effect=[BadRequest("Can't parse entities"), None]
+    )
+    channel._stream_bufs["123"] = _StreamBuf(text="hello <bad>", message_id=7, last_edit=0.0)
+
+    await channel.send_delta("123", "", {"_stream_end": True})
+
+    # edit_message_text should have been called twice: once for HTML, once for plain fallback
+    assert channel._app.bot.edit_message_text.call_count == 2
+    # Second call should not use parse_mode="HTML"
+    second_call_kwargs = channel._app.bot.edit_message_text.call_args_list[1].kwargs
+    assert "parse_mode" not in second_call_kwargs or second_call_kwargs.get("parse_mode") is None
+    # Buffer should be cleaned up on success
+    assert "123" not in channel._stream_bufs
+
+
 @pytest.mark.asyncio
 async def test_send_delta_stream_end_splits_oversized_reply() -> None:
     """Final streamed reply exceeding Telegram limit is split into chunks."""
@@ -1112,3 +1192,204 @@ async def test_on_help_includes_restart_command() -> None:
     assert "/dream" in help_text
     assert "/dream-log" in help_text
     assert "/dream-restore" in help_text
+
+
+@pytest.mark.asyncio
+async def test_on_message_location_content() -> None:
+    """Location messages are forwarded as [location: lat, lon] content."""
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"], group_policy="open"),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+    handled = []
+    async def capture_handle(**kwargs) -> None:
+        handled.append(kwargs)
+    channel._handle_message = capture_handle
+    channel._start_typing = lambda _chat_id: None
+
+    location = SimpleNamespace(latitude=48.8566, longitude=2.3522)
+    update = _make_telegram_update(location=location)
+    await channel._on_message(update, None)
+
+    assert len(handled) == 1
+    assert handled[0]["content"] == "[location: 48.8566, 2.3522]"
+
+
+@pytest.mark.asyncio
+async def test_on_message_location_with_text() -> None:
+    """Location messages with accompanying text include both in content."""
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"], group_policy="open"),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+    handled = []
+    async def capture_handle(**kwargs) -> None:
+        handled.append(kwargs)
+    channel._handle_message = capture_handle
+    channel._start_typing = lambda _chat_id: None
+
+    location = SimpleNamespace(latitude=51.5074, longitude=-0.1278)
+    update = _make_telegram_update(text="meet me here", location=location)
+    await channel._on_message(update, None)
+
+    assert len(handled) == 1
+    assert "meet me here" in handled[0]["content"]
+    assert "[location: 51.5074, -0.1278]" in handled[0]["content"]
+
+
+# ---------------------------------------------------------------------------
+# Tests for retry amplification fix (issue #3050)
+# ---------------------------------------------------------------------------
+
+@pytest.mark.asyncio
+async def test_send_text_does_not_fallback_on_network_timeout() -> None:
+    """TimedOut should propagate immediately, NOT trigger plain-text fallback.
+
+    Before the fix, _send_text caught ALL exceptions (including TimedOut)
+    and retried as plain text, doubling connection demand during pool
+    exhaustion — see issue #3050.
+    """
+    from telegram.error import TimedOut
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+
+    call_count = 0
+
+    async def always_timeout(**kwargs):
+        nonlocal call_count
+        call_count += 1
+        raise TimedOut()
+
+    channel._app.bot.send_message = always_timeout
+
+    import nanobot.channels.telegram as tg_mod
+    orig_delay = tg_mod._SEND_RETRY_BASE_DELAY
+    tg_mod._SEND_RETRY_BASE_DELAY = 0.01
+    try:
+        with pytest.raises(TimedOut):
+            await channel._send_text(123, "hello", None, {})
+    finally:
+        tg_mod._SEND_RETRY_BASE_DELAY = orig_delay
+
+    # With the fix: only _call_with_retry's 3 HTML attempts (no plain fallback).
+    # Before the fix: 3 HTML + 3 plain = 6 attempts.
+    assert call_count == 3, (
+        f"Expected 3 calls (HTML retries only), got {call_count} "
+        "(plain-text fallback should not trigger on TimedOut)"
+    )
+
+
+@pytest.mark.asyncio
+async def test_send_text_does_not_fallback_on_network_error() -> None:
+    """NetworkError should propagate immediately, NOT trigger plain-text fallback."""
+    from telegram.error import NetworkError
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+
+    call_count = 0
+
+    async def always_network_error(**kwargs):
+        nonlocal call_count
+        call_count += 1
+        raise NetworkError("Connection reset")
+
+    channel._app.bot.send_message = always_network_error
+
+    import nanobot.channels.telegram as tg_mod
+    orig_delay = tg_mod._SEND_RETRY_BASE_DELAY
+    tg_mod._SEND_RETRY_BASE_DELAY = 0.01
+    try:
+        with pytest.raises(NetworkError):
+            await channel._send_text(123, "hello", None, {})
+    finally:
+        tg_mod._SEND_RETRY_BASE_DELAY = orig_delay
+
+    # _call_with_retry does NOT retry NetworkError (only TimedOut/RetryAfter),
+    # so it raises after 1 attempt. The fix prevents plain-text fallback.
+    # Before the fix: 1 HTML + 1 plain = 2. After the fix: 1 HTML only.
+    assert call_count == 1, (
+        f"Expected 1 call (HTML only, no plain fallback), got {call_count}"
+    )
+
+
+@pytest.mark.asyncio
+async def test_send_text_falls_back_on_bad_request() -> None:
+    """BadRequest (HTML parse error) should still trigger plain-text fallback."""
+    from telegram.error import BadRequest
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+
+    original_send = channel._app.bot.send_message
+    html_call_count = 0
+
+    async def html_fails(**kwargs):
+        nonlocal html_call_count
+        if kwargs.get("parse_mode") == "HTML":
+            html_call_count += 1
+            raise BadRequest("Can't parse entities")
+        return await original_send(**kwargs)
+
+    channel._app.bot.send_message = html_fails
+
+    import nanobot.channels.telegram as tg_mod
+    orig_delay = tg_mod._SEND_RETRY_BASE_DELAY
+    tg_mod._SEND_RETRY_BASE_DELAY = 0.01
+    try:
+        await channel._send_text(123, "hello **world**", None, {})
+    finally:
+        tg_mod._SEND_RETRY_BASE_DELAY = orig_delay
+
+    # HTML attempt failed with BadRequest → fallback to plain text succeeds.
+    assert html_call_count == 1, f"Expected 1 HTML attempt, got {html_call_count}"
+    assert len(channel._app.bot.sent_messages) == 1
+    # Plain text send should NOT have parse_mode
+    assert channel._app.bot.sent_messages[0].get("parse_mode") is None
+
+
+@pytest.mark.asyncio
+async def test_send_text_bad_request_plain_fallback_exhausted() -> None:
+    """When both HTML and plain-text fallback fail with BadRequest, the error propagates."""
+    from telegram.error import BadRequest
+
+    channel = TelegramChannel(
+        TelegramConfig(enabled=True, token="123:abc", allow_from=["*"]),
+        MessageBus(),
+    )
+    channel._app = _FakeApp(lambda: None)
+
+    call_count = 0
+
+    async def always_bad_request(**kwargs):
+        nonlocal call_count
+        call_count += 1
+        raise BadRequest("Bad request")
+
+    channel._app.bot.send_message = always_bad_request
+
+    import nanobot.channels.telegram as tg_mod
+    orig_delay = tg_mod._SEND_RETRY_BASE_DELAY
+    tg_mod._SEND_RETRY_BASE_DELAY = 0.01
+    try:
+        with pytest.raises(BadRequest):
+            await channel._send_text(123, "hello", None, {})
+    finally:
+        tg_mod._SEND_RETRY_BASE_DELAY = orig_delay
+
+    # _call_with_retry does NOT retry BadRequest (only TimedOut/RetryAfter),
+    # so HTML fails after 1 attempt → fallback to plain also fails after 1 attempt.
+    # Before the fix: 2 total. After the fix: still 2 (BadRequest SHOULD fallback).
+    assert call_count == 2, f"Expected 2 calls (1 HTML + 1 plain), got {call_count}"

tests/channels/test_websocket_integration.py

@@ -8,6 +8,7 @@ from __future__ import annotations
 
 import asyncio
 import json
+from typing import Any
 from unittest.mock import AsyncMock, MagicMock
 
 import pytest

tests/channels/test_wecom_channel.py

@@ -541,6 +541,50 @@ async def test_process_voice_message() -> None:
     assert "[voice]" in msg.content
 
 
+@pytest.mark.asyncio
+async def test_process_mixed_message() -> None:
+    """Mixed message: contains picture and text message types."""
+    channel = WecomChannel(WecomConfig(bot_id="b", secret="s", allow_from=["user1"]), MessageBus())
+    client = _FakeWeComClient()
+
+    with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as f:
+        f.write(b"\x89PNG\r\n")
+        saved = f.name
+
+    client.download_file.return_value = (b"\x89PNG\r\n", "photo.png")
+    channel._client = client
+
+    try:
+        with patch("nanobot.channels.wecom.get_media_dir", return_value=Path(os.path.dirname(saved))):
+            frame = _FakeFrame(body={
+                "msgid": "msg_mixed_1",
+                "chatid": "chat1",
+                "msgtype": "mixed",
+                "from": {"userid": "user1"},
+                "mixed": {
+                    "msg_item": [
+                        {"msgtype": "text", "text": {"content": "hello wecom"}},
+                        {"msgtype": "image", "image": {"url": "https://example.com/img.png", "aeskey": "key123"}}
+                    ]
+                }
+            })
+            await channel._process_message(frame, "mixed")
+
+        msg = await channel.bus.consume_inbound()
+        assert msg.sender_id == "user1"
+        assert msg.chat_id == "chat1"
+        assert msg.content.startswith("hello wecom")
+        assert msg.metadata["msg_type"] == "mixed"
+        assert len(msg.media) == 1
+        assert msg.media[0].endswith("photo.png")
+        assert "[image:" in msg.content
+    finally:
+        # Clean up any photo.png in tempdir
+        p = os.path.join(os.path.dirname(saved), "photo.png")
+        if os.path.exists(p):
+            os.unlink(p)
+
+
 @pytest.mark.asyncio
 async def test_process_message_deduplication() -> None:
     """Same msg_id is not processed twice."""

tests/channels/test_weixin_channel.py

@@ -1003,3 +1003,185 @@ async def test_download_media_item_non_image_requires_aes_key_even_with_full_url
 
     assert saved_path is None
     channel._client.get.assert_not_awaited()
+
+
+# ---------------------------------------------------------------------------
+# Tests for media-send error classification (network vs non-network errors)
+# ---------------------------------------------------------------------------
+
+
+def _make_outbound_msg(chat_id: str = "wx-user", content: str = "", media: list | None = None):
+    """Build a minimal OutboundMessage-like object for send() tests."""
+    from nanobot.bus.events import OutboundMessage
+
+    return OutboundMessage(
+        channel="weixin",
+        chat_id=chat_id,
+        content=content,
+        media=media or [],
+        metadata={},
+    )
+
+
+@pytest.mark.asyncio
+async def test_send_media_timeout_error_propagates_without_text_fallback() -> None:
+    """httpx.TimeoutException during media send must re-raise immediately,
+    NOT fall back to _send_text (which would also fail during network issues)."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+    channel._send_media_file = AsyncMock(side_effect=httpx.TimeoutException("timed out"))
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", media=["/tmp/photo.jpg"])
+
+    with pytest.raises(httpx.TimeoutException, match="timed out"):
+        await channel.send(msg)
+
+    # _send_text must NOT have been called as a fallback
+    channel._send_text.assert_not_awaited()
+
+
+@pytest.mark.asyncio
+async def test_send_media_transport_error_propagates_without_text_fallback() -> None:
+    """httpx.TransportError during media send must re-raise immediately."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+    channel._send_media_file = AsyncMock(
+        side_effect=httpx.TransportError("connection reset")
+    )
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", media=["/tmp/photo.jpg"])
+
+    with pytest.raises(httpx.TransportError, match="connection reset"):
+        await channel.send(msg)
+
+    channel._send_text.assert_not_awaited()
+
+
+@pytest.mark.asyncio
+async def test_send_media_5xx_http_status_error_propagates_without_text_fallback() -> None:
+    """httpx.HTTPStatusError with a 5xx status must re-raise immediately."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+
+    fake_response = httpx.Response(
+        status_code=503,
+        request=httpx.Request("POST", "https://example.test/upload"),
+    )
+    channel._send_media_file = AsyncMock(
+        side_effect=httpx.HTTPStatusError(
+            "Service Unavailable", request=fake_response.request, response=fake_response
+        )
+    )
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", media=["/tmp/photo.jpg"])
+
+    with pytest.raises(httpx.HTTPStatusError, match="Service Unavailable"):
+        await channel.send(msg)
+
+    channel._send_text.assert_not_awaited()
+
+
+@pytest.mark.asyncio
+async def test_send_media_4xx_http_status_error_falls_back_to_text() -> None:
+    """httpx.HTTPStatusError with a 4xx status should fall back to text, not re-raise."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+
+    fake_response = httpx.Response(
+        status_code=400,
+        request=httpx.Request("POST", "https://example.test/upload"),
+    )
+    channel._send_media_file = AsyncMock(
+        side_effect=httpx.HTTPStatusError(
+            "Bad Request", request=fake_response.request, response=fake_response
+        )
+    )
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", media=["/tmp/photo.jpg"])
+
+    # Should NOT raise — 4xx is a client error, non-retryable
+    await channel.send(msg)
+
+    # _send_text should have been called with the fallback message
+    channel._send_text.assert_awaited_once_with(
+        "wx-user", "[Failed to send: photo.jpg]", "ctx-1"
+    )
+
+
+@pytest.mark.asyncio
+async def test_send_media_file_not_found_falls_back_to_text() -> None:
+    """FileNotFoundError (a non-network error) should fall back to text."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+    channel._send_media_file = AsyncMock(
+        side_effect=FileNotFoundError("Media file not found: /tmp/missing.jpg")
+    )
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", media=["/tmp/missing.jpg"])
+
+    # Should NOT raise
+    await channel.send(msg)
+
+    channel._send_text.assert_awaited_once_with(
+        "wx-user", "[Failed to send: missing.jpg]", "ctx-1"
+    )
+
+
+@pytest.mark.asyncio
+async def test_send_media_value_error_falls_back_to_text() -> None:
+    """ValueError (e.g. unsupported format) should fall back to text."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+    channel._send_media_file = AsyncMock(
+        side_effect=ValueError("Unsupported media format")
+    )
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", media=["/tmp/file.xyz"])
+
+    # Should NOT raise
+    await channel.send(msg)
+
+    channel._send_text.assert_awaited_once_with(
+        "wx-user", "[Failed to send: file.xyz]", "ctx-1"
+    )
+
+
+@pytest.mark.asyncio
+async def test_send_media_network_error_does_not_double_api_calls() -> None:
+    """During network issues, media send should make exactly 1 API call attempt,
+    not 2 (media + text fallback).  Verify total call count."""
+    channel, _bus = _make_channel()
+    channel._client = object()
+    channel._token = "token"
+    channel._context_tokens["wx-user"] = "ctx-1"
+    channel._send_media_file = AsyncMock(
+        side_effect=httpx.ConnectError("connection refused")
+    )
+    channel._send_text = AsyncMock()
+
+    msg = _make_outbound_msg(chat_id="wx-user", content="hello", media=["/tmp/img.png"])
+
+    with pytest.raises(httpx.ConnectError):
+        await channel.send(msg)
+
+    # _send_media_file called once, _send_text never called
+    channel._send_media_file.assert_awaited_once()
+    channel._send_text.assert_not_awaited()

tests/cli/test_commands.py

@@ -257,6 +257,28 @@ def test_config_accepts_camel_case_explicit_provider_name_for_coding_plan():
     assert config.get_api_base() == "https://ark.cn-beijing.volces.com/api/coding/v3"
 
 
+def test_config_accepts_lm_studio_without_api_key_and_uses_default_localhost_api_base():
+    config = Config.model_validate(
+        {
+            "agents": {
+                "defaults": {
+                    "provider": "lm_studio",
+                    "model": "local-model",
+                }
+            },
+            "providers": {
+                "lmStudio": {
+                    "apiKey": None,
+                }
+            },
+        }
+    )
+
+    assert config.get_provider_name() == "lm_studio"
+    assert config.get_api_key() is None
+    assert config.get_api_base() == "http://localhost:1234/v1"
+
+
 def test_find_by_name_accepts_camel_case_and_hyphen_aliases():
     assert find_by_name("volcengineCodingPlan") is not None
     assert find_by_name("volcengineCodingPlan").name == "volcengine_coding_plan"
@@ -1126,6 +1148,153 @@ def test_gateway_cli_port_overrides_configured_port(monkeypatch, tmp_path: Path)
     assert "port 18792" in result.stdout
 
 
+def test_gateway_health_endpoint_binds_and_serves_expected_responses(
+    monkeypatch, tmp_path: Path
+) -> None:
+    config_file = _write_instance_config(tmp_path)
+    config = Config()
+    config.gateway.port = 18791
+    captured: dict[str, object] = {}
+
+    class _FakeDream:
+        model = None
+        max_batch_size = 0
+        max_iterations = 0
+
+        async def run(self) -> None:
+            return None
+
+    class _FakeAgentLoop:
+        def __init__(self, **_kwargs) -> None:
+            self.model = "test-model"
+            self.dream = _FakeDream()
+
+        async def run(self) -> None:
+            await asyncio.Event().wait()
+
+        async def close_mcp(self) -> None:
+            return None
+
+        def stop(self) -> None:
+            return None
+
+    class _FakeChannelManager:
+        def __init__(self, _config, _bus) -> None:
+            self.enabled_channels = ["telegram", "discord"]
+
+        async def start_all(self) -> None:
+            await asyncio.Event().wait()
+
+        async def stop_all(self) -> None:
+            return None
+
+    class _FakeCronService:
+        def __init__(self, _store_path: Path) -> None:
+            self.on_job = None
+
+        async def start(self) -> None:
+            return None
+
+        def stop(self) -> None:
+            return None
+
+        def status(self) -> dict[str, int]:
+            return {"jobs": 0}
+
+        def register_system_job(self, _job) -> None:
+            return None
+
+    class _FakeHeartbeatService:
+        def __init__(self, **_kwargs) -> None:
+            return None
+
+        async def start(self) -> None:
+            return None
+
+        def stop(self) -> None:
+            return None
+
+    class _FakeServer:
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, exc_type, exc, tb) -> bool:
+            return False
+
+        async def serve_forever(self) -> None:
+            raise _StopGatewayError("stop")
+
+    async def _fake_start_server(handler, host: str, port: int):
+        captured["handler"] = handler
+        captured["host"] = host
+        captured["port"] = port
+        return _FakeServer()
+
+    class _FakeReader:
+        def __init__(self, payload: bytes) -> None:
+            self.payload = payload
+
+        async def read(self, _size: int) -> bytes:
+            return self.payload
+
+    class _FakeWriter:
+        def __init__(self) -> None:
+            self.output = b""
+            self.closed = False
+
+        def write(self, data: bytes) -> None:
+            self.output += data
+
+        async def drain(self) -> None:
+            return None
+
+        def close(self) -> None:
+            self.closed = True
+
+    _patch_cli_command_runtime(
+        monkeypatch,
+        config,
+        message_bus=lambda: object(),
+        session_manager=lambda _workspace: object(),
+    )
+    monkeypatch.setattr("nanobot.agent.loop.AgentLoop", _FakeAgentLoop)
+    monkeypatch.setattr("nanobot.channels.manager.ChannelManager", _FakeChannelManager)
+    monkeypatch.setattr("nanobot.cron.service.CronService", _FakeCronService)
+    monkeypatch.setattr("nanobot.heartbeat.service.HeartbeatService", _FakeHeartbeatService)
+    monkeypatch.setattr("asyncio.start_server", _fake_start_server)
+
+    result = runner.invoke(app, ["gateway", "--config", str(config_file)])
+
+    assert result.exit_code == 0
+    assert captured["host"] == "127.0.0.1"
+    assert captured["port"] == 18791
+    assert "Health endpoint: http://127.0.0.1:18791/health" in result.stdout
+
+    def _call_handler(path: str) -> tuple[str, _FakeWriter]:
+        request = f"GET {path} HTTP/1.1\r\nHost: localhost\r\n\r\n".encode()
+        writer = _FakeWriter()
+        handler = captured["handler"]
+        assert callable(handler)
+        asyncio.run(handler(_FakeReader(request), writer))
+        return writer.output.decode(), writer
+
+    root_response, root_writer = _call_handler("/")
+    assert root_writer.closed is True
+    assert "HTTP/1.0 404 Not Found" in root_response
+    assert root_response.endswith("\r\n\r\nNot Found")
+
+    health_response, health_writer = _call_handler("/health")
+    assert health_writer.closed is True
+    assert "HTTP/1.0 200 OK" in health_response
+    health_body = json.loads(health_response.split("\r\n\r\n", 1)[1])
+    assert health_body == {"status": "ok"}
+
+    missing_response, missing_writer = _call_handler("/missing")
+    assert missing_writer.closed is True
+    assert "HTTP/1.0 404 Not Found" in missing_response
+    assert missing_response.endswith("\r\n\r\nNot Found")
+
+
 def test_serve_uses_api_config_defaults_and_workspace_override(
     monkeypatch, tmp_path: Path
 ) -> None:

tests/cli/test_restart_command.py

@@ -140,6 +140,7 @@ class TestRestartCommand:
         loop.consolidator.estimate_session_prompt_tokens = MagicMock(
             return_value=(20500, "tiktoken")
         )
+        loop.subagents.get_running_count_by_session.return_value = 0
 
         msg = InboundMessage(channel="telegram", sender_id="u1", chat_id="c1", content="/status")
 
@@ -148,11 +149,36 @@ class TestRestartCommand:
         assert response is not None
         assert "Model: test-model" in response.content
         assert "Tokens: 0 in / 0 out" in response.content
-        assert "Context: 20k/64k (31%)" in response.content
+        assert "Context: 20k/65k (31% of input budget)" in response.content
         assert "Session: 3 messages" in response.content
         assert "Uptime: 2m 5s" in response.content
+        assert "Tasks: 0 active" in response.content
         assert response.metadata == {"render_as": "text"}
 
+    @pytest.mark.asyncio
+    async def test_status_counts_running_dispatch_and_subagent_tasks(self):
+        loop, _bus = _make_loop()
+        session = MagicMock()
+        session.get_history.return_value = [{"role": "user"}]
+        loop.sessions.get_or_create.return_value = session
+        loop.consolidator.estimate_session_prompt_tokens = MagicMock(
+            return_value=(1000, "tiktoken")
+        )
+
+        running_task = MagicMock()
+        running_task.done.return_value = False
+        finished_task = MagicMock()
+        finished_task.done.return_value = True
+
+        msg = InboundMessage(channel="telegram", sender_id="u1", chat_id="c1", content="/status")
+        loop._active_tasks[msg.session_key] = [running_task, finished_task]
+        loop.subagents.get_running_count_by_session.return_value = 2
+
+        response = await loop._process_message(msg)
+
+        assert response is not None
+        assert "Tasks: 3 active" in response.content
+
     @pytest.mark.asyncio
     async def test_run_agent_loop_resets_usage_when_provider_omits_it(self):
         loop, _bus = _make_loop()
@@ -179,6 +205,7 @@ class TestRestartCommand:
         loop.consolidator.estimate_session_prompt_tokens = MagicMock(
             return_value=(0, "none")
         )
+        loop.subagents.get_running_count_by_session.return_value = 0
 
         response = await loop._process_message(
             InboundMessage(channel="telegram", sender_id="u1", chat_id="c1", content="/status")
@@ -186,7 +213,8 @@ class TestRestartCommand:
 
         assert response is not None
         assert "Tokens: 1200 in / 34 out" in response.content
-        assert "Context: 1k/64k (1%)" in response.content
+        assert "Context: 1k/65k (1% of input budget)" in response.content
+        assert "Tasks: 0 active" in response.content
 
     @pytest.mark.asyncio
     async def test_process_direct_preserves_render_metadata(self):
@@ -195,6 +223,7 @@ class TestRestartCommand:
         session.get_history.return_value = []
         loop.sessions.get_or_create.return_value = session
         loop.subagents.get_running_count.return_value = 0
+        loop.subagents.get_running_count_by_session.return_value = 0
 
         response = await loop.process_direct("/status", session_key="cli:test")
 

tests/config/test_config_migration.py

@@ -140,6 +140,71 @@ def test_onboard_refresh_backfills_missing_channel_fields(tmp_path, monkeypatch)
     assert saved["channels"]["qq"]["msgFormat"] == "plain"
 
 
+def test_load_config_migrates_legacy_my_tool_keys(tmp_path) -> None:
+    config_path = tmp_path / "config.json"
+    config_path.write_text(
+        json.dumps(
+            {
+                "tools": {
+                    "myEnabled": False,
+                    "mySet": True,
+                }
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    config = load_config(config_path)
+
+    assert config.tools.my.enable is False
+    assert config.tools.my.allow_set is True
+
+
+def test_save_config_rewrites_legacy_my_tool_keys(tmp_path) -> None:
+    config_path = tmp_path / "config.json"
+    config_path.write_text(
+        json.dumps(
+            {
+                "tools": {
+                    "myEnabled": False,
+                    "mySet": True,
+                }
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    config = load_config(config_path)
+    save_config(config, config_path)
+    saved = json.loads(config_path.read_text(encoding="utf-8"))
+
+    tools = saved["tools"]
+    assert "myEnabled" not in tools
+    assert "mySet" not in tools
+    assert tools["my"] == {"enable": False, "allowSet": True}
+
+
+def test_new_my_tool_keys_take_precedence_over_legacy(tmp_path) -> None:
+    config_path = tmp_path / "config.json"
+    config_path.write_text(
+        json.dumps(
+            {
+                "tools": {
+                    "myEnabled": False,
+                    "mySet": False,
+                    "my": {"enable": True, "allowSet": True},
+                }
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    config = load_config(config_path)
+
+    assert config.tools.my.enable is True
+    assert config.tools.my.allow_set is True
+
+
 def test_load_config_resets_ssrf_whitelist_when_next_config_is_empty(tmp_path) -> None:
     whitelisted = tmp_path / "whitelisted.json"
     whitelisted.write_text(

tests/cron/test_cron_service.py

@@ -1,5 +1,6 @@
 import asyncio
 import json
+import time
 
 import pytest
 
@@ -114,6 +115,41 @@ async def test_run_history_persisted_to_disk(tmp_path) -> None:
     assert loaded.state.run_history[0].status == "ok"
 
 
+@pytest.mark.asyncio
+async def test_run_job_disabled_does_not_flip_running_state(tmp_path) -> None:
+    store_path = tmp_path / "cron" / "jobs.json"
+    service = CronService(store_path, on_job=lambda _: asyncio.sleep(0))
+    job = service.add_job(
+        name="disabled",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    service.enable_job(job.id, enabled=False)
+
+    result = await service.run_job(job.id)
+
+    assert result is False
+    assert service._running is False
+
+
+@pytest.mark.asyncio
+async def test_run_job_preserves_running_service_state(tmp_path) -> None:
+    store_path = tmp_path / "cron" / "jobs.json"
+    service = CronService(store_path, on_job=lambda _: asyncio.sleep(0))
+    service._running = True
+    job = service.add_job(
+        name="manual",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+
+    result = await service.run_job(job.id, force=True)
+
+    assert result is True
+    assert service._running is True
+    service.stop()
+
+
 @pytest.mark.asyncio
 async def test_running_service_honors_external_disable(tmp_path) -> None:
     store_path = tmp_path / "cron" / "jobs.json"
@@ -156,3 +192,375 @@ def test_remove_job_refuses_system_jobs(tmp_path) -> None:
 
     assert result == "protected"
     assert service.get_job("dream") is not None
+
+
+@pytest.mark.asyncio
+async def test_start_server_not_jobs(tmp_path):
+    store_path = tmp_path / "cron" / "jobs.json"
+    called = []
+    async def on_job(job):
+        called.append(job.name)
+
+    service = CronService(store_path, on_job=on_job, max_sleep_ms=1000)
+    await service.start()
+    assert len(service.list_jobs()) == 0
+
+    service2 = CronService(tmp_path / "cron" / "jobs.json")
+    service2.add_job(
+        name="hist",
+        schedule=CronSchedule(kind="every", every_ms=500),
+        message="hello",
+    )
+    assert len(service.list_jobs()) == 1
+    await asyncio.sleep(2)
+    assert len(called) != 0
+    service.stop()
+
+
+@pytest.mark.asyncio
+async def test_subsecond_job_not_delayed_to_one_second(tmp_path):
+    store_path = tmp_path / "cron" / "jobs.json"
+    called = []
+
+    async def on_job(job):
+        called.append(job.name)
+
+    service = CronService(store_path, on_job=on_job, max_sleep_ms=5000)
+    service.add_job(
+        name="fast",
+        schedule=CronSchedule(kind="every", every_ms=100),
+        message="hello",
+    )
+    await service.start()
+    try:
+        await asyncio.sleep(0.35)
+        assert called
+    finally:
+        service.stop()
+
+
+@pytest.mark.asyncio
+async def test_running_service_picks_up_external_add(tmp_path):
+    """A running service should detect and execute a job added by another instance."""
+    store_path = tmp_path / "cron" / "jobs.json"
+    called: list[str] = []
+
+    async def on_job(job):
+        called.append(job.name)
+
+    service = CronService(store_path, on_job=on_job)
+    service.add_job(
+        name="heartbeat",
+        schedule=CronSchedule(kind="every", every_ms=150),
+        message="tick",
+    )
+    await service.start()
+    try:
+        await asyncio.sleep(0.05)
+
+        external = CronService(store_path)
+        external.add_job(
+            name="external",
+            schedule=CronSchedule(kind="every", every_ms=150),
+            message="ping",
+        )
+
+        await asyncio.sleep(2)
+        assert "external" in called
+    finally:
+        service.stop()
+
+
+@pytest.mark.asyncio
+async def test_add_job_during_jobs_exec(tmp_path):
+    store_path = tmp_path / "cron" / "jobs.json"
+    run_once = True
+
+    async def on_job(job):
+        nonlocal run_once
+        if run_once:
+            service2 = CronService(store_path, on_job=lambda x: asyncio.sleep(0))
+            service2.add_job(
+                name="test",
+                schedule=CronSchedule(kind="every", every_ms=150),
+                message="tick",
+            )
+            run_once = False
+
+    service = CronService(store_path, on_job=on_job)
+    service.add_job(
+        name="heartbeat",
+        schedule=CronSchedule(kind="every", every_ms=150),
+        message="tick",
+    )
+    assert len(service.list_jobs()) == 1
+    await service.start()
+    try:
+        await asyncio.sleep(3)
+        jobs = service.list_jobs()
+        assert len(jobs) == 2
+        assert "test" in [j.name for j in jobs]
+    finally:
+        service.stop()
+
+
+@pytest.mark.asyncio
+async def test_external_update_preserves_run_history_records(tmp_path):
+    store_path = tmp_path / "cron" / "jobs.json"
+    service = CronService(store_path, on_job=lambda _: asyncio.sleep(0))
+    job = service.add_job(
+        name="history",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    await service.run_job(job.id, force=True)
+
+    external = CronService(store_path)
+    updated = external.enable_job(job.id, enabled=False)
+    assert updated is not None
+
+    fresh = CronService(store_path)
+    loaded = fresh.get_job(job.id)
+    assert loaded is not None
+    assert loaded.state.run_history
+    assert loaded.state.run_history[0].status == "ok"
+
+    fresh._running = True
+    fresh._save_store()
+
+
+# ── timer race regression tests ──
+
+
+@pytest.mark.asyncio
+async def test_timer_execution_is_not_rolled_back_by_list_jobs_reload(tmp_path):
+    """list_jobs() during _on_timer should not replace the active store and re-run the same due job."""
+    store_path = tmp_path / "cron" / "jobs.json"
+    calls: list[str] = []
+
+    async def on_job(job):
+        calls.append(job.id)
+        # Simulate frontend polling list_jobs while the timer callback is mid-execution.
+        service.list_jobs(include_disabled=True)
+        await asyncio.sleep(0)
+
+    service = CronService(store_path, on_job=on_job)
+    service._running = True
+    service._load_store()
+    service._arm_timer = lambda: None
+
+    job = service.add_job(
+        name="race",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    job.state.next_run_at_ms = max(1, int(time.time() * 1000) - 1_000)
+    service._save_store()
+
+    await service._on_timer()
+    await service._on_timer()
+
+    assert calls == [job.id]
+    loaded = service.get_job(job.id)
+    assert loaded is not None
+    assert loaded.state.last_run_at_ms is not None
+    assert loaded.state.next_run_at_ms is not None
+    assert loaded.state.next_run_at_ms > loaded.state.last_run_at_ms
+
+
+# ── update_job tests ──
+
+
+def test_update_job_changes_name(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="old name",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    result = service.update_job(job.id, name="new name")
+    assert isinstance(result, CronJob)
+    assert result.name == "new name"
+    assert result.payload.message == "hello"
+
+
+def test_update_job_changes_schedule(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="sched",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    old_next = job.state.next_run_at_ms
+
+    new_sched = CronSchedule(kind="every", every_ms=120_000)
+    result = service.update_job(job.id, schedule=new_sched)
+    assert isinstance(result, CronJob)
+    assert result.schedule.every_ms == 120_000
+    assert result.state.next_run_at_ms != old_next
+
+
+def test_update_job_changes_message(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="msg",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="old message",
+    )
+    result = service.update_job(job.id, message="new message")
+    assert isinstance(result, CronJob)
+    assert result.payload.message == "new message"
+
+
+def test_update_job_changes_cron_expression(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="cron-job",
+        schedule=CronSchedule(kind="cron", expr="0 9 * * *", tz="UTC"),
+        message="hello",
+    )
+    result = service.update_job(
+        job.id,
+        schedule=CronSchedule(kind="cron", expr="0 18 * * *", tz="UTC"),
+    )
+    assert isinstance(result, CronJob)
+    assert result.schedule.expr == "0 18 * * *"
+    assert result.state.next_run_at_ms is not None
+
+
+def test_update_job_not_found(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    result = service.update_job("nonexistent", name="x")
+    assert result == "not_found"
+
+
+def test_update_job_rejects_system_job(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    service.register_system_job(CronJob(
+        id="dream",
+        name="dream",
+        schedule=CronSchedule(kind="cron", expr="0 */2 * * *", tz="UTC"),
+        payload=CronPayload(kind="system_event"),
+    ))
+    result = service.update_job("dream", name="hacked")
+    assert result == "protected"
+    assert service.get_job("dream").name == "dream"
+
+
+def test_update_job_validates_schedule(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="validate",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    with pytest.raises(ValueError, match="unknown timezone"):
+        service.update_job(
+            job.id,
+            schedule=CronSchedule(kind="cron", expr="0 9 * * *", tz="Bad/Zone"),
+        )
+
+
+@pytest.mark.asyncio
+async def test_update_job_preserves_run_history(tmp_path) -> None:
+    import asyncio
+    store_path = tmp_path / "cron" / "jobs.json"
+    service = CronService(store_path, on_job=lambda _: asyncio.sleep(0))
+    job = service.add_job(
+        name="hist",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    await service.run_job(job.id)
+
+    result = service.update_job(job.id, name="renamed")
+    assert isinstance(result, CronJob)
+    assert len(result.state.run_history) == 1
+    assert result.state.run_history[0].status == "ok"
+
+
+def test_update_job_offline_writes_action(tmp_path) -> None:
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="offline",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+    )
+    service.update_job(job.id, name="updated-offline")
+
+    action_path = tmp_path / "cron" / "action.jsonl"
+    assert action_path.exists()
+    lines = [l for l in action_path.read_text().strip().split("\n") if l]
+    last = json.loads(lines[-1])
+    assert last["action"] == "update"
+    assert last["params"]["name"] == "updated-offline"
+
+
+def test_update_job_sentinel_channel_and_to(tmp_path) -> None:
+    """Passing None clears channel/to; omitting leaves them unchanged."""
+    service = CronService(tmp_path / "cron" / "jobs.json")
+    job = service.add_job(
+        name="sentinel",
+        schedule=CronSchedule(kind="every", every_ms=60_000),
+        message="hello",
+        channel="telegram",
+        to="user123",
+    )
+    assert job.payload.channel == "telegram"
+    assert job.payload.to == "user123"
+
+    result = service.update_job(job.id, name="renamed")
+    assert isinstance(result, CronJob)
+    assert result.payload.channel == "telegram"
+    assert result.payload.to == "user123"
+
+    result = service.update_job(job.id, channel=None, to=None)
+    assert isinstance(result, CronJob)
+    assert result.payload.channel is None
+    assert result.payload.to is None
+
+
+@pytest.mark.asyncio
+async def test_list_jobs_during_on_job_does_not_cause_stale_reload(tmp_path) -> None:
+    """Regression: if the bot calls list_jobs (which reloads from disk) during
+    on_job execution, the in-memory next_run_at_ms update must not be lost.
+    Previously this caused an infinite re-trigger loop."""
+    store_path = tmp_path / "cron" / "jobs.json"
+    execution_count = 0
+
+    async def on_job_that_lists(job):
+        nonlocal execution_count
+        execution_count += 1
+        # Simulate the bot calling cron(action=list) mid-execution
+        service.list_jobs()
+
+    service = CronService(store_path, on_job=on_job_that_lists, max_sleep_ms=100)
+    await service.start()
+
+    # Add two jobs scheduled in the past so they're immediately due
+    now_ms = int(time.time() * 1000)
+    for name in ("job-a", "job-b"):
+        service.add_job(
+            name=name,
+            schedule=CronSchedule(kind="every", every_ms=3_600_000),
+            message="test",
+        )
+    # Force next_run to the past so _on_timer picks them up
+    for job in service._store.jobs:
+        job.state.next_run_at_ms = now_ms - 1000
+    service._save_store()
+    service._arm_timer()
+
+    # Let the timer fire once
+    await asyncio.sleep(0.3)
+    service.stop()
+
+    # Each job should have run exactly once, not looped
+    assert execution_count == 2
+
+    # Verify next_run_at_ms was persisted correctly (in the future)
+    raw = json.loads(store_path.read_text())
+    for j in raw["jobs"]:
+        next_run = j["state"]["nextRunAtMs"]
+        assert next_run is not None
+        assert next_run > now_ms, f"Job '{j['name']}' next_run should be in the future"

tests/cron/test_cron_tool_list.py

@@ -2,9 +2,12 @@
 
 from datetime import datetime, timezone
 
+import pytest
+
 from nanobot.agent.tools.cron import CronTool
 from nanobot.cron.service import CronService
 from nanobot.cron.types import CronJob, CronJobState, CronPayload, CronSchedule
+from tests.test_openai_api import pytest_plugins
 
 
 def _make_tool(tmp_path) -> CronTool:
@@ -215,8 +218,10 @@ def test_list_at_job_shows_iso_timestamp(tmp_path) -> None:
     assert "Asia/Shanghai" in result
 
 
-def test_list_shows_last_run_state(tmp_path) -> None:
+@pytest.mark.asyncio
+async def test_list_shows_last_run_state(tmp_path) -> None:
     tool = _make_tool(tmp_path)
+    tool._cron._running = True
     job = tool._cron.add_job(
         name="Stateful job",
         schedule=CronSchedule(kind="cron", expr="0 9 * * *", tz="UTC"),
@@ -232,9 +237,10 @@ def test_list_shows_last_run_state(tmp_path) -> None:
     assert "ok" in result
     assert "(UTC)" in result
 
-
-def test_list_shows_error_message(tmp_path) -> None:
+@pytest.mark.asyncio
+async def test_list_shows_error_message(tmp_path) -> None:
     tool = _make_tool(tmp_path)
+    tool._cron._running = True
     job = tool._cron.add_job(
         name="Failed job",
         schedule=CronSchedule(kind="cron", expr="0 9 * * *", tz="UTC"),

tests/providers/test_anthropic_thinking.py

@@ -0,0 +1,65 @@
+"""Tests for Anthropic provider thinking / reasoning_effort modes."""
+
+from __future__ import annotations
+
+from unittest.mock import patch
+
+from nanobot.providers.anthropic_provider import AnthropicProvider
+
+
+def _make_provider(model: str = "claude-sonnet-4-6") -> AnthropicProvider:
+    with patch("anthropic.AsyncAnthropic"):
+        return AnthropicProvider(api_key="sk-test", default_model=model)
+
+
+def _build(provider: AnthropicProvider, reasoning_effort: str | None, **overrides):
+    defaults = dict(
+        messages=[{"role": "user", "content": "hello"}],
+        tools=None,
+        model=None,
+        max_tokens=4096,
+        temperature=0.7,
+        reasoning_effort=reasoning_effort,
+        tool_choice=None,
+        supports_caching=False,
+    )
+    defaults.update(overrides)
+    return provider._build_kwargs(**defaults)
+
+
+def test_adaptive_sets_type_adaptive() -> None:
+    kw = _build(_make_provider(), "adaptive")
+    assert kw["thinking"] == {"type": "adaptive"}
+
+
+def test_adaptive_forces_temperature_one() -> None:
+    kw = _build(_make_provider(), "adaptive")
+    assert kw["temperature"] == 1.0
+
+
+def test_adaptive_does_not_inflate_max_tokens() -> None:
+    kw = _build(_make_provider(), "adaptive", max_tokens=2048)
+    assert kw["max_tokens"] == 2048
+
+
+def test_adaptive_no_budget_tokens() -> None:
+    kw = _build(_make_provider(), "adaptive")
+    assert "budget_tokens" not in kw["thinking"]
+
+
+def test_high_uses_enabled_with_budget() -> None:
+    kw = _build(_make_provider(), "high", max_tokens=4096)
+    assert kw["thinking"]["type"] == "enabled"
+    assert kw["thinking"]["budget_tokens"] == max(8192, 4096)
+    assert kw["max_tokens"] >= kw["thinking"]["budget_tokens"] + 4096
+
+
+def test_low_uses_small_budget() -> None:
+    kw = _build(_make_provider(), "low")
+    assert kw["thinking"] == {"type": "enabled", "budget_tokens": 1024}
+
+
+def test_none_does_not_enable_thinking() -> None:
+    kw = _build(_make_provider(), None)
+    assert "thinking" not in kw
+    assert kw["temperature"] == 0.7

tests/providers/test_custom_provider.py

@@ -4,6 +4,7 @@ from types import SimpleNamespace
 from unittest.mock import patch
 
 from nanobot.providers.openai_compat_provider import OpenAICompatProvider
+from nanobot.providers.registry import find_by_name
 
 
 def test_custom_provider_parse_handles_empty_choices() -> None:
@@ -53,3 +54,20 @@ def test_custom_provider_parse_chunks_accepts_plain_text_chunks() -> None:
 
     assert result.finish_reason == "stop"
     assert result.content == "hello world"
+
+
+def test_local_provider_502_error_includes_reachability_hint() -> None:
+    spec = find_by_name("ollama")
+    with patch("nanobot.providers.openai_compat_provider.AsyncOpenAI"):
+        provider = OpenAICompatProvider(api_base="http://localhost:11434/v1", spec=spec)
+
+    result = provider._handle_error(
+        Exception("Error code: 502"),
+        spec=spec,
+        api_base="http://localhost:11434/v1",
+    )
+
+    assert result.finish_reason == "error"
+    assert "local model endpoint" in result.content
+    assert "http://localhost:11434/v1" in result.content
+    assert "proxy/tunnel" in result.content

tests/providers/test_enforce_role_alternation.py

@@ -0,0 +1,197 @@
+"""Tests for LLMProvider._enforce_role_alternation."""
+
+from nanobot.providers.base import LLMProvider
+
+
+class TestEnforceRoleAlternation:
+    """Verify trailing-assistant removal and consecutive same-role merging."""
+
+    def test_empty_messages(self):
+        assert LLMProvider._enforce_role_alternation([]) == []
+
+    def test_no_change_needed(self):
+        msgs = [
+            {"role": "system", "content": "You are helpful."},
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "Hello!"},
+            {"role": "user", "content": "Bye"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 4
+        assert result[-1]["role"] == "user"
+
+    def test_trailing_assistant_removed(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "Hello!"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 1
+        assert result[0]["role"] == "user"
+
+    def test_multiple_trailing_assistants_removed(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "A"},
+            {"role": "assistant", "content": "B"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 1
+        assert result[0]["role"] == "user"
+
+    def test_consecutive_user_messages_merged(self):
+        msgs = [
+            {"role": "user", "content": "Hello"},
+            {"role": "user", "content": "How are you?"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 1
+        assert "Hello" in result[0]["content"]
+        assert "How are you?" in result[0]["content"]
+
+    def test_consecutive_assistant_messages_merged(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "Hello!"},
+            {"role": "assistant", "content": "How can I help?"},
+            {"role": "user", "content": "Thanks"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 3
+        assert "Hello!" in result[1]["content"]
+        assert "How can I help?" in result[1]["content"]
+
+    def test_system_messages_not_merged(self):
+        msgs = [
+            {"role": "system", "content": "System A"},
+            {"role": "system", "content": "System B"},
+            {"role": "user", "content": "Hi"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 3
+        assert result[0]["content"] == "System A"
+        assert result[1]["content"] == "System B"
+
+    def test_tool_messages_not_merged(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": None, "tool_calls": [{"id": "1"}]},
+            {"role": "tool", "content": "result1", "tool_call_id": "1"},
+            {"role": "tool", "content": "result2", "tool_call_id": "2"},
+            {"role": "user", "content": "Next"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        tool_msgs = [m for m in result if m["role"] == "tool"]
+        assert len(tool_msgs) == 2
+
+    def test_consecutive_assistant_keeps_later_tool_call_message(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "Previous reply"},
+            {"role": "assistant", "content": None, "tool_calls": [{"id": "1"}]},
+            {"role": "tool", "content": "result1", "tool_call_id": "1"},
+            {"role": "user", "content": "Next"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert result[1]["role"] == "assistant"
+        assert result[1]["tool_calls"] == [{"id": "1"}]
+        assert result[1]["content"] is None
+        assert result[2]["role"] == "tool"
+
+    def test_consecutive_assistant_does_not_overwrite_existing_tool_call_message(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": None, "tool_calls": [{"id": "1"}]},
+            {"role": "assistant", "content": "Later plain assistant"},
+            {"role": "tool", "content": "result1", "tool_call_id": "1"},
+            {"role": "user", "content": "Next"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert result[1]["role"] == "assistant"
+        assert result[1]["tool_calls"] == [{"id": "1"}]
+        assert result[1]["content"] is None
+        assert result[2]["role"] == "tool"
+
+    def test_non_string_content_uses_latest(self):
+        msgs = [
+            {"role": "user", "content": [{"type": "text", "text": "A"}]},
+            {"role": "user", "content": "B"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 1
+        assert result[0]["content"] == "B"
+
+    def test_original_messages_not_mutated(self):
+        msgs = [
+            {"role": "user", "content": "Hello"},
+            {"role": "user", "content": "World"},
+        ]
+        original_first = dict(msgs[0])
+        LLMProvider._enforce_role_alternation(msgs)
+        assert msgs[0] == original_first
+        assert len(msgs) == 2
+
+    def test_trailing_assistant_recovered_as_user_when_only_system_remains(self):
+        """Subagent result injected as assistant message must not be silently dropped.
+
+        When build_messages(current_role="assistant") produces [system, assistant],
+        _enforce_role_alternation would drop the assistant, leaving only [system].
+        Most providers (e.g. Zhipu/GLM error 1214) reject such requests.
+        The trailing assistant should be recovered as a user message instead.
+        """
+        msgs = [
+            {"role": "system", "content": "You are helpful."},
+            {"role": "assistant", "content": "Subagent completed successfully."},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 2
+        assert result[0]["role"] == "system"
+        assert result[1]["role"] == "user"
+        assert "Subagent completed successfully." in result[1]["content"]
+
+    def test_trailing_assistant_not_recovered_when_user_message_present(self):
+        """Recovery should NOT happen when a user message already exists."""
+        msgs = [
+            {"role": "system", "content": "You are helpful."},
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "Hello!"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 2
+        assert result[-1]["role"] == "user"
+
+    def test_trailing_assistant_recovered_with_tool_result_preceding(self):
+        """When only [system, tool, assistant] remains, recovery is not needed
+        because tool messages are valid non-system content."""
+        msgs = [
+            {"role": "system", "content": "You are helpful."},
+            {"role": "tool", "content": "result", "tool_call_id": "1"},
+            {"role": "assistant", "content": "Done."},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 2
+        assert result[-1]["role"] == "tool"
+
+    def test_only_assistant_messages(self):
+        msgs = [
+            {"role": "assistant", "content": "A"},
+            {"role": "assistant", "content": "B"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert result == []
+
+    def test_realistic_conversation(self):
+        msgs = [
+            {"role": "system", "content": "You are helpful."},
+            {"role": "user", "content": "What is 2+2?"},
+            {"role": "assistant", "content": "4"},
+            {"role": "user", "content": "And 3+3?"},
+            {"role": "user", "content": "(please be quick)"},
+            {"role": "assistant", "content": "6"},
+        ]
+        result = LLMProvider._enforce_role_alternation(msgs)
+        assert len(result) == 4
+        assert result[2]["role"] == "assistant"
+        assert result[3]["role"] == "user"
+        assert "And 3+3?" in result[3]["content"]
+        assert "(please be quick)" in result[3]["content"]