fix(agent): align LLM wall timeout with sustained goals for main + subagents

Centralize runner_wall_llm_timeout_s in session goal_state metadata helpers so
spawned subagents inherit the same policy as AgentLoop without coupling to
long_task. Pass optional resolver into SubagentManager and add tests.

Co-authored-by: Cursor <cursoragent@cursor.com>

nanobot/agent/loop.py

@@ -35,7 +35,7 @@ from nanobot.providers.factory import ProviderSnapshot
 from nanobot.session.goal_state import (
     goal_state_runtime_lines,
     goal_state_ws_blob,
-    sustained_goal_active,
+    runner_wall_llm_timeout_s,
 )
 from nanobot.session.manager import Session, SessionManager
 from nanobot.utils.artifacts import generated_image_paths_from_messages
@@ -253,6 +253,7 @@ class AgentLoop:
             restrict_to_workspace=restrict_to_workspace,
             disabled_skills=disabled_skills,
             max_iterations=self.max_iterations,
+            llm_wall_timeout_for_session=lambda sk: runner_wall_llm_timeout_s(self.sessions, sk),
         )
         self._unified_session = unified_session
         self._max_messages = max_messages if max_messages > 0 else 120
@@ -795,10 +796,10 @@ class AgentLoop:
                 injection_callback=_drain_pending,
                 # Sustained goals may legitimately exceed NANOBOT_LLM_TIMEOUT_S; idle stall
                 # is still capped by NANOBOT_STREAM_IDLE_TIMEOUT_S in streaming providers.
-                llm_timeout_s=(
-                    0.0
-                    if session is not None and sustained_goal_active(session.metadata)
-                    else None
+                llm_timeout_s=runner_wall_llm_timeout_s(
+                    self.sessions,
+                    session.key if session is not None else session_key,
+                    metadata=(session.metadata if session is not None else None),
                 ),
             ))
         finally:

nanobot/agent/subagent.py

@@ -6,7 +6,7 @@ import time
 import uuid
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import Any
+from typing import Any, Callable
 
 from loguru import logger
 
@@ -79,6 +79,7 @@ class SubagentManager:
         restrict_to_workspace: bool = False,
         disabled_skills: list[str] | None = None,
         max_iterations: int | None = None,
+        llm_wall_timeout_for_session: Callable[[str | None], float | None] | None = None,
     ):
         defaults = AgentDefaults()
         self.provider = provider
@@ -96,6 +97,7 @@ class SubagentManager:
         )
         self.max_concurrent_subagents = defaults.max_concurrent_subagents
         self.runner = AgentRunner(provider)
+        self._llm_wall_timeout_for_session = llm_wall_timeout_for_session
         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, ...}
@@ -196,6 +198,12 @@ class SubagentManager:
                 {"role": "user", "content": task},
             ]
 
+            sess_key = origin.get("session_key")
+            llm_timeout = (
+                self._llm_wall_timeout_for_session(sess_key)
+                if self._llm_wall_timeout_for_session
+                else None
+            )
             result = await self.runner.run(AgentRunSpec(
                 initial_messages=messages,
                 tools=tools,
@@ -207,6 +215,8 @@ class SubagentManager:
                 error_message=None,
                 fail_on_tool_error=True,
                 checkpoint_callback=_on_checkpoint,
+                session_key=sess_key,
+                llm_timeout_s=llm_timeout,
             ))
             status.phase = "done"
             status.stop_reason = result.stop_reason

nanobot/session/goal_state.py

@@ -1,8 +1,8 @@
 """Session metadata helpers for sustained goals (e.g. ``long_task`` / ``complete_goal``).
 
 Tools set ``metadata[GOAL_STATE_KEY]``. Reads accept the legacy session key ``thread_goal``
-for older sessions. The agent uses ``goal_state_runtime_lines`` and
-``goal_state_ws_blob`` without importing tool implementations.
+for older sessions. Callers use ``goal_state_runtime_lines``, ``goal_state_ws_blob``, and
+``runner_wall_llm_timeout_s`` without importing tool implementations.
 """
 
 from __future__ import annotations
@@ -10,6 +10,8 @@ from __future__ import annotations
 import json
 from typing import Any, Mapping, MutableMapping
 
+from nanobot.session.manager import SessionManager
+
 GOAL_STATE_KEY = "goal_state"
 # Older builds stored the same JSON blob under this key.
 _LEGACY_GOAL_STATE_SESSION_KEY = "thread_goal"
@@ -89,3 +91,21 @@ def goal_state_ws_blob(metadata: Mapping[str, Any] | None) -> dict[str, Any]:
             blob["objective"] = objective
         return blob
     return {"active": False}
+
+
+def runner_wall_llm_timeout_s(
+    sessions: SessionManager,
+    session_key: str | None,
+    *,
+    metadata: Mapping[str, Any] | None = None,
+) -> float | None:
+    """Wall-clock cap for :class:`~nanobot.agent.runner.AgentRunner` when streaming an LLM.
+
+    Returns ``0.0`` to disable ``asyncio.wait_for`` around the request when a sustained goal is
+    active; ``None`` means use ``NANOBOT_LLM_TIMEOUT_S``. Pass in-memory ``metadata`` when the
+    caller already holds :attr:`~nanobot.session.manager.Session.metadata` for this turn.
+    """
+    meta: Mapping[str, Any] | None = metadata
+    if meta is None and session_key:
+        meta = sessions.get_or_create(session_key).metadata
+    return 0.0 if sustained_goal_active(meta) else None

tests/agent/test_loop_goal_wall_timeout.py

@@ -0,0 +1,46 @@
+"""Subagent forwards loop-provided LLM wall-timeout resolver into AgentRunSpec."""
+
+from pathlib import Path
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+
+from nanobot.agent.runner import AgentRunResult
+from nanobot.agent.subagent import SubagentManager, SubagentStatus
+from nanobot.bus.queue import MessageBus
+
+
+@pytest.mark.asyncio
+async def test_subagent_forwards_resolver_to_agent_run_spec(tmp_path: Path) -> None:
+    provider = MagicMock()
+    provider.get_default_model.return_value = "m"
+    mgr = SubagentManager(
+        provider=provider,
+        workspace=tmp_path,
+        bus=MessageBus(),
+        max_tool_result_chars=64,
+        llm_wall_timeout_for_session=lambda sk: 0.0 if sk == "cli:direct" else None,
+    )
+
+    mgr.runner.run = AsyncMock(
+        return_value=AgentRunResult(final_content="ok", messages=[], stop_reason="completed")
+    )
+    mgr._announce_result = AsyncMock()
+
+    status = SubagentStatus(
+        task_id="t1",
+        label="lbl",
+        task_description="task",
+        started_at=0.0,
+    )
+    await mgr._run_subagent(
+        "t1",
+        "task",
+        "lbl",
+        {"channel": "cli", "chat_id": "direct", "session_key": "cli:direct"},
+        status,
+    )
+    mgr.runner.run.assert_called_once()
+    spec = mgr.runner.run.call_args[0][0]
+    assert spec.session_key == "cli:direct"
+    assert spec.llm_timeout_s == 0.0

tests/session/test_goal_state.py

@@ -8,8 +8,10 @@ from nanobot.session.goal_state import (
     goal_state_runtime_lines,
     goal_state_ws_blob,
     parse_goal_state,
+    runner_wall_llm_timeout_s,
     sustained_goal_active,
 )
+from nanobot.session.manager import SessionManager
 
 
 def test_runtime_lines_empty_when_no_metadata():
@@ -105,3 +107,25 @@ def test_sustained_goal_active_true_when_active():
 def test_sustained_goal_active_respects_legacy_thread_goal_key():
     meta = {"thread_goal": {"status": "active", "objective": "Legacy."}}
     assert sustained_goal_active(meta) is True
+
+
+def test_runner_wall_llm_timeout_uses_metadata_override(tmp_path):
+    sm = SessionManager(tmp_path)
+    assert (
+        runner_wall_llm_timeout_s(
+            sm,
+            "cli:test",
+            metadata={GOAL_STATE_KEY: {"status": "active", "objective": "x"}},
+        )
+        == 0.0
+    )
+    assert runner_wall_llm_timeout_s(sm, "cli:test", metadata={}) is None
+
+
+def test_runner_wall_llm_timeout_reads_session_when_metadata_missing(tmp_path):
+    sm = SessionManager(tmp_path)
+    sess = sm.get_or_create("c:d")
+    sess.metadata = {GOAL_STATE_KEY: {"status": "active", "objective": "z"}}
+    assert runner_wall_llm_timeout_s(sm, "c:d") == 0.0
+    sess.metadata = {}
+    assert runner_wall_llm_timeout_s(sm, "c:d") is None