refactor(agent): unify runner request fitting (#5612)

* refactor(agent): unify runner request fitting

* fix(agent): fit every runner model request

* fix(agent): count resumed state during request fitting

* refactor(agent): consolidate request fitting state

nanobot/agent/context_governance.py

@@ -13,6 +13,7 @@ from typing import TYPE_CHECKING, Any, cast
 
 from loguru import logger
 
+from nanobot.providers.base import LLMUsage
 from nanobot.utils.helpers import (
     estimate_message_tokens,
     estimate_prompt_tokens_chain,
@@ -27,12 +28,6 @@ if TYPE_CHECKING:
     from nanobot.providers.base import LLMProvider
 
 SNIP_SAFETY_BUFFER = 1024
-MICROCOMPACT_MIN_CHARS = 500
-INFLIGHT_COMPACT_TARGET_RATIO = 0.85
-COMPACTABLE_TOOLS = frozenset({
-    "read_file", "exec", "grep", "find_files",
-    "web_search", "web_fetch", "list_dir", "list_exec_sessions",
-})
 # read_file is the recovery path for persisted results; exempting it prevents persist->read->persist loops.
 TOOL_RESULT_OFFLOAD_EXEMPT_TOOLS = frozenset({"read_file"})
 BACKFILL_CONTENT = "[Tool result unavailable — call was interrupted or lost]"
@@ -41,6 +36,27 @@ PLACEHOLDER_TEXTS = frozenset({
 })
 
 
+class ContextWindowExceededError(RuntimeError):
+    """Raised before a locally fitted request that still exceeds its budget."""
+
+    def __init__(
+        self,
+        *,
+        session_key: str | None,
+        estimated_tokens: int,
+        input_budget: int,
+        source: str,
+    ) -> None:
+        self.session_key = session_key
+        self.estimated_tokens = estimated_tokens
+        self.input_budget = input_budget
+        self.source = source
+        super().__init__(
+            "Model input still exceeds the local context budget after request fitting "
+            f"for {session_key or 'default'}: {estimated_tokens}/{input_budget} via {source}"
+        )
+
+
 def _tool_call_name_is_valid(tool_call: Any) -> bool:
     """Whether a persisted OpenAI-style tool_call carries a usable name.
 
@@ -67,7 +83,6 @@ class ContextGovernanceConfig:
     context_window_tokens: int | None = None
     context_block_limit: int | None = None
     max_tokens: int | None = None
-    inflight_start_index: int = 0
 
 
 class ContextGovernor:
@@ -77,17 +92,85 @@ class ContextGovernor:
         self,
         config: ContextGovernanceConfig,
         messages: list[dict[str, Any]],
-        compacted_tool_call_ids: set[str],
     ) -> list[dict[str, Any]]:
         updated = self.strip_placeholder_assistant_messages(messages)
         updated = self.strip_malformed_tool_calls(updated)
         updated = self.drop_orphan_tool_results(updated)
         updated = self.backfill_missing_tool_results(updated)
-        updated = self.apply_tool_result_budget(config, updated)
+        return self.apply_tool_result_budget(config, updated)
-        updated = self.compact_inflight_overflow(config, updated, compacted_tool_call_ids)
+
-        updated = self.snip_history(config, updated)
+    def fit_to_budget(
+        self,
+        config: ContextGovernanceConfig,
+        messages: list[dict[str, Any]],
+        *,
+        tool_definitions: list[dict[str, Any]] | None,
+    ) -> list[dict[str, Any]]:
+        """Fit a model-facing copy while keeping the source transcript intact."""
+        updated = self.snip_history(
+            config,
+            messages,
+            tool_definitions=tool_definitions,
+            force=True,
+        )
         updated = self.drop_orphan_tool_results(updated)
-        return self.backfill_missing_tool_results(updated)
+        updated = self.backfill_missing_tool_results(updated)
+        if not config.context_window_tokens:
+            return updated
+        budget = self.input_budget(config)
+        estimated, source = estimate_prompt_tokens_chain(
+            config.provider,
+            config.model,
+            updated,
+            tool_definitions,
+        )
+        if budget > 0 and estimated <= budget:
+            return updated
+        raise ContextWindowExceededError(
+            session_key=config.session_key,
+            estimated_tokens=estimated,
+            input_budget=budget,
+            source=source,
+        )
+
+    def fit_request(
+        self,
+        config: ContextGovernanceConfig,
+        messages: list[dict[str, Any]],
+        usage: LLMUsage | None,
+        *,
+        usage_matches_messages: bool,
+        tool_definitions: list[dict[str, Any]] | None,
+        request_context_tokens: int | None = None,
+    ) -> tuple[list[dict[str, Any]], bool]:
+        """Fit the request when its measured or estimated input is pressured."""
+        if not config.context_window_tokens:
+            return messages, False
+        budget = self.input_budget(config)
+        if (
+            request_context_tokens is None
+            and usage_matches_messages
+            and usage is not None
+            and usage.context_tokens is not None
+        ):
+            pressured = budget <= 0 or usage.context_tokens >= budget
+        else:
+            estimated, _ = estimate_prompt_tokens_chain(
+                config.provider,
+                config.model,
+                messages,
+                tool_definitions,
+            )
+            if request_context_tokens is not None:
+                estimated = max(estimated, request_context_tokens)
+            pressured = budget <= 0 or estimated >= budget
+        if not pressured:
+            return messages, False
+        return self.fit_to_budget(
+            config,
+            messages,
+            tool_definitions=tool_definitions,
+        ), True
 
     @staticmethod
     def input_budget(config: ContextGovernanceConfig) -> int:
@@ -326,71 +409,13 @@ class ContextGovernor:
                 updated[idx]["content"] = normalized
         return updated
 
-    def compact_inflight_overflow(
-        self,
-        config: ContextGovernanceConfig,
-        messages: list[dict[str, Any]],
-        compacted_tool_call_ids: set[str],
-    ) -> list[dict[str, Any]]:
-        """Compact in-flight tool results only when the request would overflow."""
-        budget = self.input_budget(config)
-        if budget <= 0:
-            return messages
-
-        tools = config.tools.get_definitions()
-        updated = self._apply_recorded_compactions(messages, compacted_tool_call_ids)
-        estimate, source = estimate_prompt_tokens_chain(
-            config.provider,
-            config.model,
-            updated,
-            tools,
-        )
-        if estimate <= budget:
-            return updated
-
-        target = int(budget * INFLIGHT_COMPACT_TARGET_RATIO)
-        candidates = self._inflight_compaction_candidates(
-            config,
-            updated,
-            compacted_tool_call_ids,
-        )
-        if not candidates:
-            return updated
-
-        for candidate_idx, (idx, tool_call_id) in enumerate(candidates):
-            is_newest_candidate = candidate_idx == len(candidates) - 1
-            if is_newest_candidate and estimate <= budget:
-                break
-            if tool_call_id in compacted_tool_call_ids:
-                continue
-            if updated is messages:
-                updated = [dict(m) for m in messages]
-            compacted_tool_call_ids.add(tool_call_id)
-            self._compact_tool_result_at(updated, idx)
-            estimate, source = estimate_prompt_tokens_chain(
-                config.provider,
-                config.model,
-                updated,
-                tools,
-            )
-            if estimate <= target:
-                break
-
-        logger.debug(
-            "In-flight context compaction for {}: prompt={} budget={} target={} via {}, ids={}",
-            config.session_key or "default",
-            estimate,
-            budget,
-            target,
-            source,
-            len(compacted_tool_call_ids),
-        )
-        return updated
-
     def snip_history(
         self,
         config: ContextGovernanceConfig,
         messages: list[dict[str, Any]],
+        *,
+        tool_definitions: list[dict[str, Any]] | None,
+        force: bool = False,
     ) -> list[dict[str, Any]]:
         if not messages or not config.context_window_tokens:
             return messages
@@ -399,14 +424,13 @@ class ContextGovernor:
         if budget <= 0:
             return messages
 
-        tools = config.tools.get_definitions()
         estimate, _ = estimate_prompt_tokens_chain(
             config.provider,
             config.model,
             messages,
-            tools,
+            tool_definitions,
         )
-        if estimate <= budget:
+        if not force and estimate <= budget:
             return messages
 
         system_messages = [dict(msg) for msg in messages if msg.get("role") == "system"]
@@ -419,7 +443,7 @@ class ContextGovernor:
             config.provider,
             config.model,
             system_messages,
-            tools,
+            tool_definitions,
         )
         remaining_budget = max(0, budget - max(system_tokens, fixed_tokens))
         kept: list[dict[str, Any]] = []
@@ -434,16 +458,6 @@ class ContextGovernor:
 
         return system_messages + self._legal_history_tail(kept, non_system)
 
-    @staticmethod
-    def _tool_result_compaction_message(message: dict[str, Any]) -> str:
-        name = message.get("name", "tool")
-        return (
-            f"Error: The previous {name} result was compacted to fit context because it was too "
-            "large. Do not repeat the same call unchanged. Retry with a narrower path, query, "
-            "range, or result limit, use another tool, or tell the user the task cannot fit in "
-            "the available context."
-        )
-
     def _legal_history_tail(
         self,
         kept: list[dict[str, Any]],
@@ -462,50 +476,3 @@ class ContextGovernor:
             if messages[idx].get("role") == "user":
                 return messages[idx:]
         return []
-
-    def _apply_recorded_compactions(
-        self,
-        messages: list[dict[str, Any]],
-        compacted_tool_call_ids: set[str],
-    ) -> list[dict[str, Any]]:
-        if not compacted_tool_call_ids:
-            return messages
-        updated = messages
-        for idx, msg in enumerate(messages):
-            if msg.get("role") != "tool":
-                continue
-            tool_call_id = msg.get("tool_call_id")
-            if not tool_call_id or str(tool_call_id) not in compacted_tool_call_ids:
-                continue
-            compaction_message = self._tool_result_compaction_message(msg)
-            if msg.get("content") == compaction_message:
-                continue
-            if updated is messages:
-                updated = [dict(m) for m in messages]
-            updated[idx]["content"] = compaction_message
-        return updated
-
-    def _inflight_compaction_candidates(
-        self,
-        config: ContextGovernanceConfig,
-        messages: list[dict[str, Any]],
-        compacted_tool_call_ids: set[str],
-    ) -> list[tuple[int, str]]:
-        compactable: list[tuple[int, str]] = []
-        for idx, msg in enumerate(messages):
-            if idx < config.inflight_start_index:
-                continue
-            if msg.get("role") != "tool" or msg.get("name") not in COMPACTABLE_TOOLS:
-                continue
-            tool_call_id = msg.get("tool_call_id")
-            if not tool_call_id or str(tool_call_id) in compacted_tool_call_ids:
-                continue
-            content = msg.get("content")
-            if not isinstance(content, str) or len(content) < MICROCOMPACT_MIN_CHARS:
-                continue
-            compactable.append((idx, str(tool_call_id)))
-
-        return compactable
-
-    def _compact_tool_result_at(self, messages: list[dict[str, Any]], idx: int) -> None:
-        messages[idx]["content"] = self._tool_result_compaction_message(messages[idx])

nanobot/agent/runner.py

@@ -139,6 +139,17 @@ class AgentRunResult:
     provider_state: ProviderConversationState | None = field(default=None, repr=False)
 
 
+@dataclass(slots=True)
+class _ModelRequestState:
+    """Per-run state used to govern the next provider request."""
+
+    config: ContextGovernanceConfig
+    conversation: ProviderConversationStateController
+    usage: LLMUsage | None = None
+    messages: list[dict[str, Any]] | None = None
+    tool_definitions: list[dict[str, Any]] | None = None
+
+
 class AgentRunner:
     """Run a tool-capable LLM loop without product-layer concerns."""
 
@@ -500,7 +511,6 @@ class AgentRunner:
         length_recovery_parts: list[str] = []
         had_injections = False
         injection_cycles = 0
-        compacted_tool_call_ids: set[str] = set()
         pending_stream_content: str | None = None
         conversation_state = ProviderConversationStateController(
             provider=spec.runtime.provider,
@@ -519,39 +529,29 @@ class AgentRunner:
             context_window_tokens=spec.runtime.context_window_tokens,
             context_block_limit=spec.context_block_limit,
             max_tokens=spec.runtime.generation.max_tokens,
-            inflight_start_index=len(messages),
+        )
+        request_state = _ModelRequestState(
+            config=governance_config,
+            conversation=conversation_state,
         )
 
         for iteration in range(spec.max_iterations):
-            # 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. A governance
-            # failure must stop the run instead of sending an ungoverned copy.
-            messages_for_model = self.context_governor.prepare_for_model(
-                governance_config,
-                messages,
-                compacted_tool_call_ids,
-            )
             context = AgentHookContext(
                 iteration=iteration,
                 messages=messages,
                 session_key=spec.session_key,
             )
             await hook.before_iteration(context)
-            provider_context = conversation_state.prepare_request(
-                messages,
-                context_window_tokens=spec.runtime.context_window_tokens,
-                model_messages=messages_for_model,
-            )
             response = await self._request_model(
                 spec,
-                messages_for_model,
+                messages,
                 hook,
                 context,
-                conversation_state=conversation_state,
+                request_state=request_state,
-                provider_context=provider_context,
+                transcript=messages,
             )
+            assert request_state.messages is not None
+            messages_for_model = request_state.messages
             conversation_state.observe_response(response, messages)
             context.response = response
             context.tool_calls = list(response.tool_calls)
@@ -563,7 +563,7 @@ class AgentRunner:
                 response.content,
             )
             response.content = cleaned_content
-            raw_usage = self._usage_or_estimate(spec, messages_for_model, response)
+            raw_usage = self._record_request_usage(spec, request_state, response)
             context.usage = raw_usage
             usage = self._merge_usage(usage, raw_usage)
             if reasoning_text and not context.streamed_reasoning:
@@ -637,7 +637,6 @@ class AgentRunner:
                     self.context_governor.prepare_for_model(
                         governance_config,
                         messages,
-                        compacted_tool_call_ids,
                     )
                     if response.provider_state is not None
                     else None
@@ -703,14 +702,13 @@ class AgentRunner:
                 )
                 if hook.wants_streaming():
                     await hook.on_stream_end(context, resuming=False)
-                retry_messages = self._finalization_retry_messages(messages_for_model)
                 response = await self._request_finalization_retry(
                     spec,
                     messages_for_model,
+                    request_state=request_state,
                     transcript=messages,
-                    conversation_state=conversation_state,
                 )
-                retry_usage = self._usage_or_estimate(spec, retry_messages, response)
+                retry_usage = self._record_request_usage(spec, request_state, response)
                 usage = self._merge_usage(usage, retry_usage)
                 raw_usage = self._merge_usage(raw_usage, retry_usage)
                 context.response = response
@@ -897,7 +895,7 @@ class AgentRunner:
                     hook,
                     messages,
                     usage,
-                    conversation_state,
+                    request_state=request_state,
                 )
             if terminal_content is None:
                 terminal_content = self._max_iterations_fallback(spec)
@@ -944,6 +942,60 @@ class AgentRunner:
         kwargs["reasoning_effort"] = generation.reasoning_effort
         return kwargs
 
+    def _prepare_model_request(
+        self,
+        state: _ModelRequestState,
+        messages: list[dict[str, Any]],
+        *,
+        tool_definitions: list[dict[str, Any]] | None,
+        transcript: list[dict[str, Any]] | None = None,
+    ) -> tuple[list[dict[str, Any]], ProviderCallContext | None]:
+        """Prepare, fit, and record the exact payload sent to a provider."""
+        prepared = self.context_governor.prepare_for_model(state.config, messages)
+        supplemental_messages = (
+            [prepared[-1]] if transcript is not None and tool_definitions is None else None
+        )
+        model_messages = None if supplemental_messages is not None else prepared
+        request_context_tokens = (
+            state.conversation.estimate_request_context_tokens(
+                transcript,
+                model_messages=model_messages,
+                supplemental_messages=supplemental_messages,
+                tool_definitions=tool_definitions,
+            )
+            if transcript is not None
+            else None
+        )
+        usage_matches_messages = (
+            state.messages is not None
+            and prepared == state.messages
+            and tool_definitions == state.tool_definitions
+        )
+        prepared, fitted = self.context_governor.fit_request(
+            state.config,
+            prepared,
+            state.usage,
+            usage_matches_messages=usage_matches_messages,
+            tool_definitions=tool_definitions,
+            request_context_tokens=request_context_tokens,
+        )
+        provider_context = (
+            state.conversation.prepare_request(
+                transcript,
+                context_window_tokens=state.config.context_window_tokens,
+                model_messages=model_messages,
+                supplemental_messages=supplemental_messages,
+                resume_state=not fitted,
+            )
+            if transcript is not None
+            else state.conversation.independent_request_context(
+                context_window_tokens=state.config.context_window_tokens,
+            )
+        )
+        state.messages = deepcopy(prepared)
+        state.tool_definitions = deepcopy(tool_definitions)
+        return prepared, provider_context
+
     async def _request_model(
         self,
         spec: AgentRunSpec,
@@ -951,16 +1003,23 @@ class AgentRunner:
         hook: AgentHook,
         context: AgentHookContext,
         *,
+        request_state: _ModelRequestState,
         malformed_retry: bool = False,
-        conversation_state: ProviderConversationStateController,
+        transcript: list[dict[str, Any]] | None,
-        provider_context: ProviderCallContext | None = None,
     ) -> LLMResponse:
         timeout_s = self._resolve_llm_timeout_s(spec)
+        tool_definitions = spec.tools.get_definitions()
+        messages, provider_context = self._prepare_model_request(
+            request_state,
+            messages,
+            tool_definitions=tool_definitions,
+            transcript=transcript,
+        )
 
         kwargs = self._build_request_kwargs(
             spec,
             messages,
-            tools=spec.tools.get_definitions(),
+            tools=tool_definitions,
         )
         wants_streaming = hook.wants_streaming()
 
@@ -1138,11 +1197,9 @@ class AgentRunner:
             )
             return await self._request_model(
                 spec, retry_messages, hook, context,
+                request_state=request_state,
                 malformed_retry=True,
-                conversation_state=conversation_state,
+                transcript=None,
-                provider_context=conversation_state.independent_request_context(
-                    context_window_tokens=spec.runtime.context_window_tokens,
-                ),
             )
         if (
             all_dropped
@@ -1158,9 +1215,7 @@ class AgentRunner:
             return await self._request_no_tools(
                 spec,
                 fallback_messages,
-                provider_context=conversation_state.independent_request_context(
+                request_state=request_state,
-                    context_window_tokens=spec.runtime.context_window_tokens,
-                ),
             )
         return response
 
@@ -1228,21 +1283,17 @@ class AgentRunner:
         spec: AgentRunSpec,
         messages: list[dict[str, Any]],
         *,
+        request_state: _ModelRequestState,
         transcript: list[dict[str, Any]],
-        conversation_state: ProviderConversationStateController,
     ) -> LLMResponse:
         retry_messages = self._finalization_retry_messages(messages)
-        provider_context = conversation_state.prepare_request(
-            transcript,
-            context_window_tokens=spec.runtime.context_window_tokens,
-            supplemental_messages=[retry_messages[-1]],
-        )
         response = await self._request_no_tools(
             spec,
             retry_messages,
-            provider_context=provider_context,
+            request_state=request_state,
+            transcript=transcript,
         )
-        conversation_state.observe_response(
+        request_state.conversation.observe_response(
             response,
             transcript,
             adopt_candidate_state=False,
@@ -1261,16 +1312,15 @@ class AgentRunner:
         hook: AgentHook,
         messages: list[dict[str, Any]],
         usage: LLMUsage | None,
-        conversation_state: ProviderConversationStateController,
+        *,
+        request_state: _ModelRequestState,
     ) -> tuple[str | None, LLMUsage | None]:
         retry_messages = self._budget_exhausted_finalization_messages(messages)
         try:
             response = await self._request_no_tools(
                 spec,
                 retry_messages,
-                provider_context=conversation_state.independent_request_context(
+                request_state=request_state,
-                    context_window_tokens=spec.runtime.context_window_tokens,
-                ),
             )
         except Exception:
             logger.exception(
@@ -1279,7 +1329,7 @@ class AgentRunner:
             )
             return None, usage
 
-        raw_usage = self._usage_or_estimate(spec, retry_messages, response)
+        raw_usage = self._record_request_usage(spec, request_state, response)
         usage = self._merge_usage(usage, raw_usage)
         if response.finish_reason == "error" or response.has_tool_calls:
             logger.warning(
@@ -1308,8 +1358,15 @@ class AgentRunner:
         spec: AgentRunSpec,
         messages: list[dict[str, Any]],
         *,
-        provider_context: ProviderCallContext | None = None,
+        request_state: _ModelRequestState,
+        transcript: list[dict[str, Any]] | None = None,
     ) -> LLMResponse:
+        messages, provider_context = self._prepare_model_request(
+            request_state,
+            messages,
+            tool_definitions=None,
+            transcript=transcript,
+        )
         kwargs = self._build_request_kwargs(
             spec,
             messages,
@@ -1321,17 +1378,18 @@ class AgentRunner:
         )
         timeout_s = self._resolve_llm_timeout_s(spec)
         try:
-            return (
+            response = (
                 await coro
                 if timeout_s is None
                 else await asyncio.wait_for(coro, timeout=timeout_s)
             )
         except asyncio.TimeoutError:
-            return LLMResponse(
+            response = LLMResponse(
                 content=f"Error calling LLM: timed out after {timeout_s:g}s",
                 finish_reason="error",
                 error_kind="timeout",
             )
+        return response
 
     @staticmethod
     def _resolve_llm_timeout_s(spec: AgentRunSpec) -> float | None:
@@ -1373,33 +1431,53 @@ class AgentRunner:
         spec: AgentRunSpec,
         messages: list[dict[str, Any]],
         response: LLMResponse,
+        *,
+        tool_definitions: list[dict[str, Any]] | None,
     ) -> LLMUsage | None:
         usage = response.usage
         if response.finish_reason == "error":
             if usage is None or usage.total_tokens == 0:
                 usage = LLMUsage.empty_request()
         elif usage is None or usage.total_tokens == 0:
-            usage = self._estimate_response_usage(spec, messages, response)
+            usage = self._estimate_response_usage(
+                spec,
+                messages,
+                response,
+                tool_definitions=tool_definitions,
+            )
         return usage.with_timing(
             generation_ms=response.generation_ms,
             ttft_ms=response.ttft_ms,
         )
 
+    def _record_request_usage(
+        self,
+        spec: AgentRunSpec,
+        state: _ModelRequestState,
+        response: LLMResponse,
+    ) -> LLMUsage | None:
+        assert state.messages is not None
+        state.usage = self._usage_or_estimate(
+            spec,
+            state.messages,
+            response,
+            tool_definitions=state.tool_definitions,
+        )
+        return state.usage
+
     def _estimate_response_usage(
         self,
         spec: AgentRunSpec,
         messages: list[dict[str, Any]],
         response: LLMResponse,
+        *,
+        tool_definitions: list[dict[str, Any]] | None,
     ) -> LLMUsage:
-        try:
-            tools = spec.tools.get_definitions()
-        except Exception:
-            tools = None
         prompt_tokens, _ = estimate_prompt_tokens_chain(
             spec.runtime.provider,
             spec.runtime.model,
             messages,
-            tools,
+            tool_definitions,
         )
         assistant_message = build_assistant_message(
             response.content or "",

nanobot/providers/conversation_state.py

@@ -11,6 +11,7 @@ from nanobot.providers.base import (
     ProviderCallContext,
     ProviderConversationState,
 )
+from nanobot.utils.helpers import estimate_prompt_tokens_chain
 
 _PROVIDER_STATE_OUTPUT_META = "provider_state_output"
 _PROVIDER_STATE_BOUNDARY_META = "provider_state_boundary"
@@ -69,6 +70,37 @@ class ProviderConversationStateController:
             session_id=self._session_id,
         )
 
+    def estimate_request_context_tokens(
+        self,
+        messages: list[dict[str, Any]],
+        *,
+        model_messages: list[dict[str, Any]] | None = None,
+        supplemental_messages: list[dict[str, Any]] | None = None,
+        tool_definitions: list[dict[str, Any]] | None = None,
+    ) -> int | None:
+        """Estimate resumed state plus the pending delta for the next request."""
+        state = self.checkpoint(messages, model_messages=model_messages)
+        if state is None:
+            return None
+        context_tokens = state.payload.get("context_tokens")
+        if (
+            isinstance(context_tokens, bool)
+            or not isinstance(context_tokens, int)
+            or context_tokens < 0
+        ):
+            return None
+        pending_messages = [
+            *state.pending_messages,
+            *(supplemental_messages or []),
+        ]
+        delta_tokens, _ = estimate_prompt_tokens_chain(
+            self._provider,
+            self._model,
+            pending_messages,
+            tool_definitions,
+        )
+        return context_tokens + max(0, delta_tokens)
+
     def prepare_request(
         self,
         messages: list[dict[str, Any]],
@@ -76,11 +108,20 @@ class ProviderConversationStateController:
         context_window_tokens: int | None,
         model_messages: list[dict[str, Any]] | None = None,
         supplemental_messages: list[dict[str, Any]] | None = None,
+        resume_state: bool = True,
     ) -> ProviderCallContext | None:
-        """Build typed context for the next request and remember its durable delta."""
+        """Build context for the next request and remember its durable delta.
+
+        ``resume_state=False`` abandons opaque history when local request
+        fitting has produced a new independent model-facing context.
+        """
         independent_context = self.independent_request_context(
             context_window_tokens=context_window_tokens,
         )
+        if not resume_state:
+            self._state = None
+            self._request_messages = []
+            return independent_context
         if self._state is None:
             self._request_messages = []
             return independent_context

tests/agent/test_dream.py

@@ -426,7 +426,7 @@ class TestEphemeralDirect:
                 bus=bus,
                 provider=provider,
                 workspace=tmp_path,
-                context_window_tokens=8000,
+                context_window_tokens=32_000,
             )
 
         return loop, store
@@ -606,7 +606,7 @@ class TestEphemeralDirect:
             bus=MessageBus(),
             provider=provider,
             workspace=tmp_path,
-            context_window_tokens=8000,
+            context_window_tokens=32_000,
         )
 
         await loop.process_direct(
@@ -666,7 +666,7 @@ class TestEphemeralHooks:
                 bus=bus,
                 provider=provider,
                 workspace=tmp_path,
-                context_window_tokens=8000,
+                context_window_tokens=32_000,
                 hooks=[spy],
             )
 

tests/agent/test_loop_consolidation_tokens.py

@@ -29,6 +29,9 @@ def _make_loop(
         workspace=tmp_path,
         model="test-model",
         context_window_tokens=context_window_tokens,
+        # These tests isolate Memory consolidation; Runner request fitting is
+        # covered separately with realistic context windows.
+        context_block_limit=10_000,
     )
     loop.tools.get_definitions = MagicMock(return_value=[])
     loop.consolidator._SAFETY_BUFFER = 0

tests/agent/test_runner_core.py

@@ -11,6 +11,7 @@ import pytest
 
 from agent.runner_helpers import make_run_spec
 from nanobot.agent.context import TranscriptInput
+from nanobot.agent.context_governance import ContextWindowExceededError
 from nanobot.config.schema import AgentDefaults
 from nanobot.providers.base import (
     LLMProvider,
@@ -86,6 +87,7 @@ def test_usage_or_estimate_replaces_reported_zero_for_content(monkeypatch) -> No
         _make_usage_spec(provider, tools),
         [{"role": "user", "content": "hello"}],
         response,
+        tool_definitions=tools.get_definitions(),
     )
 
     assert usage == LLMUsage.estimated(input_tokens=12, output_tokens=7).with_timing(
@@ -130,6 +132,7 @@ def test_usage_or_estimate_counts_tool_call_output_for_reported_zero(monkeypatch
         _make_usage_spec(provider, tools),
         [{"role": "user", "content": "hello"}],
         response,
+        tool_definitions=tools.get_definitions(),
     )
 
     assert usage == LLMUsage.estimated(input_tokens=13, output_tokens=9)
@@ -162,6 +165,7 @@ def test_usage_or_estimate_counts_error_without_estimating_tokens(
         _make_usage_spec(provider, tools),
         [{"role": "user", "content": "hello"}],
         response,
+        tool_definitions=tools.get_definitions(),
     )
 
     assert usage is not None
@@ -197,6 +201,7 @@ def test_usage_or_estimate_trusts_positive_reported_total(monkeypatch) -> None:
         _make_usage_spec(provider, tools),
         [{"role": "user", "content": "hello"}],
         response,
+        tool_definitions=tools.get_definitions(),
     )
 
     assert usage is not None
@@ -366,14 +371,12 @@ async def test_runner_replays_provider_state_without_chat_projection_duplicates(
 
 
 @pytest.mark.asyncio
-async def test_runner_governs_tool_result_before_adding_it_to_provider_state():
+async def test_runner_preserves_tool_result_before_rejecting_unfit_followup():
     from nanobot.agent.runner import AgentRunner
 
     provider = MagicMock(spec=LLMProvider)
     provider.can_resume_conversation_state.return_value = True
-    provider.supports_native_compaction.return_value = False
     calls = 0
-    captured_context: ProviderCallContext | None = None
     checkpoints: list[dict] = []
     state = ProviderConversationState(
         kind="openai_responses",
@@ -384,7 +387,7 @@ async def test_runner_governs_tool_result_before_adding_it_to_provider_state():
     )
 
     async def chat_with_retry(**kwargs):
-        nonlocal calls, captured_context
+        nonlocal calls
         calls += 1
         if calls == 1:
             return LLMResponse(
@@ -398,7 +401,6 @@ async def test_runner_governs_tool_result_before_adding_it_to_provider_state():
                 ],
                 provider_state=state,
             )
-        captured_context = kwargs["provider_context"]
         return LLMResponse(content="done")
 
     provider.chat_with_retry = chat_with_retry
@@ -409,37 +411,36 @@ async def test_runner_governs_tool_result_before_adding_it_to_provider_state():
     async def checkpoint(payload: dict) -> None:
         checkpoints.append(payload)
 
-    await AgentRunner().run(make_run_spec(
+    with pytest.raises(ContextWindowExceededError):
-        provider,
+        await AgentRunner().run(make_run_spec(
-        initial_messages=[
+            provider,
-            {"role": "system", "content": "system"},
+            initial_messages=[
-            {"role": "user", "content": "read the file"},
+                {"role": "system", "content": "system"},
-        ],
+                {"role": "user", "content": "read the file"},
-        tools=tools,
+            ],
-        model="gpt-5.6",
+            tools=tools,
-        context_window_tokens=3_000,
+            model="gpt-5.6",
-        context_block_limit=200,
+            context_window_tokens=3_000,
-        max_tokens=1_000,
+            context_block_limit=200,
-        max_iterations=3,
+            max_tokens=1_000,
-        max_tool_result_chars=10_000,
+            max_iterations=3,
-        checkpoint_callback=checkpoint,
+            max_tool_result_chars=10_000,
-    ))
+            checkpoint_callback=checkpoint,
+        ))
 
-    assert captured_context is not None
+    assert calls == 1
-    assert captured_context.conversation_state is not None
-    pending = captured_context.conversation_state.pending_messages
-    assert len(pending) == 1
-    assert pending[0]["role"] == "tool"
-    assert "compacted to fit context" in pending[0]["content"]
-    assert pending[0]["content"] != "x" * 5_000
     completed_checkpoint = next(
         checkpoint
         for checkpoint in checkpoints
         if checkpoint["phase"] == "tools_completed"
     )
     checkpoint_pending = completed_checkpoint["provider_state"].pending_messages
-    assert "compacted to fit context" in checkpoint_pending[0]["content"]
+    assert checkpoint_pending == [{
-    assert checkpoint_pending[0]["content"] != "x" * 5_000
+        "role": "tool",
+        "tool_call_id": "call_1",
+        "name": "read_file",
+        "content": "x" * 5_000,
+    }]
 
 
 @pytest.mark.asyncio

tests/agent/test_runner_governance.py

@@ -1,8 +1,7 @@
-"""Tests for AgentRunner context governance: backfill, orphan cleanup, microcompact, snip_history."""
+"""Tests for AgentRunner context governance: repair and request fitting."""
 
 from __future__ import annotations
 
-from types import SimpleNamespace
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
@@ -12,11 +11,14 @@ from nanobot.agent.context_governance import (
     BACKFILL_CONTENT,
     ContextGovernanceConfig,
     ContextGovernor,
+    ContextWindowExceededError,
 )
 from nanobot.agent.runner import AgentRunSpec
 from nanobot.config.schema import AgentDefaults
 from nanobot.providers.base import (
+    LLMProvider,
     LLMResponse,
+    LLMUsage,
     ProviderConversationState,
     ToolCallRequest,
 )
@@ -28,8 +30,6 @@ def _governance_config(
     provider,
     tools,
     spec: AgentRunSpec,
-    *,
-    inflight_start_index: int = 0,
 ) -> ContextGovernanceConfig:
     return ContextGovernanceConfig(
         provider=provider,
@@ -41,7 +41,6 @@ def _governance_config(
         context_window_tokens=spec.runtime.context_window_tokens,
         context_block_limit=spec.context_block_limit,
         max_tokens=spec.runtime.generation.max_tokens,
-        inflight_start_index=inflight_start_index,
     )
 
 
@@ -89,6 +88,508 @@ async def test_runner_propagates_context_governance_failure():
     provider.chat_with_retry.assert_not_awaited()
 
 
+@pytest.mark.asyncio
+async def test_runner_locally_fits_oversized_initial_transcript(monkeypatch):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    provider.chat_with_retry = AsyncMock(return_value=LLMResponse(content="done"))
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    old_content = "x" * 20_000
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        lambda _provider, _model, messages, _tools: (
+            (600, "test-counter")
+            if any(message.get("content") == old_content for message in messages)
+            else (100, "test-counter")
+        ),
+    )
+
+    result = await AgentRunner().run(make_run_spec(
+        provider,
+        initial_messages=[
+            {"role": "system", "content": "system"},
+            {"role": "user", "content": "old question"},
+            {"role": "assistant", "content": old_content},
+            {"role": "user", "content": "continue"},
+        ],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_tokens=100,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+    ))
+
+    assert provider.chat_with_retry.await_args.kwargs["messages"] == [
+        {"role": "system", "content": "system"},
+        {"role": "user", "content": "continue"},
+    ]
+    assert any(message.get("content") == old_content for message in result.messages)
+
+
+@pytest.mark.asyncio
+async def test_runner_governs_messages_added_by_before_iteration_hook(monkeypatch):
+    from nanobot.agent.hook import AgentHook, AgentHookContext
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    provider.chat_with_retry = AsyncMock(return_value=LLMResponse(content="unexpected"))
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    oversized = "hook-added-oversized-message"
+
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        lambda _provider, _model, messages, _tools: (
+            (2_000, "test-counter")
+            if any(message.get("content") == oversized for message in messages)
+            else (100, "test-counter")
+        ),
+    )
+
+    class MutatingHook(AgentHook):
+        async def before_iteration(self, context: AgentHookContext) -> None:
+            context.messages.append({"role": "user", "content": oversized})
+
+    with pytest.raises(ContextWindowExceededError):
+        await AgentRunner().run(make_run_spec(
+            provider,
+            initial_messages=[{"role": "user", "content": "hello"}],
+            tools=tools,
+            model="local-model",
+            context_window_tokens=2_000,
+            context_block_limit=500,
+            max_iterations=1,
+            max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+            hook=MutatingHook(),
+        ))
+
+    provider.chat_with_retry.assert_not_awaited()
+
+
+@pytest.mark.asyncio
+async def test_runner_drops_resumable_provider_state_when_request_is_fitted(monkeypatch):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    provider.can_resume_conversation_state.return_value = True
+    captured_contexts = []
+    old_content = "old-oversized-history"
+    candidate = ProviderConversationState(
+        kind="openai_responses",
+        provider="openai:test",
+        model="local-model",
+        version=1,
+        payload={"items": [{"type": "message", "content": "fresh state"}]},
+    )
+
+    async def chat_with_retry(*, provider_context=None, **_kwargs):
+        captured_contexts.append(provider_context)
+        return LLMResponse(
+            content="done",
+            usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+            provider_state=candidate,
+        )
+
+    provider.chat_with_retry = chat_with_retry
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        lambda _provider, _model, messages, _tools: (
+            (600, "test-counter")
+            if any(message.get("content") == old_content for message in messages)
+            else (100, "test-counter")
+        ),
+    )
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_message_tokens",
+        lambda message: 450 if message.get("content") == old_content else 50,
+    )
+    saved_state = ProviderConversationState(
+        kind="openai_responses",
+        provider="openai:test",
+        model="local-model",
+        version=1,
+        payload={"items": [{"type": "message", "content": "stale state"}]},
+    )
+
+    result = await AgentRunner().run(make_run_spec(
+        provider,
+        initial_messages=[
+            {"role": "assistant", "content": old_content},
+            {"role": "user", "content": "continue"},
+        ],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+        provider_state=saved_state,
+    ))
+
+    assert captured_contexts[0].conversation_state is None
+    assert result.provider_state is not None
+    assert result.provider_state.payload == candidate.payload
+
+
+@pytest.mark.asyncio
+async def test_runner_fits_each_malformed_retry_with_its_actual_tools(monkeypatch):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    calls: list[dict] = []
+    estimated_tools: list[object] = []
+    definitions = [{"type": "function", "function": {"name": "read_file"}}]
+
+    async def chat_with_retry(*, messages, tools=None, **_kwargs):
+        calls.append({"messages": [dict(message) for message in messages], "tools": tools})
+        if len(calls) < 3:
+            return LLMResponse(
+                content="bad tool request",
+                tool_calls=[ToolCallRequest(id=f"bad_{len(calls)}", name=None, arguments={})],
+                finish_reason="tool_calls",
+                usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+            )
+        return LLMResponse(
+            content="recovered",
+            usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+        )
+
+    def estimate(_provider, _model, messages, _tools):
+        estimated_tools.append(_tools)
+        user_count = sum(message.get("role") == "user" for message in messages)
+        return (600 if user_count > 1 else 100), "test-counter"
+
+    provider.chat_with_retry = chat_with_retry
+    tools = MagicMock()
+    tools.get_definitions.return_value = definitions
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        estimate,
+    )
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_message_tokens",
+        lambda _message: 300,
+    )
+
+    result = await AgentRunner().run(make_run_spec(
+        provider,
+        initial_messages=[{"role": "user", "content": "use a tool"}],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+    ))
+
+    assert [call["tools"] for call in calls] == [definitions, definitions, None]
+    assert definitions in estimated_tools
+    assert None in estimated_tools
+    assert [len(call["messages"]) for call in calls] == [1, 1, 1]
+    assert result.final_content == "recovered"
+    assert result.messages == [
+        {"role": "user", "content": "use a tool"},
+        {"role": "assistant", "content": "recovered"},
+    ]
+
+
+@pytest.mark.asyncio
+async def test_runner_fits_empty_response_finalization_before_dispatch(monkeypatch):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    calls: list[dict] = []
+
+    async def chat_with_retry(*, messages, tools=None, **_kwargs):
+        calls.append({"messages": [dict(message) for message in messages], "tools": tools})
+        if len(calls) < 3:
+            return LLMResponse(
+                content=None,
+                usage=LLMUsage.reported(input_tokens=100, output_tokens=1),
+            )
+        return LLMResponse(
+            content="finalized",
+            usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+        )
+
+    def estimate(_provider, _model, messages, _tools):
+        contents = [str(message.get("content") or "") for message in messages]
+        has_original = "do task" in contents
+        has_finalization = any("conversation above" in content for content in contents)
+        return (600 if has_original and has_finalization else 100), "test-counter"
+
+    provider.chat_with_retry = chat_with_retry
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        estimate,
+    )
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_message_tokens",
+        lambda _message: 300,
+    )
+
+    result = await AgentRunner().run(make_run_spec(
+        provider,
+        initial_messages=[{"role": "user", "content": "do task"}],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=3,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+    ))
+
+    assert len(calls) == 3
+    assert calls[-1]["tools"] is None
+    assert all(message.get("content") != "do task" for message in calls[-1]["messages"])
+    assert result.final_content == "finalized"
+
+
+@pytest.mark.asyncio
+async def test_runner_fits_max_iteration_finalization_before_dispatch(monkeypatch):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    calls: list[dict] = []
+    oversized_result = "oversized-current-tool-result"
+
+    async def chat_with_retry(*, messages, tools=None, **_kwargs):
+        calls.append({"messages": [dict(message) for message in messages], "tools": tools})
+        if len(calls) == 1:
+            return LLMResponse(
+                content="working",
+                tool_calls=[ToolCallRequest(id="call_1", name="read_file", arguments={})],
+                finish_reason="tool_calls",
+                usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+            )
+        return LLMResponse(
+            content="safe summary",
+            usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+        )
+
+    def estimate(_provider, _model, messages, _tools):
+        has_oversized = any(
+            message.get("content") == oversized_result for message in messages
+        )
+        return (600 if has_oversized else 100), "test-counter"
+
+    provider.chat_with_retry = chat_with_retry
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    tools.execute = AsyncMock(return_value=oversized_result)
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        estimate,
+    )
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_message_tokens",
+        lambda message: 600 if message.get("content") == oversized_result else 50,
+    )
+
+    result = await AgentRunner().run(make_run_spec(
+        provider,
+        initial_messages=[{"role": "user", "content": "inspect"}],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+    ))
+
+    assert len(calls) == 2
+    assert calls[-1]["tools"] is None
+    assert all(
+        message.get("content") != oversized_result
+        for message in calls[-1]["messages"]
+    )
+    assert any(message.get("content") == oversized_result for message in result.messages)
+    assert result.final_content == "safe summary"
+
+
+@pytest.mark.parametrize(
+    ("input_tokens", "expected_fitted"),
+    [(500, True), (100, False)],
+)
+def test_matching_reported_provider_usage_avoids_local_estimate(
+    monkeypatch,
+    input_tokens,
+    expected_fitted,
+):
+    provider = MagicMock(spec=LLMProvider)
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    spec = make_run_spec(
+        provider,
+        initial_messages=[{"role": "user", "content": "hello"}],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+    )
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        lambda *_args, **_kwargs: (_ for _ in ()).throw(
+            AssertionError("matching provider usage must be authoritative")
+        ),
+    )
+
+    governor = ContextGovernor()
+    monkeypatch.setattr(governor, "fit_to_budget", lambda *_args, **_kwargs: [])
+    _messages, fitted = governor.fit_request(
+        _governance_config(provider, tools, spec),
+        spec.initial_messages,
+        LLMUsage.reported(input_tokens=input_tokens, output_tokens=10),
+        usage_matches_messages=True,
+        tool_definitions=tools.get_definitions(),
+    )
+
+    assert fitted is expected_fitted
+
+
+def test_changed_messages_use_local_estimate_after_reported_usage(monkeypatch):
+    provider = MagicMock(spec=LLMProvider)
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    spec = make_run_spec(
+        provider,
+        initial_messages=[{"role": "user", "content": "new tool output"}],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+    )
+    estimate = MagicMock(return_value=(600, "test-counter"))
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        estimate,
+    )
+
+    governor = ContextGovernor()
+    monkeypatch.setattr(governor, "fit_to_budget", lambda *_args, **_kwargs: [])
+    _messages, fitted = governor.fit_request(
+        _governance_config(provider, tools, spec),
+        spec.initial_messages,
+        LLMUsage.reported(input_tokens=900, output_tokens=10),
+        usage_matches_messages=False,
+        tool_definitions=tools.get_definitions(),
+    )
+
+    assert fitted is True
+    estimate.assert_called_once()
+
+
+@pytest.mark.asyncio
+async def test_runner_counts_resumed_provider_state_before_dispatch(monkeypatch):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    provider.can_resume_conversation_state.return_value = True
+    captured_contexts = []
+
+    async def chat_with_retry(*, provider_context=None, **_kwargs):
+        captured_contexts.append(provider_context)
+        return LLMResponse(
+            content="done",
+            usage=LLMUsage.reported(input_tokens=100, output_tokens=10),
+        )
+
+    provider.chat_with_retry = chat_with_retry
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    current_message = {"role": "user", "content": "new delta"}
+    saved_state = ProviderConversationState(
+        kind="openai_responses",
+        provider="openai:test",
+        model="local-model",
+        version=1,
+        payload={
+            "items": [{"type": "reasoning", "encrypted_content": "opaque"}],
+            "context_tokens": 450,
+        },
+        pending_messages=[current_message],
+    )
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        lambda *_args, **_kwargs: (100, "test-counter"),
+    )
+    monkeypatch.setattr(
+        "nanobot.providers.conversation_state.estimate_prompt_tokens_chain",
+        lambda *_args, **_kwargs: (100, "test-counter"),
+    )
+
+    result = await AgentRunner().run(make_run_spec(
+        provider,
+        initial_messages=[current_message],
+        tools=tools,
+        model="local-model",
+        context_window_tokens=2_000,
+        context_block_limit=500,
+        max_iterations=1,
+        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+        provider_state=saved_state,
+    ))
+
+    assert captured_contexts[0].conversation_state is None
+    assert result.messages == [
+        current_message,
+        {"role": "assistant", "content": "done"},
+    ]
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    ("context_block_limit", "expected_budget"),
+    [(500, 500), (None, 0)],
+)
+async def test_runner_refuses_locally_fitted_request_that_still_cannot_fit(
+    monkeypatch,
+    context_block_limit,
+    expected_budget,
+):
+    from nanobot.agent.runner import AgentRunner
+
+    provider = MagicMock(spec=LLMProvider)
+    provider.chat_with_retry = AsyncMock(return_value=LLMResponse(content="unexpected"))
+    tools = MagicMock()
+    tools.get_definitions.return_value = []
+    monkeypatch.setattr(
+        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
+        lambda *_args, **_kwargs: (2_000, "test-counter"),
+    )
+
+    with pytest.raises(ContextWindowExceededError) as exc_info:
+        await AgentRunner().run(make_run_spec(
+            provider,
+            initial_messages=[
+                {"role": "system", "content": "oversized system"},
+                {"role": "user", "content": "oversized user"},
+            ],
+            tools=tools,
+            model="local-model",
+            context_window_tokens=1_000,
+            context_block_limit=context_block_limit,
+            max_iterations=1,
+            max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
+        ))
+
+    assert exc_info.value.estimated_tokens == 2_000
+    assert exc_info.value.input_budget == expected_budget
+    provider.chat_with_retry.assert_not_awaited()
+
+
 def test_snip_history_drops_orphaned_tool_results_from_trimmed_slice(monkeypatch):
     provider = MagicMock()
     tools = MagicMock()
@@ -130,7 +631,11 @@ def test_snip_history_drops_orphaned_tool_results_from_trimmed_slice(monkeypatch
         lambda msg: token_sizes.get(str(msg.get("content")), 40),
     )
 
-    trimmed = ContextGovernor().snip_history(_governance_config(provider, tools, spec), messages)
+    trimmed = ContextGovernor().snip_history(
+        _governance_config(provider, tools, spec),
+        messages,
+        tool_definitions=tools.get_definitions(),
+    )
 
     # After the fix, the user message is recovered so the sequence is valid
     # for providers that require system → user (e.g. GLM error 1214).
@@ -182,7 +687,11 @@ def test_snip_history_reserves_budget_for_tool_definitions(monkeypatch):
         lambda msg: token_sizes.get(str(msg.get("content")), 40),
     )
 
-    trimmed = ContextGovernor().snip_history(_governance_config(provider, tools, spec), messages)
+    trimmed = ContextGovernor().snip_history(
+        _governance_config(provider, tools, spec),
+        messages,
+        tool_definitions=tools.get_definitions(),
+    )
 
     contents = [message.get("content") for message in trimmed]
     assert contents == ["system", "recent two"]
@@ -465,260 +974,6 @@ async def test_runner_backfill_only_mutates_model_context_not_returned_messages(
     ]
 
 
-# ---------------------------------------------------------------------------
-# Microcompact (stale tool result compaction)
-# ---------------------------------------------------------------------------
-
-
-def _microcompact_messages(*, total: int, tool_name: str, content: str) -> list[dict]:
-    messages: list[dict] = [{"role": "system", "content": "sys"}]
-    for i in range(total):
-        messages.append({
-            "role": "assistant",
-            "content": "",
-            "tool_calls": [{
-                "id": f"c{i}",
-                "type": "function",
-                "function": {"name": tool_name, "arguments": "{}"},
-            }],
-        })
-        messages.append({
-            "role": "tool",
-            "tool_call_id": f"c{i}",
-            "name": tool_name,
-            "content": content,
-        })
-    return messages
-
-
-def test_microcompact_skips_when_prompt_under_hard_budget(monkeypatch):
-    """Cache-friendly path: in-flight tool results stay stable while prompt fits."""
-    provider = MagicMock()
-    provider.generation = SimpleNamespace(max_tokens=0)
-    tools = MagicMock()
-    tools.get_definitions.return_value = []
-
-    total = 15
-    long_content = "x" * 600
-    messages = _microcompact_messages(total=total, tool_name="read_file", content=long_content)
-    spec = make_run_spec(provider,
-        initial_messages=messages,
-        tools=tools,
-        model="test-model",
-        max_iterations=1,
-        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
-        max_tokens=0,
-        context_window_tokens=20_000,
-    )
-
-    monkeypatch.setattr(
-        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
-        lambda *_args, **_kwargs: (1000, "test"),
-    )
-
-    result = ContextGovernor().compact_inflight_overflow(
-        _governance_config(provider, tools, spec),
-        messages,
-        set(),
-    )
-
-    assert result is messages
-
-
-def test_microcompact_overflow_compacts_to_low_watermark(monkeypatch):
-    """Overflow path: compact in-flight stale results with headroom for later calls."""
-    provider = MagicMock()
-    provider.generation = SimpleNamespace(max_tokens=0)
-    tools = MagicMock()
-    tools.get_definitions.return_value = []
-
-    total = 18
-    long_content = "x" * 600
-    messages = _microcompact_messages(total=total, tool_name="read_file", content=long_content)
-    spec = make_run_spec(provider,
-        initial_messages=messages,
-        tools=tools,
-        model="test-model",
-        max_iterations=1,
-        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
-        max_tokens=0,
-        context_window_tokens=2224,  # input budget 1200, low target 1020
-    )
-
-    def estimate(_provider, _model, msgs, _tools):
-        return sum(
-            100 if (content := msg.get("content")) == long_content
-            else 1 if isinstance(content, str) and "compacted to fit context" in content
-            else 0
-            for msg in msgs
-            if msg.get("role") == "tool"
-        ), "test"
-
-    monkeypatch.setattr("nanobot.agent.context_governance.estimate_prompt_tokens_chain", estimate)
-
-    result = ContextGovernor().compact_inflight_overflow(
-        _governance_config(provider, tools, spec),
-        messages,
-        set(),
-    )
-    tool_msgs = [m for m in result if m.get("role") == "tool"]
-    compacted = [m for m in tool_msgs if "compacted to fit context" in str(m.get("content", ""))]
-    preserved = [m for m in tool_msgs if m.get("content") == long_content]
-
-    assert len(compacted) == 8
-    assert len(preserved) == total - 8
-    assert [m["tool_call_id"] for m in compacted] == [f"c{i}" for i in range(8)]
-
-
-def test_microcompact_compacts_newest_when_it_alone_overflows(monkeypatch):
-    """An unfit newest result tells the model to retry narrowly or report the limit."""
-    provider = MagicMock()
-    provider.generation = SimpleNamespace(max_tokens=0)
-    tools = MagicMock()
-    tools.get_definitions.return_value = []
-
-    long_content = "x" * 600
-    messages = _microcompact_messages(total=1, tool_name="read_file", content=long_content)
-    spec = make_run_spec(provider,
-        initial_messages=messages,
-        tools=tools,
-        model="test-model",
-        max_iterations=1,
-        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
-        max_tokens=0,
-        context_window_tokens=2000,
-        context_block_limit=500,
-    )
-
-    def estimate(_provider, _model, msgs, _tools):
-        return sum(
-            1000 if msg.get("content") == long_content else 1
-            for msg in msgs
-            if msg.get("role") == "tool"
-        ), "test"
-
-    monkeypatch.setattr("nanobot.agent.context_governance.estimate_prompt_tokens_chain", estimate)
-
-    compacted_tool_call_ids: set[str] = set()
-    result = ContextGovernor().compact_inflight_overflow(
-        _governance_config(provider, tools, spec),
-        messages,
-        compacted_tool_call_ids,
-    )
-
-    tool_msg = next(m for m in result if m.get("role") == "tool")
-    assert "compacted to fit context" in tool_msg["content"]
-    assert "Do not repeat the same call unchanged" in tool_msg["content"]
-    assert "Retry with a narrower path, query, range, or result limit" in tool_msg["content"]
-    assert "tell the user the task cannot fit" in tool_msg["content"]
-    assert compacted_tool_call_ids == {"c0"}
-
-
-def test_context_governor_keeps_compaction_boundary_stable(monkeypatch):
-    provider = MagicMock()
-    provider.generation = SimpleNamespace(max_tokens=0)
-    tools = MagicMock()
-    tools.get_definitions.return_value = []
-
-    total = 18
-    long_content = "x" * 600
-    messages = _microcompact_messages(total=total, tool_name="read_file", content=long_content)
-    spec = make_run_spec(provider,
-        initial_messages=messages,
-        tools=tools,
-        model="test-model",
-        max_iterations=1,
-        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
-        max_tokens=0,
-        context_window_tokens=2224,
-    )
-
-    def estimate(_provider, _model, msgs, _tools):
-        return sum(
-            100 if msg.get("content") == long_content else 1
-            for msg in msgs
-            if msg.get("role") == "tool"
-        ), "test"
-
-    monkeypatch.setattr("nanobot.agent.context_governance.estimate_prompt_tokens_chain", estimate)
-
-    governor = ContextGovernor()
-    compacted_tool_call_ids: set[str] = set()
-    config = _governance_config(provider, tools, spec, inflight_start_index=0)
-    first = governor.compact_inflight_overflow(config, messages, compacted_tool_call_ids)
-    first_ids = set(compacted_tool_call_ids)
-
-    second = governor.compact_inflight_overflow(config, messages, compacted_tool_call_ids)
-
-    assert compacted_tool_call_ids == first_ids
-    assert [m.get("content") for m in second] == [m.get("content") for m in first]
-
-
-def test_microcompact_preserves_short_results(monkeypatch):
-    """Short tool results below the compaction threshold should not be replaced."""
-    provider = MagicMock()
-    provider.generation = SimpleNamespace(max_tokens=0)
-    tools = MagicMock()
-    tools.get_definitions.return_value = []
-
-    total = 15
-    messages = _microcompact_messages(total=total, tool_name="exec", content="short")
-    spec = make_run_spec(provider,
-        initial_messages=messages,
-        tools=tools,
-        model="test-model",
-        max_iterations=1,
-        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
-        max_tokens=0,
-        context_window_tokens=2024,
-    )
-
-    monkeypatch.setattr(
-        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
-        lambda *_args, **_kwargs: (2000, "test"),
-    )
-
-    result = ContextGovernor().compact_inflight_overflow(
-        _governance_config(provider, tools, spec),
-        messages,
-        set(),
-    )
-    assert result is messages  # no copy needed — all stale results are short
-
-
-def test_microcompact_skips_non_compactable_tools(monkeypatch):
-    """Non-compactable tools (e.g. 'message') should never be replaced."""
-    provider = MagicMock()
-    provider.generation = SimpleNamespace(max_tokens=0)
-    tools = MagicMock()
-    tools.get_definitions.return_value = []
-
-    total = 15
-    long_content = "y" * 1000
-    messages = _microcompact_messages(total=total, tool_name="message", content=long_content)
-    spec = make_run_spec(provider,
-        initial_messages=messages,
-        tools=tools,
-        model="test-model",
-        max_iterations=1,
-        max_tool_result_chars=_MAX_TOOL_RESULT_CHARS,
-        max_tokens=0,
-        context_window_tokens=2024,
-    )
-
-    monkeypatch.setattr(
-        "nanobot.agent.context_governance.estimate_prompt_tokens_chain",
-        lambda *_args, **_kwargs: (2000, "test"),
-    )
-
-    result = ContextGovernor().compact_inflight_overflow(
-        _governance_config(provider, tools, spec),
-        messages,
-        set(),
-    )
-    assert result is messages  # no compactable tools found
-
-
 def test_governance_repairs_orphans_after_snip():
     """After snipping clips an assistant+tool_calls, orphan repair cleans up the tail."""
     # Simulate snipping that keeps only the tail: drop the assistant with
@@ -818,7 +1073,11 @@ def test_snip_history_preserves_user_message_after_truncation(monkeypatch):
         lambda msg: token_sizes.get(str(msg.get("content")), 100),
     )
 
-    trimmed = ContextGovernor().snip_history(_governance_config(provider, tools, spec), messages)
+    trimmed = ContextGovernor().snip_history(
+        _governance_config(provider, tools, spec),
+        messages,
+        tool_definitions=tools.get_definitions(),
+    )
 
     # The first non-system message MUST be user (not assistant).
     non_system = [m for m in trimmed if m.get("role") != "system"]
@@ -863,7 +1122,11 @@ def test_snip_history_no_user_at_all_falls_back_gracefully(monkeypatch):
         lambda msg: 100,
     )
 
-    trimmed = ContextGovernor().snip_history(_governance_config(provider, tools, spec), messages)
+    trimmed = ContextGovernor().snip_history(
+        _governance_config(provider, tools, spec),
+        messages,
+        tool_definitions=tools.get_definitions(),
+    )
 
     # Should not crash.  The result should still be a valid list.
     assert isinstance(trimmed, list)
@@ -871,7 +1134,6 @@ def test_snip_history_no_user_at_all_falls_back_gracefully(monkeypatch):
     assert any(m.get("role") == "system" for m in trimmed)
     # The _enforce_role_alternation safety net must be able to fix whatever
     # _snip_history returns here — verify it produces a valid sequence.
-    from nanobot.providers.base import LLMProvider
     fixed = LLMProvider._enforce_role_alternation(trimmed)
     non_system = [m for m in fixed if m["role"] != "system"]
     if non_system:

tests/agent/test_session_model_runtime.py

@@ -114,7 +114,7 @@ async def test_removed_session_model_preset_falls_back_and_clears_metadata(tmp_p
         provider=base,
         workspace=tmp_path,
         model="base-model",
-        context_window_tokens=8_000,
+        context_window_tokens=16_000,
     )
     loop.schedule_background = lambda coro: coro.close()  # type: ignore[method-assign]
     session_key = "sdk:removed-preset"
@@ -196,7 +196,7 @@ async def test_sdk_custom_model_preset_metadata_does_not_select_runtime(
         provider=base,
         workspace=tmp_path,
         model="base-model",
-        context_window_tokens=8_000,
+        context_window_tokens=16_000,
     )
     loop.schedule_background = lambda coro: coro.close()  # type: ignore[method-assign]
     bot = Nanobot(loop)

tests/providers/test_conversation_state.py

@@ -146,6 +146,51 @@ def test_controller_uses_governed_messages_for_provider_state_delta() -> None:
     assert governed_checkpoint.pending_messages[-1]["content"] == "compacted result"
 
 
+def test_controller_estimates_active_state_plus_pending_delta(monkeypatch) -> None:
+    provider = _provider()
+    current_message = {"role": "user", "content": "new delta"}
+    state = ProviderConversationState(
+        kind="openai_responses",
+        provider="openai:test",
+        model="gpt-5.6",
+        version=1,
+        payload={
+            "items": [{"type": "reasoning", "encrypted_content": "opaque"}],
+            "context_tokens": 450,
+        },
+        pending_messages=[current_message],
+    )
+    controller = ProviderConversationStateController(
+        provider=provider,
+        model="gpt-5.6",
+        messages=[current_message],
+        state=state,
+    )
+    seen = {}
+
+    def estimate(_provider, _model, messages, tools):
+        seen["messages"] = messages
+        seen["tools"] = tools
+        return 100, "test-counter"
+
+    monkeypatch.setattr(
+        "nanobot.providers.conversation_state.estimate_prompt_tokens_chain",
+        estimate,
+    )
+
+    tokens = controller.estimate_request_context_tokens(
+        [current_message],
+        model_messages=[current_message],
+        tool_definitions=[{"type": "web_search"}],
+    )
+
+    assert tokens == 550
+    assert seen == {
+        "messages": [current_message],
+        "tools": [{"type": "web_search"}],
+    }
+
+
 def test_transient_response_preserves_only_durable_request_messages() -> None:
     provider = _provider()
     current_message = {"role": "user", "content": "continue"}