nanobot/nanobot/agent/loop.py

"""Agent loop: the core processing engine."""

from __future__ import annotations

import asyncio
import dataclasses
import os
import time
from collections.abc import Mapping
from contextlib import AbstractContextManager, ExitStack, nullcontext, suppress
from dataclasses import dataclass, field
from enum import Enum, auto
from functools import partial
from pathlib import Path
from typing import TYPE_CHECKING, Any, Awaitable, Callable

from loguru import logger

from nanobot.agent import context as agent_context
from nanobot.agent import model_presets as preset_helpers
from nanobot.agent.autocompact import AutoCompact
from nanobot.agent.automation_turns import publish_next_deferred_turn
from nanobot.agent.context import ContextBuilder
from nanobot.agent.cron_turns import CronTurnCoordinator
from nanobot.agent.hook import AgentHook, AgentTurnHookFactory
from nanobot.agent.memory import Consolidator
from nanobot.agent.model_runtime import ModelRuntimeResolver
from nanobot.agent.runner import _MAX_INJECTIONS_PER_TURN, AgentRunner, AgentRunSpec
from nanobot.agent.subagent import SubagentManager
from nanobot.agent.tools.context import RequestContext, bind_request_context, reset_request_context
from nanobot.agent.tools.exec_session import ExecSessionManager
from nanobot.agent.tools.file_state import FileStateStore, bind_file_states, reset_file_states
from nanobot.agent.tools.message import MessageTool
from nanobot.agent.tools.registry import ToolRegistry
from nanobot.agent.tools.self import MyTool
from nanobot.agent.turn_delivery import (
    TurnDelivery,
    TurnDeliveryFactory,
)
from nanobot.agent.turn_delivery import TurnRoute as TurnRoute
from nanobot.agent.turn_hooks import AgentTurnHookSpec, build_agent_turn_hook
from nanobot.bus.events import InboundMessage, OutboundMessage
from nanobot.bus.outbound_events import StreamedResponseEvent
from nanobot.bus.queue import MessageBus
from nanobot.bus.runtime_events import (
    RuntimeEventBus,
    RuntimeEventPublisher,
    ensure_runtime_event_publisher,
)
from nanobot.command import CommandContext, CommandRouter, register_builtin_commands
from nanobot.config.schema import AgentDefaults, ModelPresetConfig
from nanobot.providers.base import LLMProvider
from nanobot.providers.factory import ProviderSnapshot
from nanobot.runtime_context import (
    RUNTIME_CONTEXT_HISTORY_META,
    RUNTIME_CONTEXT_MESSAGE_META,
    RuntimeContextBlock,
    RuntimeContextProvider,
    append_runtime_context,
    resolve_runtime_context,
    runtime_context_blocks_from_metadata,
)
from nanobot.security.workspace_access import (
    WorkspaceScopeResolver,
    bind_workspace_scope,
    reset_workspace_scope,
)
from nanobot.session import turn_continuation
from nanobot.session.automation_turns import automation_history_overrides
from nanobot.session.goal_state import (
    goal_state_runtime_lines,
    runner_wall_llm_timeout_s,
    sustained_goal_active,
)
from nanobot.session.history_visibility import HIDDEN_HISTORY_META
from nanobot.session.keys import UNIFIED_SESSION_KEY, remember_last_channel
from nanobot.session.manager import (
    Session,
    SessionManager,
    replay_max_messages_for_context,
)
from nanobot.session.model_selection import (
    SESSION_MODEL_PRESET_METADATA_KEY,
    model_preset_from_metadata,
)
from nanobot.triggers.local_turns import LocalTriggerTurnCoordinator
from nanobot.utils.cancellation import task_is_cancelling
from nanobot.utils.document import extract_documents, reference_non_image_attachments
from nanobot.utils.helpers import image_placeholder_text
from nanobot.utils.helpers import truncate_text as truncate_text_fn
from nanobot.utils.llm_runtime import LLMRuntime
from nanobot.utils.runtime import (
    EMPTY_FINAL_RESPONSE_MESSAGE,
)

if TYPE_CHECKING:
    from nanobot.agent.tools.mcp import MCPConnection
    from nanobot.config.schema import (
        ChannelsConfig,
        ProviderConfig,
        ToolsConfig,
    )
    from nanobot.cron.service import CronService

class TurnState(Enum):
    RESTORE = auto()
    COMPACT = auto()
    COMMAND = auto()
    BUILD = auto()
    RUN = auto()
    SAVE = auto()
    RESPOND = auto()
    DONE = auto()


class TurnKind(Enum):
    USER = auto()
    SYSTEM = auto()


@dataclass
class StateTraceEntry:
    state: TurnState
    started_at: float
    duration_ms: float
    event: str
    error: str | None = None


@dataclass
class TurnContext:
    msg: InboundMessage
    session_key: str
    state: TurnState
    turn_id: str
    runtime: LLMRuntime | None
    kind: TurnKind
    delivery: TurnDelivery
    original_user_text: str | None = None
    session: Session | None = None

    history: list[dict[str, Any]] = field(default_factory=list)
    initial_messages: list[dict[str, Any]] = field(default_factory=list)
    request_context: RequestContext | None = None
    runtime_context_blocks: list[RuntimeContextBlock] = field(default_factory=list)

    final_content: str | None = None
    tools_used: list[str] = field(default_factory=list)
    all_messages: list[dict[str, Any]] = field(default_factory=list)
    stop_reason: str = ""
    had_injections: bool = False
    streamed_content: bool = False

    input_persisted_early: bool = False
    save_skip: int = 0

    outbound: OutboundMessage | None = None
    suppress_response: bool = False

    on_progress: Callable[..., Awaitable[None]] | None = None
    on_stream: Callable[[str], Awaitable[None]] | None = None
    on_stream_end: Callable[..., Awaitable[None]] | None = None
    on_runtime_admitted: Callable[[LLMRuntime], Awaitable[None]] | None = None
    on_retry_wait: Callable[[str], Awaitable[None]] | None = None

    pending_queue: asyncio.Queue | None = None
    pending_summary: str | None = None

    ephemeral: bool = False
    run_extra_hooks_for_ephemeral: bool = False
    hooks: list[AgentHook] = field(default_factory=list)
    hook_factories: list[AgentTurnHookFactory] = field(default_factory=list)
    turn_scopes: list[AbstractContextManager[Any]] = field(default_factory=list)
    tools: ToolRegistry | None = None

    turn_wall_started_at: float = field(default_factory=time.time)
    visible_run_started_at: float | None = None
    turn_latency_ms: int | None = None

    trace: list[StateTraceEntry] = field(default_factory=list)


class AgentLoop:
    """
    The agent loop is the core processing engine.

    It:
    1. Receives messages from the bus
    2. Builds context with history, memory, skills
    3. Calls the LLM
    4. Executes tool calls
    5. Sends responses back
    """

    @property
    def current_iteration(self) -> int:
        return self._current_iteration

    @property
    def tool_names(self) -> list[str]:
        return self.tools.tool_names

    @property
    def provider(self) -> LLMProvider:
        """Provider selected for future turn admissions."""
        return self.runtime_resolver.runtime.provider

    @property
    def model(self) -> str:
        """Model selected for future turn admissions."""
        return self.runtime_resolver.runtime.model

    @property
    def context_window_tokens(self) -> int:
        """Context limit selected for future turn admissions."""
        return self.runtime_resolver.runtime.context_window_tokens

    @property
    def model_presets(self) -> Mapping[str, ModelPresetConfig]:
        """Configured model presets exposed for selection and display."""
        return self.runtime_resolver.model_presets

    @property
    def model_preset(self) -> str | None:
        return self.runtime_resolver.model_preset

    @model_preset.setter
    def model_preset(self, name: str | None) -> None:
        self.set_model_preset(name)

    def llm_runtime(self) -> LLMRuntime:
        """Resolve the immutable default used to admit the next turn."""
        previous = self.runtime_resolver.runtime
        runtime = self.runtime_resolver.admit()
        if (
            runtime.model != previous.model
            or runtime.model_preset != previous.model_preset
            or runtime.snapshot_signature != previous.snapshot_signature
        ):
            self._publish_runtime_selection(runtime)
        return runtime

    _RUNTIME_CHECKPOINT_KEY = "runtime_checkpoint"
    _PENDING_USER_TURN_KEY = "pending_user_turn"

    # Event-driven state transition table.
    # Handlers return an event string; the driver looks up the next state here.
    _TRANSITIONS: dict[tuple[TurnState, str], TurnState] = {
        (TurnState.RESTORE, "ok"): TurnState.COMPACT,
        (TurnState.COMPACT, "ok"): TurnState.COMMAND,
        (TurnState.COMMAND, "dispatch"): TurnState.BUILD,
        (TurnState.COMMAND, "shortcut"): TurnState.DONE,
        (TurnState.BUILD, "ok"): TurnState.RUN,
        (TurnState.RUN, "ok"): TurnState.SAVE,
        (TurnState.SAVE, "ok"): TurnState.RESPOND,
        (TurnState.RESPOND, "ok"): TurnState.DONE,
    }

    def __init__(
        self,
        bus: MessageBus,
        provider: LLMProvider,
        workspace: Path,
        model: str | None = None,
        max_iterations: int | None = None,
        max_concurrent_subagents: int | None = None,
        context_window_tokens: int | None = None,
        context_block_limit: int | None = None,
        max_tool_result_chars: int | None = None,
        fail_on_tool_error: bool | None = None,
        provider_retry_mode: str = "standard",
        tool_hint_max_length: int | None = None,
        cron_service: CronService | None = None,
        restrict_to_workspace: bool = False,
        session_manager: SessionManager | None = None,
        mcp_servers: dict | None = None,
        channels_config: ChannelsConfig | None = None,
        timezone: str | None = None,
        session_ttl_minutes: int = 0,
        consolidation_ratio: float = 0.5,
        hooks: list[AgentHook] | None = None,
        hook_factories: list[AgentTurnHookFactory] | None = None,
        unified_session: bool = False,
        disabled_skills: list[str] | None = None,
        tools_config: ToolsConfig | None = None,
        image_generation_provider_config: ProviderConfig | None = None,
        image_generation_provider_configs: dict[str, ProviderConfig] | None = None,
        provider_snapshot_loader: Callable[..., ProviderSnapshot] | None = None,
        provider_signature: tuple[object, ...] | None = None,
        model_presets: dict[str, ModelPresetConfig] | None = None,
        preset_catalog_loader: preset_helpers.PresetCatalogLoader | None = None,
        model_preset: str | None = None,
        preset_snapshot_loader: preset_helpers.PresetSnapshotLoader | None = None,
        runtime_events: RuntimeEventBus | None = None,
        turn_delivery_factory: TurnDeliveryFactory | None = None,
        runtime_model_publisher: Callable[[str, str | None], None] | None = None,
        restart_mode: str = "auto",
        local_trigger_store: Any | None = None,
    ):
        from nanobot.config.schema import ToolsConfig

        _tc = tools_config or ToolsConfig()
        defaults = AgentDefaults()
        self.bus = bus
        if turn_delivery_factory is not None:
            if turn_delivery_factory.bus is not bus:
                raise ValueError("turn delivery factory must use the agent message bus")
            if (
                runtime_events is not None
                and turn_delivery_factory.runtime_events is not runtime_events
            ):
                raise ValueError("turn delivery factory must use the agent runtime event bus")
            self.turn_delivery_factory = turn_delivery_factory
            self.runtime_events = turn_delivery_factory.runtime_events
        else:
            self.runtime_events = runtime_events or RuntimeEventBus()
            self.turn_delivery_factory = TurnDeliveryFactory(bus, self.runtime_events)
        self.runtime_event_publisher = self.turn_delivery_factory.runtime_event_publisher
        self.channels_config = channels_config
        self.restart_mode = restart_mode
        self._runtime_model_publisher = runtime_model_publisher
        self.workspace = workspace
        initial_model = model or provider.get_default_model()
        self.max_iterations = (
            max_iterations if max_iterations is not None else defaults.max_tool_iterations
        )
        initial_context_window = (
            context_window_tokens
            if context_window_tokens is not None
            else defaults.context_window_tokens
        )
        configured_presets = model_presets or {}
        self.runtime_resolver = ModelRuntimeResolver(
            LLMRuntime.capture(
                provider,
                initial_model,
                context_window_tokens=initial_context_window,
                snapshot_signature=provider_signature,
            ),
            model_presets=configured_presets,
            preset_catalog_loader=preset_catalog_loader,
            configured_default_preset=model_preset,
            provider_snapshot_loader=provider_snapshot_loader,
            preset_snapshot_loader=preset_snapshot_loader,
        )
        self.context_block_limit = context_block_limit
        self.max_tool_result_chars = (
            max_tool_result_chars
            if max_tool_result_chars is not None
            else defaults.max_tool_result_chars
        )
        self.provider_retry_mode = provider_retry_mode
        self.tool_hint_max_length = (
            tool_hint_max_length if tool_hint_max_length is not None
            else defaults.tool_hint_max_length
        )
        self.tools_config = _tc
        self.web_config = _tc.web
        self.exec_config = _tc.exec
        self._image_generation_provider_configs = dict(image_generation_provider_configs or {})
        if (
            image_generation_provider_config is not None
            and "openrouter" not in self._image_generation_provider_configs
        ):
            self._image_generation_provider_configs["openrouter"] = image_generation_provider_config
        self.cron_service = cron_service
        self.local_trigger_store = local_trigger_store
        self.restrict_to_workspace = restrict_to_workspace
        self.workspace_scopes = WorkspaceScopeResolver(
            default_workspace=workspace,
            default_restrict_to_workspace=restrict_to_workspace,
        )
        self._start_time = time.time()
        self._last_usage: dict[str, int] = {}
        self._extra_hooks: list[AgentHook] = hooks or []
        self._hook_factories: list[AgentTurnHookFactory] = hook_factories or []

        self.context = ContextBuilder(workspace, timezone=timezone, disabled_skills=disabled_skills)
        self.sessions = session_manager or SessionManager(workspace)
        self.sessions.set_file_cap_archiver(self.context.memory.raw_archive)
        self.tools = ToolRegistry()
        # One file-read/write tracker per logical session. The tool registry is
        # shared by this loop, so tools resolve the active state via contextvars.
        self._file_state_store = FileStateStore()
        self._exec_session_manager = ExecSessionManager()
        self.runner = AgentRunner()
        self.subagents = SubagentManager(
            workspace=workspace,
            bus=bus,
            tools_config=_tc,
            max_tool_result_chars=self.max_tool_result_chars,
            restrict_to_workspace=restrict_to_workspace,
            disabled_skills=disabled_skills,
            max_iterations=self.max_iterations,
            max_concurrent_subagents=max_concurrent_subagents,
            fail_on_tool_error=fail_on_tool_error,
            llm_wall_timeout_for_session=lambda sk: runner_wall_llm_timeout_s(self.sessions, sk),
        )
        self._unified_session = unified_session
        self._running = False
        self._mcp_servers = mcp_servers or {}
        self._mcp_stacks: dict[str, MCPConnection] = {}
        self._mcp_connecting = False
        self._runtime_context_providers: list[RuntimeContextProvider] = []
        self._active_tasks: dict[str, list[asyncio.Task]] = {}  # session_key -> tasks
        self._background_tasks: list[asyncio.Task] = []
        self._session_locks: dict[str, asyncio.Lock] = {}
        # Per-session pending queues for mid-turn message injection.
        # When a session has an active task, new messages for that session
        # are routed here instead of creating a new task.
        self._pending_queues: dict[str, asyncio.Queue] = {}
        self._deferred_automation_turns: dict[str, list[InboundMessage]] = {}
        self._cron_turns = CronTurnCoordinator(
            publish_inbound=self.bus.publish_inbound,
            dispatch=self._dispatch,
            is_running=lambda: self._running,
            deferred_queues=self._deferred_automation_turns,
        )
        self._local_trigger_turns = LocalTriggerTurnCoordinator(
            publish_inbound=self.bus.publish_inbound,
            dispatch=self._dispatch,
            is_running=lambda: self._running,
            deferred_queues=self._deferred_automation_turns,
        )
        self._automation_turn_coordinators = (
            ("cron", self._cron_turns),
            ("local trigger", self._local_trigger_turns),
        )
        # NANOBOT_MAX_CONCURRENT_REQUESTS: <=0 means unlimited; default 3.
        _max = int(os.environ.get("NANOBOT_MAX_CONCURRENT_REQUESTS", "3"))
        self._concurrency_gate: asyncio.Semaphore | None = (
            asyncio.Semaphore(_max) if _max > 0 else None
        )
        self.consolidator = Consolidator(
            store=self.context.memory,
            sessions=self.sessions,
            build_messages=self.context.build_messages,
            get_tool_definitions=self.tools.get_definitions,
            consolidation_ratio=consolidation_ratio,
            unified_session=unified_session,
        )
        self.auto_compact = AutoCompact(
            sessions=self.sessions,
            consolidator=self.consolidator,
            session_ttl_minutes=session_ttl_minutes,
        )
        if model_preset:
            self.set_model_preset(model_preset, publish_update=False)
        self._register_default_tools(provider_snapshot_loader=provider_snapshot_loader)
        self._runtime_vars: dict[str, Any] = {}
        self._current_iteration: int = 0
        self.commands = CommandRouter()
        register_builtin_commands(self.commands)

    @classmethod
    def from_config(
        cls,
        config: Any,
        bus: MessageBus | None = None,
        **extra: Any,
    ) -> AgentLoop:
        """Create an AgentLoop from config with the common parameter set.

        Extra keyword arguments are forwarded to ``AgentLoop.__init__``,
        allowing callers to override or extend the standard config-derived
        parameters (e.g. ``cron_service``, ``session_manager``).
        """
        from nanobot.providers.factory import make_provider

        if bus is None:
            bus = MessageBus()
        defaults = config.agents.defaults
        provider = extra.pop("provider", None) or make_provider(config)
        resolved = config.resolve_preset()
        model = extra.pop("model", None) or resolved.model
        context_window_tokens = extra.pop("context_window_tokens", None) or resolved.context_window_tokens
        provider_snapshot_loader = extra.pop("provider_snapshot_loader", None)
        preset_snapshot_loader = extra.pop("preset_snapshot_loader", None) or preset_helpers.make_preset_snapshot_loader(
            config,
            provider_snapshot_loader,
        )
        return cls(
            bus=bus,
            provider=provider,
            workspace=config.workspace_path,
            model=model,
            max_iterations=defaults.max_tool_iterations,
            max_concurrent_subagents=defaults.max_concurrent_subagents,
            context_window_tokens=context_window_tokens,
            context_block_limit=defaults.context_block_limit,
            max_tool_result_chars=defaults.max_tool_result_chars,
            fail_on_tool_error=defaults.fail_on_tool_error,
            provider_retry_mode=defaults.provider_retry_mode,
            tool_hint_max_length=defaults.tool_hint_max_length,
            restrict_to_workspace=config.tools.restrict_to_workspace,
            mcp_servers=config.tools.mcp_servers,
            channels_config=config.channels,
            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,
            model_presets=preset_helpers.configured_model_presets(config),
            model_preset=defaults.model_preset,
            restart_mode=config.gateway.restart_mode,
            provider_snapshot_loader=provider_snapshot_loader,
            preset_snapshot_loader=preset_snapshot_loader,
            **extra,
        )

    def _sync_subagent_runtime_limits(self) -> None:
        """Keep subagent runtime limits aligned with mutable loop settings."""
        self.subagents.max_iterations = self.max_iterations

    def invalidate_runtime_config(self) -> None:
        """Invalidate runtime config and notify clients to refresh its catalog."""
        self.runtime_resolver.invalidate()
        self._publish_runtime_selection(self.runtime_resolver.runtime)

    def runtime_for_session(
        self,
        session: Session,
        *,
        recover_removed: bool = True,
    ) -> LLMRuntime:
        """Resolve the immutable runtime selected by one session."""
        name = model_preset_from_metadata(session.metadata)
        if name is None:
            return self.llm_runtime()
        try:
            return self.runtime_resolver.resolve_preset(name)
        except KeyError:
            if not recover_removed or name in self.runtime_resolver.model_presets:
                raise
            logger.warning(
                "Session '{}' references removed model preset '{}'; falling back to default",
                session.key,
                name,
            )
            session.metadata.pop(SESSION_MODEL_PRESET_METADATA_KEY, None)
            self.sessions.save(session)
            return self.llm_runtime()

    def set_session_model_preset(
        self,
        session_key: str,
        name: str,
    ) -> LLMRuntime:
        """Validate and persist one session's preset selection."""
        runtime = self.runtime_resolver.resolve_preset(name)
        session = self.sessions.get_or_create(session_key)
        session.metadata[SESSION_MODEL_PRESET_METADATA_KEY] = runtime.model_preset
        self.sessions.save(session)
        return runtime

    def _publish_runtime_selection(
        self,
        runtime: LLMRuntime,
        *,
        publish_update: bool = True,
    ) -> None:
        if not publish_update:
            return
        if self._runtime_model_publisher is not None:
            self._runtime_model_publisher(runtime.model, runtime.model_preset)
        self._runtime_events().runtime_model_changed(
            runtime.model,
            runtime.model_preset,
        )

    def set_model_preset(
        self,
        name: str | None,
        *,
        publish_update: bool = True,
    ) -> LLMRuntime:
        """Select a named default runtime for future turns."""
        old_model = self.model
        runtime = self.runtime_resolver.select_preset(name)
        self._publish_runtime_selection(runtime, publish_update=publish_update)
        logger.info(
            "Runtime model switched for next turn: {} -> {}",
            old_model,
            runtime.model,
        )
        return runtime

    def set_runtime_model(self, model: str) -> LLMRuntime:
        """Select a model on the current provider for future turns."""
        return self.runtime_resolver.select_model(model)

    def set_runtime_context_window(self, context_window_tokens: int) -> LLMRuntime:
        """Select a context limit for future turns."""
        return self.runtime_resolver.select_context_window(context_window_tokens)

    def _register_default_tools(
        self,
        *,
        provider_snapshot_loader: Callable[..., ProviderSnapshot] | None,
    ) -> None:
        """Register the default set of tools via plugin loader."""
        from nanobot.agent.tools.context import ToolContext
        from nanobot.agent.tools.loader import ToolLoader

        ctx = ToolContext(
            config=self.tools_config,
            workspace=str(self.workspace),
            bus=self.bus,
            subagent_manager=self.subagents,
            cron_service=self.cron_service,
            exec_session_manager=self._exec_session_manager,
            sessions=self.sessions,
            provider_snapshot_loader=provider_snapshot_loader,
            image_generation_provider_configs=self._image_generation_provider_configs,
            timezone=self.context.timezone or "UTC",
            workspace_sandbox=self.workspace_scopes.sandbox_status,
            runtime_events=self.runtime_events,
        )
        loader = ToolLoader()
        registered = loader.load(ctx, self.tools)

        # MyTool needs runtime state reference — manual registration
        if self.tools_config.my.enable:
            self.tools.register(
                MyTool(runtime_state=self, modify_allowed=self.tools_config.my.allow_set)
            )
            registered.append("my")

        logger.info("Registered {} tools: {}", len(registered), registered)

    async def _connect_mcp(self) -> None:
        """Connect configured MCP servers."""
        await agent_context.connect_mcp(self, self.tools)

    def register_runtime_context_provider(
        self,
        provider: RuntimeContextProvider,
    ) -> None:
        """Register a provider resolved once before each inbound model turn."""
        if provider not in self._runtime_context_providers:
            self._runtime_context_providers.append(provider)

    def _runtime_events(self) -> RuntimeEventPublisher:
        return ensure_runtime_event_publisher(self)

    async def submit_cron_turn(self, msg: InboundMessage) -> OutboundMessage | None:
        return await self._cron_turns.submit(msg)

    async def submit_local_trigger_turn(self, msg: InboundMessage) -> OutboundMessage | None:
        return await self._local_trigger_turns.submit(msg)

    def pending_cron_job_ids_for_session(self, session_key: str) -> set[str]:
        return self._cron_turns.pending_job_ids_for_session(session_key)

    def pending_local_trigger_ids_for_session(self, session_key: str) -> set[str]:
        return self._local_trigger_turns.pending_trigger_ids_for_session(session_key)

    async def _publish_next_deferred_automation_turn(self, session_key: str) -> None:
        await publish_next_deferred_turn(
            deferred_queues=self._deferred_automation_turns,
            publish_inbound=self.bus.publish_inbound,
            session_key=session_key,
        )

    def _persist_user_message_early(
        self,
        msg: InboundMessage,
        session: Session,
        runtime_context_blocks: list[RuntimeContextBlock] | None = None,
        **kwargs: Any,
    ) -> bool:
        """Persist the triggering user message before the turn starts.

        Returns True if the message was persisted.
        """
        if not turn_continuation.should_persist_user_message(msg.metadata):
            return False
        media_paths = [p for p in (msg.media or []) if isinstance(p, str) and p]
        has_text = isinstance(msg.content, str) and msg.content.strip()
        if has_text or media_paths or runtime_context_blocks:
            extra: dict[str, Any] = ({"media": list(media_paths)} if media_paths else {}) | agent_context.session_extra(msg.metadata)
            extra.update(kwargs)
            text = msg.content if isinstance(msg.content, str) else ""
            text_override, automation_extra = automation_history_overrides(msg.metadata)
            if text_override is not None:
                text = text_override
            extra.update(automation_extra)
            text, runtime_context_meta = append_runtime_context(
                text,
                runtime_context_blocks or (),
            )
            if runtime_context_meta is not None:
                extra[RUNTIME_CONTEXT_HISTORY_META] = runtime_context_meta
            session.add_message("user", text, **extra)
            self._mark_pending_user_turn(session)
            self.sessions.save(session)
            return True
        return False

    def _build_initial_messages(self, ctx: TurnContext) -> list[dict[str, Any]]:
        """Build the initial message list for the LLM turn."""
        assert ctx.session is not None
        scope = self.workspace_scopes.for_message(ctx.msg, ctx.session.metadata)
        return self.context.build_messages(
            history=ctx.history,
            current_message=ctx.msg.content,
            media=ctx.msg.media if ctx.kind is TurnKind.USER and ctx.msg.media else None,
            channel=ctx.delivery.route.channel,
            chat_id=str(
                ctx.msg.metadata.get("context_chat_id") or ctx.delivery.route.chat_id
            ),
            current_role="user",
            sender_id=ctx.msg.sender_id,
            session_summary=ctx.pending_summary,
            session_metadata=ctx.session.metadata,
            workspace=scope.project_path,
            runtime_context_blocks=ctx.runtime_context_blocks,
            include_memory_recent_history=not ctx.ephemeral,
            session_key=ctx.session.key,
            unified_session=self._unified_session,
        )

    def _request_context_for_turn(self, ctx: TurnContext) -> RequestContext:
        assert ctx.session is not None
        scope = self.workspace_scopes.for_turn(
            channel=ctx.delivery.route.channel,
            message_metadata=ctx.msg.metadata,
            session_metadata=ctx.session.metadata,
        )
        return RequestContext(
            channel=ctx.delivery.route.channel,
            chat_id=ctx.delivery.route.chat_id,
            message_id=ctx.msg.metadata.get("message_id"),
            session_key=ctx.session_key,
            original_user_text=ctx.original_user_text,
            runtime=ctx.runtime,
            metadata=dict(ctx.msg.metadata or {}),
            sender_id=ctx.msg.sender_id,
            turn_id=ctx.turn_id,
            workspace=scope.project_path,
        )

    async def _resolve_runtime_context_for_turn(
        self,
        ctx: TurnContext,
    ) -> list[RuntimeContextBlock]:
        assert ctx.request_context is not None
        return await self._resolve_runtime_context_for_request(
            ctx.request_context,
            ctx.tools or self.tools,
        )

    async def _resolve_runtime_context_for_request(
        self,
        request: RequestContext,
        tools: ToolRegistry,
    ) -> list[RuntimeContextBlock]:
        providers = [
            *tools.get_runtime_context_providers(),
            *self._runtime_context_providers,
        ]
        blocks = runtime_context_blocks_from_metadata(request.metadata)
        blocks.extend(await resolve_runtime_context(providers, request))
        return blocks

    async def _dispatch_command_inline(
        self,
        msg: InboundMessage,
        key: str,
        raw: str,
        dispatch_fn: Callable[[CommandContext], Awaitable[OutboundMessage | None]],
    ) -> None:
        """Dispatch a command directly from the run() loop and publish the result."""
        ctx = CommandContext(msg=msg, session=None, key=key, raw=raw, loop=self)
        result = await dispatch_fn(ctx)
        if result:
            await self.bus.publish_outbound(result)
        else:
            logger.warning("Command '{}' matched but dispatch returned None", raw)

    async def _cancel_active_tasks(self, key: str) -> int:
        """Cancel and await all active tasks and subagents for *key*.

        Returns the total number of cancelled tasks + subagents.
        """
        tasks = self._active_tasks.pop(key, [])
        cancelled = sum(1 for t in tasks if not t.done() and t.cancel())
        for t in tasks:
            with suppress(asyncio.CancelledError, Exception):
                await t
        sub_cancelled = await self.subagents.cancel_by_session(key)
        return cancelled + sub_cancelled

    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

    def _remember_unified_session_route(
        self,
        session: Session,
        msg: InboundMessage,
        *,
        is_user_turn: bool,
    ) -> None:
        """Remember the latest user-facing route for unified-session delivery."""
        if (
            not self._unified_session
            or session.key != UNIFIED_SESSION_KEY
            or not is_user_turn
            or msg.channel in {"cli", "system"}
            or msg.sender_id == "subagent"
        ):
            return
        _, automation_metadata = automation_history_overrides(msg.metadata)
        if automation_metadata:
            return
        remember_last_channel(session.metadata, msg.channel, msg.chat_id)

    @staticmethod
    def _replay_token_budget(runtime: LLMRuntime) -> int:
        """Derive a token budget for session history replay from the context window."""
        if runtime.context_window_tokens <= 0:
            return 0
        max_output = runtime.generation.max_tokens
        try:
            reserved_output = int(max_output)
        except (TypeError, ValueError):
            reserved_output = 4096
        budget = runtime.context_window_tokens - max(1, reserved_output) - 1024
        return budget if budget > 0 else max(128, runtime.context_window_tokens // 2)

    async def _run_agent_loop(
        self,
        initial_messages: list[dict],
        on_progress: Callable[..., Awaitable[None]] | None = None,
        on_stream: Callable[[str], Awaitable[None]] | None = None,
        on_stream_end: Callable[..., Awaitable[None]] | None = None,
        on_retry_wait: Callable[[str], Awaitable[None]] | None = None,
        *,
        runtime: LLMRuntime,
        session: Session | None = None,
        channel: str = "cli",
        chat_id: str = "direct",
        message_id: str | None = None,
        metadata: dict[str, Any] | None = None,
        session_key: str | None = None,
        original_user_text: str | None = None,
        pending_queue: asyncio.Queue | None = None,
        ephemeral: bool = False,
        run_extra_hooks_for_ephemeral: bool = False,
        hooks: list[AgentHook] | None = None,
        hook_factories: list[AgentTurnHookFactory] | None = None,
        turn_scopes: list[AbstractContextManager[Any]] | None = None,
        tools: ToolRegistry | None = None,
        request_context: RequestContext | None = None,
    ) -> tuple[str | None, list[str], list[dict], str, bool]:
        """Run the agent iteration loop.

        *on_stream*: called with each content delta during streaming.
        *on_stream_end(resuming)*: called when a streaming session finishes.
        ``resuming=True`` means tool calls follow (spinner should restart);
        ``resuming=False`` means this is the final response.

        Returns (final_content, tools_used, messages, stop_reason, had_injections).
        """
        self._sync_subagent_runtime_limits()

        async def _checkpoint(payload: dict[str, Any]) -> None:
            if session is None:
                return
            self._set_runtime_checkpoint(session, payload)

        async def _drain_pending(*, limit: int = _MAX_INJECTIONS_PER_TURN) -> list[dict[str, Any]]:
            """Drain follow-up messages from the pending queue.

            When no messages are immediately available but sub-agents
            spawned in this dispatch are still running, blocks until at
            least one result arrives (or timeout).  This keeps the runner
            loop alive so subsequent sub-agent completions are consumed
            in-order rather than dispatched separately.
            """
            if pending_queue is None:
                return []

            async def _to_user_message(pending_msg: InboundMessage) -> dict[str, Any]:
                content = pending_msg.content
                media = pending_msg.media if pending_msg.media else None
                if media:
                    content, media = self._prepare_message_media(content, media)
                    media = media or None
                user_content = self.context._build_user_content(content, media)
                row: dict[str, Any] = {"role": "user", "content": user_content}
                metadata = pending_msg.metadata if isinstance(pending_msg.metadata, dict) else {}
                if pending_msg.channel != "system":
                    scope = self.workspace_scopes.for_turn(
                        channel=pending_msg.channel,
                        message_metadata=metadata,
                        session_metadata=session.metadata if session is not None else None,
                    )
                    pending_request = RequestContext(
                        channel=pending_msg.channel,
                        chat_id=pending_msg.chat_id,
                        message_id=metadata.get("message_id"),
                        session_key=active_session_key,
                        original_user_text=pending_msg.content,
                        runtime=runtime,
                        metadata=dict(metadata),
                        sender_id=pending_msg.sender_id,
                        turn_id=request_ctx.turn_id,
                        workspace=scope.project_path,
                    )
                    blocks = await self._resolve_runtime_context_for_request(
                        pending_request,
                        effective_tools,
                    )
                    row["content"], marker = append_runtime_context(user_content, blocks)
                    if marker is not None:
                        row["_meta"] = {RUNTIME_CONTEXT_MESSAGE_META: marker}
                if (
                    pending_msg.sender_id == "subagent"
                    and metadata.get("injected_event") == "subagent_result"
                ):
                    marker: dict[str, Any] = {"kind": "subagent_result"}
                    task_id = metadata.get("subagent_task_id")
                    if isinstance(task_id, str) and task_id:
                        marker["subagent_task_id"] = task_id
                        row["subagent_task_id"] = task_id
                    row[HIDDEN_HISTORY_META] = marker
                    row["injected_event"] = "subagent_result"
                return row

            items: list[dict[str, Any]] = []
            while len(items) < limit:
                try:
                    items.append(await _to_user_message(pending_queue.get_nowait()))
                except asyncio.QueueEmpty:
                    break

            # Block if nothing drained but sub-agents spawned in this dispatch
            # are still running.  Keeps the runner loop alive so subsequent
            # completions are injected in-order rather than dispatched separately.
            if (not items
                    and session is not None
                    and self.subagents.get_running_count_by_session(session.key) > 0):
                try:
                    msg = await asyncio.wait_for(pending_queue.get(), timeout=300)
                except asyncio.TimeoutError:
                    logger.warning(
                        "Timeout waiting for sub-agent completion in session {}",
                        session.key,
                    )
                    return items
                items.append(await _to_user_message(msg))
                while len(items) < limit:
                    try:
                        items.append(await _to_user_message(pending_queue.get_nowait()))
                    except asyncio.QueueEmpty:
                        break

            return items

        active_session_key = session.key if session else session_key
        effective_scope = self.workspace_scopes.for_turn(
            channel=channel,
            message_metadata=metadata,
            session_metadata=session.metadata if session is not None else None,
        )
        effective_tools = tools or self.tools
        request_ctx = request_context or RequestContext(
            channel=channel,
            chat_id=chat_id,
            message_id=message_id,
            session_key=active_session_key,
            original_user_text=original_user_text,
            runtime=runtime,
            metadata=dict(metadata or {}),
            workspace=effective_scope.project_path,
        )
        file_state_token = bind_file_states(self._file_state_store.for_session(active_session_key))
        request_token = bind_request_context(request_ctx)
        workspace_token = bind_workspace_scope(effective_scope)
        turn_scope_stack = ExitStack()
        # Compute lazily because create_goal may create goal metadata during this run.
        def _goal_continue() -> str | None:
            _goal_lines = goal_state_runtime_lines(session.metadata if session is not None else None)
            if not _goal_lines:
                return None
            return (
                "You have an active sustained goal:\n\n"
                + "\n".join(_goal_lines)
                + "\n\nPlease continue working toward the objective using your tools, "
                "or call update_goal with action='complete' if the work is truly finished."
            )

        session_metadata = session.metadata if session is not None else None
        try:
            for scope in turn_scopes or ():
                turn_scope_stack.enter_context(scope)
            hook = build_agent_turn_hook(AgentTurnHookSpec(
                on_progress=on_progress,
                on_stream=on_stream,
                on_stream_end=on_stream_end,
                channel=channel,
                chat_id=chat_id,
                message_id=message_id,
                metadata=metadata,
                session_key=active_session_key,
                workspace=effective_scope.project_path,
                tool_hint_max_length=self.tool_hint_max_length,
                on_iteration=lambda iteration: setattr(self, "_current_iteration", iteration),
                registered_hook_factories=self._hook_factories,
                turn_hook_factories=list(hook_factories or []),
                registered_hooks=self._extra_hooks,
                turn_hooks=list(hooks or []),
                ephemeral=ephemeral,
                run_extra_hooks_for_ephemeral=run_extra_hooks_for_ephemeral,
            ))
            result = await self.runner.run(AgentRunSpec(
                initial_messages=initial_messages,
                tools=effective_tools,
                runtime=runtime,
                max_iterations=self.max_iterations,
                max_tool_result_chars=self.max_tool_result_chars,
                hook=hook,
                error_message="Sorry, I encountered an error calling the AI model.",
                concurrent_tools=True,
                workspace=effective_scope.project_path,
                session_key=session.key if session else None,
                context_block_limit=self.context_block_limit,
                provider_retry_mode=self.provider_retry_mode,
                progress_callback=on_progress,
                stream_progress_deltas=on_stream is not None,
                retry_wait_callback=on_retry_wait,
                checkpoint_callback=_checkpoint,
                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=runner_wall_llm_timeout_s(
                    self.sessions,
                    session.key if session is not None else session_key,
                    metadata=session_metadata,
                    message_metadata=metadata,
                ),
                goal_active_predicate=lambda: sustained_goal_active(session.metadata) if session is not None else False,
                goal_continue_message=_goal_continue,
                finalize_on_max_iterations=turn_continuation.should_finalize_on_max_iterations(
                    pending_queue_available=pending_queue is not None and session is not None,
                    session_metadata=session_metadata,
                    message_metadata=metadata,
                ),
            ))
        finally:
            turn_scope_stack.close()
            reset_workspace_scope(workspace_token)
            reset_request_context(request_token)
            reset_file_states(file_state_token)
        self._last_usage = result.usage
        if result.stop_reason == "max_iterations":
            logger.warning("Max iterations ({}) reached", self.max_iterations)
            should_stream = turn_continuation.should_stream_budget_response(
                stop_reason=result.stop_reason,
                pending_queue_available=pending_queue is not None and session is not None,
                session_metadata=session_metadata,
                message_metadata=metadata,
            )
            # 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 and should_stream:
                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

    async def run(self) -> None:
        """Run the agent loop, dispatching messages as tasks to stay responsive to /stop."""
        self._running = True
        try:
            await self._connect_mcp()
            logger.info("Agent loop started")

            while self._running:
                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.runtime_for_session,
                        active_session_keys=self._pending_queues.keys(),
                    )
                    continue
                except asyncio.CancelledError:
                    # Preserve real task cancellation so shutdown can complete cleanly.
                    # Only ignore non-task CancelledError signals that may leak from integrations.
                    if not self._running or task_is_cancelling():
                        raise
                    logger.warning(
                        "Ignoring leaked CancelledError while consuming inbound messages"
                    )
                    continue
                except Exception as e:
                    logger.warning("Error consuming inbound message: {}, continuing...", e)
                    continue

                raw = msg.content.strip()
                effective_key = self._effective_session_key(msg)
                if await agent_context.handle_runtime_control(self, msg, self.tools):
                    continue
                if self.commands.is_priority(raw):
                    await self._dispatch_command_inline(
                        msg, effective_key, raw,
                        self.commands.dispatch_priority,
                    )
                    continue
                deferred = False
                for label, coordinator in self._automation_turn_coordinators:
                    if coordinator.defer_if_active(
                        msg,
                        session_key=effective_key,
                        active_session_keys=self._pending_queues.keys(),
                    ):
                        logger.info(
                            "Deferred {} turn for active session {}",
                            label,
                            effective_key,
                        )
                        deferred = True
                        break
                if deferred:
                    continue
                # 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 effective_key in self._pending_queues:
                    # Non-priority commands must not be queued for injection;
                    # dispatch them directly (same pattern as priority commands).
                    if self.commands.is_dispatchable_command(raw):
                        await self._dispatch_command_inline(
                            msg, effective_key, raw,
                            self.commands.dispatch,
                        )
                        continue
                    pending_msg = msg
                    if effective_key != msg.session_key:
                        pending_msg = dataclasses.replace(
                            msg,
                            session_key_override=effective_key,
                        )
                    try:
                        self._pending_queues[effective_key].put_nowait(pending_msg)
                    except asyncio.QueueFull:
                        logger.warning(
                            "Pending queue full for session {}, falling back to queued task",
                            effective_key,
                        )
                    else:
                        logger.info(
                            "Routed follow-up message to pending queue for session {}",
                            effective_key,
                        )
                        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(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
                )
        finally:
            # MCP stdio transports use AnyIO cancel scopes; close them from the task that opened them.
            await self.close_mcp()

    async def _dispatch(self, msg: InboundMessage) -> None:
        """Process a message: per-session serial, cross-session concurrent."""
        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()

        delivery = self.turn_delivery_factory.unrouted(msg, session_key)
        pending: asyncio.Queue | None = None
        try:
            async with lock, gate:
                # Only the task that owns the session lock may publish the
                # active mid-turn injection queue for this session.
                pending = asyncio.Queue(maxsize=20)
                self._pending_queues[session_key] = pending
                try:
                    delivery = self.turn_delivery_factory.create(
                        msg,
                        session_key,
                        enable_stream=True,
                    )
                    response = await self._process_message(
                        msg,
                        on_stream=delivery.on_stream,
                        on_stream_end=delivery.on_stream_end,
                        pending_queue=pending,
                        delivery=delivery,
                    )
                    continuing = turn_continuation.internal_continuation_pending(msg.metadata)
                    await delivery.complete(
                        response,
                        publish_completion=not continuing,
                    )
                    for _, coordinator in self._automation_turn_coordinators:
                        coordinator.complete(msg, response=response)
                except asyncio.CancelledError:
                    for _, coordinator in self._automation_turn_coordinators:
                        coordinator.complete(msg, error=asyncio.CancelledError())
                    logger.info("Task cancelled for session {}", session_key)
                    # Preserve partial context from the interrupted turn so
                    # the user does not lose tool results and assistant
                    # messages accumulated before /stop.  The checkpoint was
                    # already persisted to session metadata by
                    # _emit_checkpoint during tool execution; materializing
                    # it into session history now makes it visible in the
                    # next conversation turn.
                    try:
                        key = self._effective_session_key(msg)
                        session = self.sessions.get_or_create(key)
                        if self._restore_runtime_checkpoint(session):
                            self._clear_pending_user_turn(session)
                            self.sessions.save(session)
                            logger.info(
                                "Restored partial context for cancelled session {}",
                                key,
                            )
                    except Exception:
                        logger.debug(
                            "Could not restore checkpoint for cancelled session {}",
                            session_key,
                            exc_info=True,
                        )
                    raise
                except Exception as exc:
                    logger.exception("Error processing message for session {}", session_key)
                    await delivery.fail(
                        publish_completion=not turn_continuation.internal_continuation_pending(
                            msg.metadata
                        )
                    )
                    for _, coordinator in self._automation_turn_coordinators:
                        coordinator.complete(msg, error=exc)
                finally:
                    # Drain any messages still in the pending queue and re-publish
                    # them to the bus so they are processed as fresh inbound messages
                    # rather than silently lost.  Only remove our own queue; a
                    # later task waiting on the lock must not be able to steal
                    # cleanup ownership.
                    queue = None
                    if self._pending_queues.get(session_key) is pending:
                        queue = self._pending_queues.pop(session_key, None)
                    else:
                        queue = pending
                    if queue is not None:
                        leftover = 0
                        while True:
                            try:
                                item = queue.get_nowait()
                            except asyncio.QueueEmpty:
                                break
                            await self.bus.publish_inbound(item)
                            leftover += 1
                        if leftover:
                            logger.info(
                                "Re-published {} leftover message(s) to bus for session {}",
                                leftover, session_key,
                            )
                    if not turn_continuation.internal_continuation_pending(msg.metadata):
                        await delivery.idle()
                    await self._publish_next_deferred_automation_turn(session_key)
        finally:
            if pending is None:
                await delivery.idle()
                await self._publish_next_deferred_automation_turn(session_key)

    async def close_mcp(self) -> None:
        """Drain background work, stop exec sessions, then close MCP connections."""
        if self._background_tasks:
            await asyncio.gather(*self._background_tasks, return_exceptions=True)
            self._background_tasks.clear()
        errors: list[BaseException] = []
        cleanup_steps = (
            self.subagents.close,
            self._exec_session_manager.close_all,
            lambda: agent_context.close_mcp(self),
        )
        for cleanup in cleanup_steps:
            try:
                await cleanup()
            except BaseException as exc:
                errors.append(exc)
        if len(errors) == 1:
            raise errors[0]
        if errors:
            raise BaseExceptionGroup("failed to close agent resources", errors)

    def _schedule_background(self, coro) -> None:
        """Schedule a coroutine as a tracked background task (drained on shutdown)."""
        task = asyncio.create_task(coro)
        self._background_tasks.append(task)
        task.add_done_callback(self._background_tasks.remove)

    def stop(self) -> None:
        """Stop the agent loop."""
        self._running = False
        logger.info("Agent loop stopping")

    async def _process_message(
        self,
        msg: InboundMessage,
        session_key: str | None = None,
        on_progress: Callable[..., Awaitable[None]] | None = None,
        on_stream: Callable[[str], Awaitable[None]] | None = None,
        on_stream_end: Callable[..., Awaitable[None]] | None = None,
        pending_queue: asyncio.Queue | None = None,
        ephemeral: bool = False,
        run_extra_hooks_for_ephemeral: bool = False,
        hooks: list[AgentHook] | None = None,
        hook_factories: list[AgentTurnHookFactory] | None = None,
        tools: ToolRegistry | None = None,
        runtime: LLMRuntime | None = None,
        delivery: TurnDelivery | None = None,
        on_runtime_admitted: Callable[[LLMRuntime], Awaitable[None]] | None = None,
    ) -> OutboundMessage | None:
        """Process a single inbound message and return the response."""
        kind = TurnKind.SYSTEM if msg.channel == "system" else TurnKind.USER
        if kind is TurnKind.SYSTEM:
            destination = (
                msg.chat_id.split(":", 1) if ":" in msg.chat_id else ("cli", msg.chat_id)
            )
            key = session_key or msg.session_key_override or f"{destination[0]}:{destination[1]}"
        else:
            key = session_key or msg.session_key
        if delivery is None:
            delivery = self.turn_delivery_factory.create(msg, key)
        elif delivery.session_key != key:
            raise ValueError("turn delivery session does not match the processing session")
        if on_stream is None:
            on_stream = delivery.on_stream
        if on_stream_end is None:
            on_stream_end = delivery.on_stream_end
        t0 = time.time()
        ctx = TurnContext(
            msg=msg,
            session=None,
            session_key=key,
            state=TurnState.RESTORE,
            turn_id=f"{key}:{time.time_ns()}",
            runtime=runtime,
            kind=kind,
            delivery=delivery,
            original_user_text=(
                None
                if kind is TurnKind.SYSTEM
                or turn_continuation.internal_continuation_inbound(msg.metadata)
                else msg.content
            ),
            turn_wall_started_at=t0,
            visible_run_started_at=turn_continuation.internal_continuation_run_started_at(
                msg.metadata,
            ),
            on_progress=on_progress,
            on_stream=on_stream,
            on_stream_end=on_stream_end,
            on_runtime_admitted=on_runtime_admitted,
            pending_queue=pending_queue,
            ephemeral=ephemeral,
            run_extra_hooks_for_ephemeral=run_extra_hooks_for_ephemeral,
            hooks=list(hooks or []),
            hook_factories=list(hook_factories or []),
            tools=tools,
        )
        # A streaming callback may be present even when the final text comes from a
        # non-streaming recovery. Only the last completed segment can suppress the
        # regular outbound message.
        if ctx.on_stream is not None:
            stream_callback = ctx.on_stream
            stream_end_callback = ctx.on_stream_end
            segment_streamed_content = False

            async def _tracked_stream(delta: str) -> None:
                nonlocal segment_streamed_content
                if delta:
                    segment_streamed_content = True
                await stream_callback(delta)

            async def _tracked_stream_end(*, resuming: bool = False) -> None:
                nonlocal segment_streamed_content
                ctx.streamed_content = segment_streamed_content
                segment_streamed_content = False
                if stream_end_callback is not None:
                    await stream_end_callback(resuming=resuming)

            ctx.on_stream = _tracked_stream
            ctx.on_stream_end = _tracked_stream_end

        while ctx.state is not TurnState.DONE:
            handler_name = f"_state_{ctx.state.name.lower()}"
            handler = getattr(self, handler_name, None)
            if handler is None:
                raise RuntimeError(f"Missing state handler for {ctx.state}")

            t0 = time.perf_counter()
            try:
                event = await handler(ctx)
            except Exception:
                duration = (time.perf_counter() - t0) * 1000
                ctx.trace.append(
                    StateTraceEntry(
                        state=ctx.state,
                        started_at=t0,
                        duration_ms=duration,
                        event="",
                        error="exception",
                    )
                )
                raise

            duration = (time.perf_counter() - t0) * 1000
            ctx.trace.append(
                StateTraceEntry(
                    state=ctx.state,
                    started_at=t0,
                    duration_ms=duration,
                    event=event,
                )
            )
            logger.debug(
                "[turn {}] State {} took {:.1f}ms -> event {}",
                ctx.turn_id,
                ctx.state.name,
                duration,
                event,
            )

            next_state = self._TRANSITIONS.get((ctx.state, event))
            if next_state is None:
                raise RuntimeError(
                    f"[turn {ctx.turn_id}] No transition from {ctx.state} "
                    f"on event {event!r}"
                )
            ctx.state = next_state

        logger.debug(
            "[turn {}] Turn completed after {} states",
            ctx.turn_id,
            len(ctx.trace),
        )
        return ctx.outbound

    def _assemble_outbound(
        self,
        msg: InboundMessage,
        final_content: str,
        all_msgs: list[dict[str, Any]],
        stop_reason: str,
        had_injections: bool,
        streamed_content: bool,
        *,
        turn_latency_ms: int | None = None,
    ) -> OutboundMessage | None:
        """Assemble the final outbound message from turn results."""
        # MessageTool suppression
        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)

        event = None
        meta = dict(msg.metadata or {})
        if streamed_content and stop_reason not in {"error", "tool_error"}:
            event = StreamedResponseEvent()
        if turn_latency_ms is not None:
            meta["latency_ms"] = int(turn_latency_ms)

        return OutboundMessage(
            channel=msg.channel,
            chat_id=msg.chat_id,
            content=final_content,
            event=event,
            metadata=meta,
        )

    async def _state_restore(self, ctx: TurnContext) -> TurnState:
        """Restore checkpoint / pending user turn; extract documents."""
        msg = ctx.msg

        if ctx.kind is TurnKind.USER and msg.media:
            new_content, image_only = self._prepare_message_media(msg.content, msg.media)
            ctx.msg = dataclasses.replace(msg, content=new_content, media=image_only)
            msg = ctx.msg

        preview = msg.content[:80] + "..." if len(msg.content) > 80 else msg.content
        if ctx.kind is TurnKind.SYSTEM:
            logger.info("Processing system message from {}", msg.sender_id)
        else:
            logger.info("Processing message from {}:{}: {}", msg.channel, msg.sender_id, preview)

        # Session is already fetched by the caller (_process_message) but
        # ensure it exists in case this handler is invoked independently.
        if ctx.session is None:
            ctx.session = self.sessions.get_or_create(ctx.session_key)
        self._remember_unified_session_route(
            ctx.session,
            msg,
            is_user_turn=ctx.kind is TurnKind.USER,
        )
        await ctx.delivery.started()
        if ctx.kind is TurnKind.USER:
            self.workspace_scopes.persist_message_scope(ctx.session, msg)

        if self._restore_runtime_checkpoint(ctx.session):
            self.sessions.save(ctx.session)
        if self._restore_pending_user_turn(ctx.session):
            self.sessions.save(ctx.session)

        return "ok"

    def _prepare_message_media(self, content: str, media: list[str]) -> tuple[str, list[str]]:
        if self._should_extract_document_text():
            return extract_documents(content, media)
        return reference_non_image_attachments(content, media)

    def _should_extract_document_text(self) -> bool:
        if self.channels_config is None:
            return True
        return self.channels_config.extract_document_text

    async def _state_compact(self, ctx: TurnContext) -> str:
        ctx.session, pending = self.auto_compact.prepare_session(ctx.session, ctx.session_key)
        ctx.pending_summary = pending
        return "ok"

    async def _state_command(self, ctx: TurnContext) -> str:
        if ctx.kind is TurnKind.SYSTEM:
            return "dispatch"
        raw = ctx.msg.content.strip()
        _, automation_metadata = automation_history_overrides(ctx.msg.metadata)
        is_user_turn = (
            ctx.original_user_text is not None
            and not automation_metadata
            and ctx.msg.channel != "system"
            and ctx.msg.sender_id != "subagent"
        )
        cmd_ctx = CommandContext(
            msg=ctx.msg,
            session=ctx.session,
            key=ctx.session_key,
            raw=raw,
            loop=self,
            runtime=ctx.runtime,
            is_user_turn=is_user_turn,
            turn_scopes=ctx.turn_scopes,
        )
        result = await self.commands.dispatch(cmd_ctx)
        if result is not None:
            ctx.outbound = result
            # Shortcut commands skip BUILD and SAVE, so we must persist the
            # turn here so WebUI history hydration after _turn_end sees the
            # message.  Mark messages with _command so get_history can filter
            # them out of LLM context.  /new is excluded because it
            # intentionally clears the session.
            if cmd_ctx.raw.lower() != "/new":
                ctx.input_persisted_early = self._persist_user_message_early(
                    ctx.msg, ctx.session, _command=True
                )
                ctx.session.add_message(
                    "assistant", result.content, _command=True
                )
                self.sessions.save(ctx.session)
                self._clear_pending_user_turn(ctx.session)
            return "shortcut"
        return "dispatch"

    async def _state_build(self, ctx: TurnContext) -> str:
        runtime = ctx.runtime
        if runtime is None:
            runtime = self.runtime_for_session(ctx.session)
            ctx.runtime = runtime
        if ctx.on_runtime_admitted is not None:
            await ctx.on_runtime_admitted(runtime)
        replay_max_messages = replay_max_messages_for_context(
            runtime.context_window_tokens
        )
        if not ctx.ephemeral:
            await self.consolidator.maybe_consolidate_by_tokens(
                ctx.session,
                runtime=runtime,
                replay_max_messages=replay_max_messages,
            )
        is_subagent = ctx.kind is TurnKind.SYSTEM and ctx.msg.sender_id == "subagent"

        if ctx.kind is TurnKind.USER and (message_tool := self.tools.get("message")):
            if isinstance(message_tool, MessageTool):
                message_tool.start_turn()

        _hist_kwargs: dict[str, Any] = {
            "max_messages": replay_max_messages,
            "max_tokens": self._replay_token_budget(runtime),
            "extend_to_user": is_subagent,
        }
        ctx.history = ctx.session.get_history(**_hist_kwargs)
        if is_subagent:
            # Keep the durable internal delivery as an assistant record, but
            # present this completion to the model as fresh follow-up input.
            # Providers without assistant-prefill support drop trailing
            # assistant messages, so using the persisted record as the current
            # prompt would hide an independently dispatched subagent result.
            if self._persist_subagent_followup(ctx.session, ctx.msg):
                logger.debug("Subagent result persisted for session {}", ctx.session_key)
                self.sessions.save(ctx.session)
            ctx.input_persisted_early = True
        ctx.delivery.record_runtime(ctx.runtime)

        ctx.request_context = self._request_context_for_turn(ctx)
        if ctx.kind is TurnKind.USER:
            ctx.runtime_context_blocks = await self._resolve_runtime_context_for_turn(ctx)
        ctx.initial_messages = self._build_initial_messages(ctx)
        if ctx.kind is TurnKind.USER:
            ctx.input_persisted_early = self._persist_user_message_early(
                ctx.msg,
                ctx.session,
                runtime_context_blocks=ctx.runtime_context_blocks,
            )

        if ctx.on_progress is None:
            ctx.on_progress = ctx.delivery.progress_callback()
        if ctx.on_retry_wait is None:
            ctx.on_retry_wait = ctx.delivery.retry_wait_callback()

        return "ok"

    async def _state_run(self, ctx: TurnContext) -> str:
        if ctx.visible_run_started_at is None:
            ctx.visible_run_started_at = time.time()
        await ctx.delivery.running(started_at=ctx.visible_run_started_at)
        result = await self._run_agent_loop(
            ctx.initial_messages,
            runtime=ctx.runtime,
            on_progress=ctx.on_progress,
            on_stream=ctx.on_stream,
            on_stream_end=ctx.on_stream_end,
            on_retry_wait=ctx.on_retry_wait,
            session=ctx.session,
            channel=ctx.delivery.route.channel,
            chat_id=ctx.delivery.route.chat_id,
            message_id=ctx.msg.metadata.get("message_id"),
            metadata=ctx.msg.metadata,
            session_key=ctx.session_key,
            original_user_text=ctx.original_user_text,
            pending_queue=ctx.pending_queue,
            ephemeral=ctx.ephemeral,
            run_extra_hooks_for_ephemeral=ctx.run_extra_hooks_for_ephemeral,
            hooks=ctx.hooks,
            hook_factories=ctx.hook_factories,
            turn_scopes=ctx.turn_scopes,
            tools=ctx.tools,
            request_context=ctx.request_context,
        )
        final_content, tools_used, all_msgs, stop_reason, had_injections = result
        ctx.final_content = final_content
        ctx.tools_used = tools_used
        ctx.all_messages = all_msgs
        ctx.stop_reason = stop_reason
        ctx.had_injections = had_injections
        if ctx.kind is TurnKind.USER:
            await turn_continuation.maybe_continue_turn(ctx)
        return "ok"

    async def _state_save(self, ctx: TurnContext) -> str:
        turn_continuation.prepare_save_boundary(ctx)

        if (
            ctx.kind is TurnKind.USER
            and (ctx.final_content is None or not ctx.final_content.strip())
            and not ctx.suppress_response
        ):
            ctx.final_content = EMPTY_FINAL_RESPONSE_MESSAGE

        latency_started_at = (
            ctx.visible_run_started_at
            if (
                ctx.kind is TurnKind.SYSTEM
                or turn_continuation.internal_continuation_inbound(ctx.msg.metadata)
            )
            and ctx.visible_run_started_at is not None
            else ctx.turn_wall_started_at
        )
        ctx.turn_latency_ms = max(0, int((time.time() - latency_started_at) * 1000))
        self._save_turn(
            ctx.session, ctx.all_messages, ctx.save_skip,
            turn_latency_ms=ctx.turn_latency_ms,
        )
        ctx.delivery.record_latency(ctx.turn_latency_ms)
        if not ctx.ephemeral:
            ctx.session.enforce_file_cap(
                on_archive=partial(self.context.memory.raw_archive, session_key=ctx.session_key)
            )
            self._schedule_background(
                self.consolidator.maybe_consolidate_by_tokens(
                    ctx.session,
                    runtime=ctx.runtime,
                    replay_max_messages=replay_max_messages_for_context(
                        ctx.runtime.context_window_tokens
                    ),
                )
            )
        self._clear_pending_user_turn(ctx.session)
        self._clear_runtime_checkpoint(ctx.session)
        self.sessions.save(ctx.session)
        return "ok"

    async def _state_respond(self, ctx: TurnContext) -> str:
        if ctx.suppress_response:
            ctx.outbound = None
            return "ok"
        if ctx.kind is TurnKind.SYSTEM:
            ctx.outbound = ctx.delivery.background_response(
                ctx.final_content,
                stop_reason=ctx.stop_reason,
                streamed=ctx.streamed_content,
                latency_ms=ctx.turn_latency_ms,
            )
            return "ok"
        ctx.outbound = self._assemble_outbound(
            ctx.msg,
            ctx.final_content,
            ctx.all_messages,
            ctx.stop_reason,
            ctx.had_injections,
            ctx.streamed_content,
            turn_latency_ms=ctx.turn_latency_ms,
        )
        if ctx.ephemeral and ctx.outbound is not None:
            ctx.outbound.metadata["_stop_reason"] = ctx.stop_reason
        return "ok"

    def _sanitize_persisted_blocks(
        self,
        content: list[dict[str, Any]],
        *,
        should_truncate_text: bool = False,
    ) -> list[dict[str, Any]]:
        """Strip volatile multimodal payloads before writing session history."""
        filtered: list[dict[str, Any]] = []
        for block in content:
            if not isinstance(block, dict):
                filtered.append(block)
                continue

            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 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

            filtered.append(block)

        return filtered

    def _save_turn(
        self,
        session: Session,
        messages: list[dict],
        skip: int,
        *,
        turn_latency_ms: int | None = None,
    ) -> None:
        """Save new-turn messages into session, truncating large tool results."""
        from datetime import datetime

        declared_tool_call_ids = {
            str(tc["id"])
            for m in session.messages
            if m.get("role") == "assistant"
            for tc in m.get("tool_calls") or []
            if isinstance(tc, dict) and tc.get("id")
        }
        fulfilled_tool_call_ids = {
            str(m["tool_call_id"])
            for m in session.messages
            if m.get("role") == "tool" and m.get("tool_call_id")
        }
        last_assistant_idx: int | None = None
        for m in messages[skip:]:
            entry = dict(m)
            internal_meta = entry.pop("_meta", None)
            runtime_context_meta = (
                internal_meta.get(RUNTIME_CONTEXT_MESSAGE_META)
                if isinstance(internal_meta, dict)
                else None
            )
            role, content = entry.get("role"), entry.get("content")
            if role == "assistant" and not content and not entry.get("tool_calls"):
                continue  # skip empty assistant messages — they poison session context
            if role == "tool":
                tool_call_id = entry.get("tool_call_id")
                tool_call_id_str = str(tool_call_id) if tool_call_id else ""
                if (
                    not tool_call_id_str
                    or tool_call_id_str not in declared_tool_call_ids
                    or tool_call_id_str in fulfilled_tool_call_ids
                ):
                    # Undeclared tool results corrupt future provider requests.
                    logger.warning(
                        "Dropping invalid tool result {} from session {} during persistence",
                        tool_call_id_str or "(missing id)",
                        session.key,
                    )
                    continue
                fulfilled_tool_call_ids.add(tool_call_id_str)
                if isinstance(content, str) and len(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, should_truncate_text=True)
                    if not filtered:
                        # Preserve the tool_call/result pair after block filtering.
                        filtered = [
                            {"type": "text", "text": "[tool result omitted during persistence]"}
                        ]
                    entry["content"] = filtered
            elif role == "user":
                if isinstance(content, list):
                    filtered = self._sanitize_persisted_blocks(content)
                    if not filtered:
                        continue
                    entry["content"] = filtered
                if isinstance(runtime_context_meta, dict):
                    entry[RUNTIME_CONTEXT_HISTORY_META] = runtime_context_meta
            entry.setdefault("timestamp", datetime.now().isoformat())
            session.messages.append(entry)
            if role == "assistant":
                last_assistant_idx = len(session.messages) - 1
                declared_tool_call_ids.update(
                    str(tc["id"])
                    for tc in entry.get("tool_calls") or []
                    if isinstance(tc, dict) and tc.get("id")
                )
        if turn_latency_ms is not None and last_assistant_idx is not None:
            session.messages[last_assistant_idx]["latency_ms"] = int(turn_latency_ms)
        session.updated_at = datetime.now()

    def _persist_subagent_followup(self, session: Session, msg: InboundMessage) -> bool:
        """Persist subagent follow-ups before prompt assembly so history stays durable.

        Returns True if a new entry was appended; False if the follow-up was
        deduped (same ``subagent_task_id`` already in session) or carries no
        content worth persisting.
        """
        if not msg.content:
            return False
        task_id = msg.metadata.get("subagent_task_id") if isinstance(msg.metadata, dict) else None
        if task_id and any(
            m.get("injected_event") == "subagent_result" and m.get("subagent_task_id") == task_id
            for m in session.messages
        ):
            return False
        session.add_message(
            "assistant",
            msg.content,
            sender_id=msg.sender_id,
            injected_event="subagent_result",
            subagent_task_id=task_id,
        )
        return True

    def _set_runtime_checkpoint(self, session: Session, payload: dict[str, Any]) -> None:
        """Persist the latest in-flight turn state into session metadata."""
        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)

    @staticmethod
    def _checkpoint_message_key(message: dict[str, Any]) -> tuple[Any, ...]:
        return (
            message.get("role"),
            message.get("content"),
            message.get("tool_call_id"),
            message.get("name"),
            message.get("tool_calls"),
            message.get("reasoning_content"),
            message.get("thinking_blocks"),
        )

    def _restore_runtime_checkpoint(self, session: Session) -> bool:
        """Materialize an unfinished turn into session history before a new request."""
        from datetime import datetime

        checkpoint = session.metadata.get(self._RUNTIME_CHECKPOINT_KEY)
        if not isinstance(checkpoint, dict):
            return False

        assistant_message = checkpoint.get("assistant_message")
        completed_tool_results = checkpoint.get("completed_tool_results") or []
        pending_tool_calls = checkpoint.get("pending_tool_calls") or []

        restored_messages: list[dict[str, Any]] = []
        if isinstance(assistant_message, dict):
            restored = dict(assistant_message)
            restored.setdefault("timestamp", datetime.now().isoformat())
            restored_messages.append(restored)
        for message in completed_tool_results:
            if isinstance(message, dict):
                restored = dict(message)
                restored.setdefault("timestamp", datetime.now().isoformat())
                restored_messages.append(restored)
        for tool_call in pending_tool_calls:
            if not isinstance(tool_call, dict):
                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(),
                }
            )

        overlap = 0
        max_overlap = min(len(session.messages), len(restored_messages))
        for size in range(max_overlap, 0, -1):
            existing = session.messages[-size:]
            restored = restored_messages[:size]
            if all(
                self._checkpoint_message_key(left) == self._checkpoint_message_key(right)
                for left, right in zip(existing, restored)
            ):
                overlap = size
                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",
        sender_id: str = "user",
        media: list[str] | None = None,
        on_progress: Callable[..., Awaitable[None]] | None = None,
        on_stream: Callable[[str], Awaitable[None]] | None = None,
        on_stream_end: Callable[..., Awaitable[None]] | None = None,
        ephemeral: bool = False,
        _run_extra_hooks_for_ephemeral: bool = False,
        hooks: list[AgentHook] | None = None,
        hook_factories: list[AgentTurnHookFactory] | None = None,
        tools: ToolRegistry | None = None,
        persist_user_message: bool = True,
        runtime: LLMRuntime | None = None,
        on_runtime_admitted: Callable[[LLMRuntime], Awaitable[None]] | None = None,
    ) -> OutboundMessage | None:
        """Process an external message directly and return the outbound payload."""
        if channel == "system":
            raise ValueError("channel 'system' is reserved for internal messages")
        await self._connect_mcp()
        metadata: dict[str, Any] = {}
        if not persist_user_message:
            metadata[turn_continuation.SKIP_USER_PERSIST_META] = True
        msg = InboundMessage(
            channel=channel, sender_id=sender_id, chat_id=chat_id,
            content=content, media=media or [], metadata=metadata,
        )
        # Share the dispatch lock so direct calls serialize with bus turns.
        lock = self._session_locks.setdefault(session_key, asyncio.Lock())
        try:
            async with lock:
                kwargs: dict[str, Any] = {
                    "session_key": session_key,
                    "on_progress": on_progress,
                    "on_stream": on_stream,
                    "on_stream_end": on_stream_end,
                    "ephemeral": ephemeral,
                }
                if _run_extra_hooks_for_ephemeral:
                    kwargs["run_extra_hooks_for_ephemeral"] = True
                if hooks is not None:
                    kwargs["hooks"] = hooks
                if hook_factories is not None:
                    kwargs["hook_factories"] = hook_factories
                if tools is not None:
                    kwargs["tools"] = tools
                if runtime is not None:
                    kwargs["runtime"] = runtime
                if on_runtime_admitted is not None:
                    kwargs["on_runtime_admitted"] = on_runtime_admitted
                return await self._process_message(
                    msg,
                    **kwargs,
                )
        finally:
            await self._runtime_events().run_status_changed(msg, session_key, "idle")
            self._runtime_events().clear_turn(session_key)