refactor(orchestrator+bitable): remove Any from type signatures

Eliminate 112 Any usages across orchestrator/ (62) and bitable/ (50) via:
- TYPE_CHECKING Protocol for Redis/LLMGateway/Plan/Dispatcher/StateManager
- object for arbitrary dict/list/value types (Pydantic v2 serializes fine)
- RecalcTask concrete import (replacing Any in recalc_worker.py)
- Coroutine[object, object, object] for async generic
- Remove unused Any imports (F401 cleanup)

Note: Avoided recursive TypeAlias (FieldValue) because Pydantic v2 cannot
build schemas for recursive named aliases (RecursionError).

Tests: 245 passed (bitable 91 + orchestrator 154), 0 regressions
ruff: All checks passed

src/agentkit/bitable/formula/engine.py

@@ -16,7 +16,6 @@ from __future__ import annotations
 
 import ast
 from collections import deque
-from typing import Any
 
 from agentkit.bitable.formula.functions import AGGREGATE_FUNCTIONS, FUNCTION_REGISTRY
 from agentkit.bitable.formula.parser import (
@@ -104,9 +103,9 @@ class FormulaEngine:
     def evaluate(
         self,
         field_id: str,
-        row_values: dict[str, Any],
-        column_values: dict[str, list[Any]] | None = None,
-    ) -> Any:
+        row_values: dict[str, object],
+        column_values: dict[str, list[object]] | None = None,
+    ) -> object:
         """Evaluate a formula field for a specific record.
 
         Args:
@@ -130,7 +129,7 @@ class FormulaEngine:
 
         # Build the field_values dict for the evaluator
         # Aggregate refs get column values (lists), row refs get row values (scalars)
-        eval_values: dict[str, Any] = {}
+        eval_values: dict[str, object] = {}
 
         # Map real field IDs to safe names
         for safe_name, real_id in entry.field_mapping.items():
@@ -143,16 +142,16 @@ class FormulaEngine:
 
     def evaluate_all_for_record(
         self,
-        row_values: dict[str, Any],
-        column_values: dict[str, list[Any]] | None = None,
-    ) -> dict[str, Any]:
+        row_values: dict[str, object],
+        column_values: dict[str, list[object]] | None = None,
+    ) -> dict[str, object]:
         """Evaluate all registered formulas for a record.
 
         Returns a dict of field_id → computed value.
         Formulas are evaluated in topological order so that formula-to-formula
         dependencies are resolved correctly.
         """
-        results: dict[str, Any] = {}
+        results: dict[str, object] = {}
         column_values = column_values or {}
 
         for field_id in self.topological_order():

src/agentkit/bitable/ingestion/api_collector.py

@@ -16,13 +16,11 @@ Usage::
 
 from __future__ import annotations
 
-from typing import Any
-
 
 def transform_records(
-    records: list[dict[str, Any]],
+    records: list[dict[str, object]],
     field_mapping: dict[str, str],
-) -> list[dict[str, Any]]:
+) -> list[dict[str, object]]:
     """Map source record keys to bitable field IDs via field_mapping.
 
     Keys not in ``field_mapping`` are dropped. Values are passed through
@@ -40,9 +38,9 @@ def transform_records(
     if not field_mapping:
         return []
 
-    transformed: list[dict[str, Any]] = []
+    transformed: list[dict[str, object]] = []
     for rec in records:
-        out: dict[str, Any] = {}
+        out: dict[str, object] = {}
         for src_key, field_id in field_mapping.items():
             if src_key in rec:
                 out[field_id] = rec[src_key]

src/agentkit/bitable/ingestion/database.py

@@ -16,7 +16,6 @@ Type mapping (KTD: DB → bitable):
 from __future__ import annotations
 
 import logging
-from typing import Any
 
 from sqlalchemy import (
     BigInteger,
@@ -56,7 +55,7 @@ DB_TYPE_MAP: dict[type, str] = {
 READ_BATCH = 1000
 
 
-def infer_field_type(sqla_type: Any) -> str:
+def infer_field_type(sqla_type: object) -> str:
     """Map a SQLAlchemy column type instance or class to a bitable field type.
 
     Handles both type instances (``Integer()``) and type classes (``Integer``).
@@ -78,7 +77,7 @@ def import_table(
     table_name: str,
     *,
     max_rows: int = 50_000,
-) -> dict[str, Any]:
+) -> dict[str, object]:
     """Reflect a single table from an external DB.
 
     Returns ``{"table_name": str, "fields": [...], "records": [...],
@@ -97,7 +96,7 @@ def import_table(
         engine.dispose()
 
 
-def _reflect_and_read(engine: Engine, table_name: str, max_rows: int) -> dict[str, Any]:
+def _reflect_and_read(engine: Engine, table_name: str, max_rows: int) -> dict[str, object]:
     """Reflect one table and read its rows."""
     insp = inspect(engine)
 
@@ -111,7 +110,7 @@ def _reflect_and_read(engine: Engine, table_name: str, max_rows: int) -> dict[st
     table = Table(table_name, metadata, autoload_with=engine)
 
     # Build field definitions
-    fields: list[dict[str, Any]] = []
+    fields: list[dict[str, object]] = []
     pk_columns = list(table.primary_key.columns)
     pk_name = pk_columns[0].name if pk_columns else None
 
@@ -131,14 +130,14 @@ def _reflect_and_read(engine: Engine, table_name: str, max_rows: int) -> dict[st
         pk_name = "id"
 
     # Read rows
-    records: list[dict[str, Any]] = []
+    records: list[dict[str, object]] = []
     with engine.connect() as conn:
         result = conn.execute(select(table))
         for i, row in enumerate(result):
             if i >= max_rows:
                 logger.warning("Table %r truncated at %d rows during import", table_name, max_rows)
                 break
-            rec: dict[str, Any] = {}
+            rec: dict[str, object] = {}
             for col in table.columns:
                 val = getattr(row, col.name, None)
                 if val is not None:
@@ -155,7 +154,7 @@ def _reflect_and_read(engine: Engine, table_name: str, max_rows: int) -> dict[st
     }
 
 
-def _serialize(val: Any) -> Any:
+def _serialize(val: object) -> object:
     """Serialize a DB value to JSON-safe form."""
     from datetime import date, datetime
     from decimal import Decimal

src/agentkit/bitable/ingestion/excel.py

@@ -18,7 +18,6 @@ import logging
 import socket
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import Any
 from urllib.parse import urlparse
 
 import httpx
@@ -36,7 +35,7 @@ class ParsedSheet:
     name: str
     columns: list[str] = field(default_factory=list)
     field_types: list[str] = field(default_factory=list)  # "text" | "number" | "date"
-    records: list[dict[str, Any]] = field(default_factory=list)
+    records: list[dict[str, object]] = field(default_factory=list)
 
 
 def parse_excel(file_path: str | Path) -> list[ParsedSheet]:
@@ -182,9 +181,9 @@ def _parse_worksheet(ws) -> ParsedSheet | None:
     col_count = len(clean_headers)
     field_types = _infer_column_types(data_rows, col_count)
 
-    records: list[dict[str, Any]] = []
+    records: list[dict[str, object]] = []
     for row in data_rows:
-        rec: dict[str, Any] = {}
+        rec: dict[str, object] = {}
         for i, col_name in enumerate(clean_headers):
             val = row[i] if i < len(row) else None
             if val is not None:
@@ -237,7 +236,7 @@ def _infer_column_types(rows: list[tuple], col_count: int) -> list[str]:
     return types
 
 
-def _coerce_value(val: Any, field_type: str) -> Any:
+def _coerce_value(val: object, field_type: str) -> object:
     """Coerce a cell value to the inferred field type. Truncate long strings."""
     if field_type == "date":
         from datetime import datetime

src/agentkit/bitable/models.py

@@ -9,10 +9,14 @@ from __future__ import annotations
 
 from datetime import datetime, timezone
 from enum import Enum
-from typing import Any
 
 from pydantic import BaseModel, ConfigDict, Field as PydanticField
 
+# ponytail: bitable JSONB columns hold arbitrary JSON. Using `object` instead of
+# a recursive TypeAlias because Pydantic v2 cannot build a schema for recursive
+# named aliases (RecursionError). `object` is the most permissive type and
+# Pydantic v2 serializes dict/list/primitive values fine at runtime.
+
 
 def _utcnow() -> datetime:
     return datetime.now(timezone.utc)
@@ -97,7 +101,7 @@ class Table(BaseModel):
 # ---------------------------------------------------------------------------
 
 # Status select field options — labels and colors match Feishu Bitable defaults.
-_STATUS_OPTIONS: list[dict[str, Any]] = [
+_STATUS_OPTIONS: list[dict[str, object]] = [
     {"label": "未开始", "value": "not_started", "color": "default"},
     {"label": "进行中", "value": "in_progress", "color": "processing"},
     {"label": "已完成", "value": "done", "color": "success"},
@@ -106,7 +110,7 @@ _STATUS_OPTIONS: list[dict[str, Any]] = [
 #: Templates for the 5 default fields created on every new table (R2).
 #: agent-owned fields (创建人/创建时间) are auto-filled by the service layer
 #: on record creation; user-owned fields are user-editable.
-DEFAULT_FIELD_TEMPLATES: list[dict[str, Any]] = [
+DEFAULT_FIELD_TEMPLATES: list[dict[str, object]] = [
     {
         "name": "标题",
         "field_type": FieldType.text,
@@ -155,7 +159,7 @@ class Field(BaseModel):
     table_id: str
     name: str
     field_type: FieldType
-    config: dict[str, Any] = PydanticField(default_factory=dict)
+    config: dict[str, object] = PydanticField(default_factory=dict)
     owner: FieldOwner = FieldOwner.user
     created_at: datetime = PydanticField(default_factory=_utcnow)
 
@@ -167,7 +171,7 @@ class Record(BaseModel):
 
     id: str
     table_id: str
-    values: dict[str, Any] = PydanticField(default_factory=dict)
+    values: dict[str, object] = PydanticField(default_factory=dict)
     created_at: datetime = PydanticField(default_factory=_utcnow)
     updated_at: datetime = PydanticField(default_factory=_utcnow)
 
@@ -181,7 +185,7 @@ class View(BaseModel):
     table_id: str
     name: str
     view_type: ViewType = ViewType.grid
-    config: dict[str, Any] = PydanticField(default_factory=dict)
+    config: dict[str, object] = PydanticField(default_factory=dict)
     created_at: datetime = PydanticField(default_factory=_utcnow)
 
 

src/agentkit/bitable/recalc_worker.py

@@ -18,11 +18,10 @@ from __future__ import annotations
 
 import asyncio
 import logging
-from typing import Any
 
 from agentkit.bitable.db import BitableDB
 from agentkit.bitable.formula.engine import FormulaEngine
-from agentkit.bitable.models import FieldType, RecalcStatus
+from agentkit.bitable.models import FieldType, RecalcStatus, RecalcTask
 from agentkit.bitable.repository import BitableRepository
 from agentkit.bitable.service import BitableService
 
@@ -124,7 +123,7 @@ class RecalcWorker:
                 logger.exception("RecalcWorker error in main loop")
                 await asyncio.sleep(self._poll_interval)
 
-    async def _sort_by_topological_order(self, tasks: list[Any]) -> list[Any]:
+    async def _sort_by_topological_order(self, tasks: list[RecalcTask]) -> list[RecalcTask]:
         """Sort claimed tasks so dependencies are processed first (P1 #7).
 
         Groups tasks by table_id, builds (or reuses) the engine to get the
@@ -146,7 +145,7 @@ class RecalcWorker:
             order = engine.topological_order()
             topo_index[tid] = {fid: i for i, fid in enumerate(order)}
 
-        def _key(t: Any) -> tuple[str, str, int]:
+        def _key(t: RecalcTask) -> tuple[str, str, int]:
             idx = topo_index.get(t.table_id, {}).get(t.field_id, 1 << 30)
             return (t.table_id, t.record_id, idx)
 
@@ -175,7 +174,7 @@ class RecalcWorker:
             except Exception:
                 logger.exception("RecalcWorker reaper error")
 
-    async def process_task(self, task: Any) -> None:
+    async def process_task(self, task: RecalcTask) -> None:
         """Process a single recalc task: evaluate formula → write result.
 
         The task is expected to already be in ``calculating`` status when
@@ -216,7 +215,7 @@ class RecalcWorker:
                 return
 
             deps = engine.get_dependencies(task.field_id)
-            column_values: dict[str, list[Any]] = {}
+            column_values: dict[str, list[object]] = {}
             for dep_field_id in deps:
                 column_values[dep_field_id] = await self._repo.get_column_values(
                     task.table_id, dep_field_id

src/agentkit/bitable/repository.py

@@ -10,7 +10,6 @@ from __future__ import annotations
 import logging
 import re
 from datetime import datetime, timedelta, timezone
-from typing import Any
 
 from sqlalchemy import delete, func, insert, select, text, update
 from sqlalchemy.dialects.postgresql import insert as pg_insert
@@ -102,7 +101,7 @@ class BitableRepository:
             result = await session.execute(stmt)
             return [BitableFile.model_validate(e) for e in result.scalars().all()]
 
-    async def update_file(self, file_id: str, **kwargs: Any) -> BitableFile | None:
+    async def update_file(self, file_id: str, **kwargs: object) -> BitableFile | None:
         """Update a file's attributes."""
         async with self._session_factory() as session:
             stmt = (
@@ -181,7 +180,7 @@ class BitableRepository:
             result = await session.execute(stmt)
             return [Table.model_validate(e) for e in result.scalars().all()]
 
-    async def update_table(self, table_id: str, **kwargs: Any) -> Table | None:
+    async def update_table(self, table_id: str, **kwargs: object) -> Table | None:
         """Update a table's attributes."""
         async with self._session_factory() as session:
             stmt = (
@@ -236,7 +235,7 @@ class BitableRepository:
         table_id: str,
         name: str,
         field_type: FieldType,
-        config: dict[str, Any] | None = None,
+        config: dict[str, object] | None = None,
         owner: FieldOwner = FieldOwner.user,
     ) -> Field:
         """Create a new field in a table."""
@@ -277,7 +276,7 @@ class BitableRepository:
             result = await session.execute(stmt)
             return [Field.model_validate(e) for e in result.scalars().all()]
 
-    async def update_field(self, field_id: str, **kwargs: Any) -> Field | None:
+    async def update_field(self, field_id: str, **kwargs: object) -> Field | None:
         """Update a field's attributes."""
         async with self._session_factory() as session:
             stmt = (
@@ -300,7 +299,7 @@ class BitableRepository:
 
     # ── Records ─────────────────────────────────────────────
 
-    async def create_record(self, table_id: str, values: dict[str, Any] | None = None) -> Record:
+    async def create_record(self, table_id: str, values: dict[str, object] | None = None) -> Record:
         """Create a new record."""
         async with self._session_factory() as session:
             stmt = (
@@ -318,7 +317,7 @@ class BitableRepository:
             return Record.model_validate(entity)
 
     async def create_records_batch(
-        self, table_id: str, records_values: list[dict[str, Any]]
+        self, table_id: str, records_values: list[dict[str, object]]
     ) -> list[Record]:
         """Batch-insert multiple records (P2 #19: eliminates per-record INSERT).
 
@@ -376,7 +375,7 @@ class BitableRepository:
 
             return [Record.model_validate(e) for e in entities], next_cursor
 
-    async def update_record_values(self, record_id: str, values: dict[str, Any]) -> Record | None:
+    async def update_record_values(self, record_id: str, values: dict[str, object]) -> Record | None:
         """Update a record's values (full replace)."""
         async with self._session_factory() as session:
             stmt = (
@@ -413,7 +412,7 @@ class BitableRepository:
         table_id: str,
         name: str,
         view_type: ViewType = ViewType.grid,
-        config: dict[str, Any] | None = None,
+        config: dict[str, object] | None = None,
     ) -> View:
         """Create a new view."""
         async with self._session_factory() as session:
@@ -451,7 +450,7 @@ class BitableRepository:
             result = await session.execute(stmt)
             return [View.model_validate(e) for e in result.scalars().all()]
 
-    async def update_view(self, view_id: str, **kwargs: Any) -> View | None:
+    async def update_view(self, view_id: str, **kwargs: object) -> View | None:
         """Update a view's attributes."""
         async with self._session_factory() as session:
             stmt = (
@@ -543,7 +542,7 @@ class BitableRepository:
     ) -> None:
         """Update a recalc task's status."""
         async with self._session_factory() as session:
-            kwargs: dict[str, Any] = {"status": status.value}
+            kwargs: dict[str, object] = {"status": status.value}
             if error_message is not None:
                 kwargs["error_message"] = error_message
             if status in (RecalcStatus.done, RecalcStatus.error):
@@ -630,7 +629,7 @@ class BitableRepository:
             return result_map
 
     async def upsert_record_agent_fields(
-        self, record_id: str, agent_field_values: dict[str, Any]
+        self, record_id: str, agent_field_values: dict[str, object]
     ) -> None:
         """Update agent-owned fields using jsonb_set (KTD8).
 
@@ -646,7 +645,7 @@ class BitableRepository:
             # Use CAST(:param AS jsonb) instead of :param::jsonb — asyncpg dialect
             # misparses the `::` as part of the param name.
             inner = "values"
-            params: dict[str, Any] = {"record_id": record_id}
+            params: dict[str, object] = {"record_id": record_id}
             for i, (field_id, value) in enumerate(agent_field_values.items()):
                 param_key = f"v{i}"
                 inner = f"jsonb_set({inner}, '{{{field_id}}}', CAST(:{param_key} AS jsonb), true)"
@@ -660,8 +659,8 @@ class BitableRepository:
     async def list_records_filtered(
         self,
         table_id: str,
-        filters: list[dict[str, Any]] | None = None,
-        sorts: list[dict[str, Any]] | None = None,
+        filters: list[dict[str, object]] | None = None,
+        sorts: list[dict[str, object]] | None = None,
         cursor: str | None = None,
         limit: int = 50,
     ) -> tuple[list[Record], str | None]:
@@ -682,7 +681,7 @@ class BitableRepository:
             # Build raw SQL with JSONB filter/sort translation.
             # ponytail: field_ids in filters/sorts are system UUIDs (validated by service layer).
             where_clauses = ["table_id = :table_id"]
-            params: dict[str, Any] = {"table_id": table_id}
+            params: dict[str, object] = {"table_id": table_id}
 
             if filters:
                 for i, f in enumerate(filters):
@@ -783,7 +782,7 @@ class BitableRepository:
                 last_mapping = last_row._mapping
                 # Build composite cursor from sort values + id.
                 # Sort values are extracted as text to match `values->>'fid'` expressions.
-                sv: list[Any] = []
+                sv: list[object] = []
                 last_values = last_mapping.get("values")
                 if isinstance(last_values, str):
                     # asyncpg may return JSONB as str in raw text() queries.
@@ -852,7 +851,7 @@ class BitableRepository:
 
     # ── Recalc support (U3) ────────────────────────────────
 
-    async def get_column_values(self, table_id: str, field_id: str) -> list[Any]:
+    async def get_column_values(self, table_id: str, field_id: str) -> list[object]:
         """Get all values for a field across all records in a table (for aggregates).
 
         Returns a list of values (preserving order by record id). Missing values
@@ -866,7 +865,7 @@ class BitableRepository:
             result = await session.execute(sql, {"field_id": field_id, "table_id": table_id})
             return [row[0] for row in result.fetchall()]
 
-    async def set_formula_value(self, record_id: str, field_id: str, value: Any) -> None:
+    async def set_formula_value(self, record_id: str, field_id: str, value: object) -> None:
         """Set a single formula field value in a record's JSONB (jsonb_set)."""
         import json
 

src/agentkit/client/sync.py

@@ -34,12 +34,19 @@ import os
 import sqlite3
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Any, Callable
+from typing import Callable, TypeAlias
 
 import httpx
 
 logger = logging.getLogger(__name__)
 
+# 缓存的配置项 — 技能/工作流配置为 JSON 反序列化后的字典，值为标量或嵌套结构。
+# 服务器返回的 skills/workflows 列表元素是 dict（model_dump/to_dict 输出），
+# 其中可能包含 list/dict 等容器，因此使用 object 作为值类型。
+SkillConfigDict: TypeAlias = dict[str, object]
+WorkflowConfigDict: TypeAlias = dict[str, object]
+SyncedConfigPayload: TypeAlias = dict[str, object]
+
 
 # ── Defaults ──────────────────────────────────────────────────────────
 
@@ -100,8 +107,8 @@ class ConfigSync:
 
         # In-memory cache (mirrors the SQLite cache for fast access)
         self._version: str | None = None
-        self._skills: list[dict[str, Any]] = []
-        self._workflows: list[dict[str, Any]] = []
+        self._skills: list[SkillConfigDict] = []
+        self._workflows: list[WorkflowConfigDict] = []
         self._last_synced_at: str | None = None
 
     # ── Lifecycle ─────────────────────────────────────────────────
@@ -232,15 +239,15 @@ class ConfigSync:
         """Return the current cached config version hash."""
         return self._version
 
-    def get_skills(self) -> list[dict[str, Any]]:
+    def get_skills(self) -> list[SkillConfigDict]:
         """Return the cached skill configs."""
         return list(self._skills)
 
-    def get_workflows(self) -> list[dict[str, Any]]:
+    def get_workflows(self) -> list[WorkflowConfigDict]:
         """Return the cached workflow configs."""
         return list(self._workflows)
 
-    def get_all(self) -> dict[str, Any]:
+    def get_all(self) -> SyncedConfigPayload:
         """Return all cached configs as a single dict."""
         return {
             "version": self._version,
@@ -249,14 +256,14 @@ class ConfigSync:
             "synced_at": self._last_synced_at,
         }
 
-    def get_skill(self, name: str) -> dict[str, Any] | None:
+    def get_skill(self, name: str) -> SkillConfigDict | None:
         """Return a single skill config by name, or ``None``."""
         for skill in self._skills:
             if skill.get("name") == name:
                 return skill
         return None
 
-    def get_workflow(self, workflow_id: str) -> dict[str, Any] | None:
+    def get_workflow(self, workflow_id: str) -> WorkflowConfigDict | None:
         """Return a single workflow config by ID, or ``None``."""
         for wf in self._workflows:
             if wf.get("workflow_id") == workflow_id:
@@ -281,7 +288,7 @@ class ConfigSync:
             conn.executescript(_CACHE_SCHEMA)
             conn.commit()
 
-    def _save_to_cache(self, data: dict[str, Any]) -> None:
+    def _save_to_cache(self, data: SyncedConfigPayload) -> None:
         """Save the synced configs to the local SQLite cache."""
         now = datetime.now(timezone.utc).isoformat()
         with sqlite3.connect(str(self.cache_db_path)) as conn:

src/agentkit/llm/cache.py

@@ -14,7 +14,7 @@ import logging
 import time
 from collections import OrderedDict
 from dataclasses import dataclass, field
-from typing import TYPE_CHECKING, Any, Protocol, runtime_checkable
+from typing import TYPE_CHECKING, Protocol, runtime_checkable
 
 from agentkit.llm.protocol import LLMResponse, TokenUsage, ToolCall
 from agentkit.utils.vector_math import compute_cosine_similarity
@@ -25,6 +25,52 @@ if TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
+# ---------------------------------------------------------------------------
+# TYPE_CHECKING Protocols — 避免 Any，描述运行时 lazy import 的第三方对象
+# ---------------------------------------------------------------------------
+
+
+if TYPE_CHECKING:
+
+    class _RedisLike(Protocol):
+        """Redis 客户端最小契约（仅覆盖本模块用到的方法）。"""
+
+        async def get(self, key: str) -> bytes | str | None: ...
+
+        async def mget(self, keys: list[str]) -> list[bytes | str | None]: ...
+
+        async def set(self, key: str, value: bytes | str, ex: int | None = None) -> None: ...
+
+        async def smembers(self, key: str) -> set[bytes | str]: ...
+
+        async def sadd(self, name: str, *values: str) -> int: ...
+
+        async def srem(self, name: str, *values: str) -> int: ...
+
+        async def scard(self, name: str) -> int: ...
+
+        async def delete(self, *names: str) -> int: ...
+
+        def pipeline(self) -> "_RedisPipelineLike": ...
+
+    class _RedisPipelineLike(Protocol):
+        """Redis pipeline 最小契约。"""
+
+        def get(self, key: str) -> "_RedisPipelineLike": ...
+
+        def set(
+            self, key: str, value: bytes | str, ex: int | None = None
+        ) -> "_RedisPipelineLike": ...
+
+        def delete(self, *names: str) -> "_RedisPipelineLike": ...
+
+        def sadd(self, name: str, *values: str) -> "_RedisPipelineLike": ...
+
+        def srem(self, name: str, *values: str) -> "_RedisPipelineLike": ...
+
+        async def execute(self) -> list[object]: ...
+
+
 # ---------------------------------------------------------------------------
 # Data Classes
 # ---------------------------------------------------------------------------
@@ -328,7 +374,7 @@ class RedisLLMCache:
         self._semantic_ttl = semantic_ttl
         self._similarity_threshold = similarity_threshold
         self._max_entries_to_scan = max_entries_to_scan
-        self._redis: Any = None
+        self._redis: _RedisLike | None = None
         self._fallback: InMemoryLLMCache | None = fallback  # For auto-degradation
         self._degraded = False  # True if Redis is unreachable
 
@@ -691,7 +737,7 @@ class LitellmCacheManager:
 
     def __init__(self, config: LitellmCacheConfig):
         self._config = config
-        self._cache_instance: Any = None  # litellm.caching.Cache 实例
+        self._cache_instance: object | None = None  # litellm.caching.Cache 实例
         self._hits = 0
         self._misses = 0
 
@@ -709,7 +755,7 @@ class LitellmCacheManager:
         litellm.cache = None
         self._cache_instance = None
 
-    def _create_cache_instance(self) -> Any:
+    def _create_cache_instance(self) -> object:
         """根据 backend 配置创建 LiteLLM Cache 实例。
 
         auto 模式按优先级尝试：RedisSemanticCache → RedisCache → InMemoryCache。

src/agentkit/llm/cache_key.py

@@ -2,14 +2,13 @@
 
 import hashlib
 import json
-from typing import Any
 
 
 def generate_cache_key(
     model: str,
     messages: list[dict[str, str]],
     temperature: float,
-    tools: list[dict[str, Any]] | None = None,
+    tools: list[dict[str, object]] | None = None,
     tool_choice: str = "auto",
     max_tokens: int = 2000,
     user_id: str | None = None,

src/agentkit/llm/config.py

@@ -3,12 +3,12 @@
 import json
 import logging
 from dataclasses import dataclass, field
-from typing import Any, TYPE_CHECKING
+from typing import TYPE_CHECKING
 
 from agentkit.llm.retry import CircuitBreakerConfig, RetryConfig
 
 if TYPE_CHECKING:
-    from agentkit.channels.secrets import SecretsStore
+    from agentkit.channels.secrets import SecretEntry, SecretsStore
 
 logger = logging.getLogger(__name__)
 
@@ -56,7 +56,7 @@ class ProviderConfig:
 
     api_key: str
     base_url: str
-    models: dict[str, dict[str, Any]] = field(default_factory=dict)
+    models: dict[str, dict[str, object]] = field(default_factory=dict)
     type: str = "openai"  # "openai" | "anthropic" | "gemini"
     max_tokens: int = 4096  # Anthropic: default max_tokens
     timeout: float = 120.0  # Anthropic: request timeout
@@ -168,18 +168,18 @@ class ProviderConfig:
         return f"llm:provider:{self.type}:api_key"
 
     @staticmethod
-    def _encode_secret_entry(entry: Any, key: str) -> str:
+    def _encode_secret_entry(entry: object, key: str) -> str:
         """把 SecretEntry 编码为 JSON 字符串（含 key 字段）。"""
         # entry 是 SecretEntry pydantic 模型，有 model_dump()
         if hasattr(entry, "model_dump"):
-            data = entry.model_dump()
+            data = entry.model_dump()  # type: ignore[attr-defined]
         else:
-            data = dict(entry)
+            data = dict(entry)  # type: ignore[call-overload]
         data["key"] = key
         return json.dumps(data)
 
     @staticmethod
-    def _decode_secret_entry(encoded: str) -> Any:
+    def _decode_secret_entry(encoded: str) -> "SecretEntry":
         """从 JSON 字符串解码 SecretEntry。返回带 .key 属性的对象。"""
         from agentkit.channels.secrets import SecretEntry
 

src/agentkit/llm/gateway.py

@@ -5,7 +5,7 @@ import logging
 import time
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Any
+from typing import TYPE_CHECKING, Protocol
 
 from agentkit.core.exceptions import LLMProviderError, ModelNotFoundError
 from agentkit.llm.config import LLMConfig
@@ -14,9 +14,40 @@ from agentkit.llm.providers.tracker import UsageSummary, UsageTracker
 from agentkit.telemetry.tracing import get_tracer, _OTEL_AVAILABLE
 from agentkit.telemetry.metrics import llm_token_histogram
 
+if TYPE_CHECKING:
+    from agentkit.llm.cache import LitellmCacheManager
+
 logger = logging.getLogger(__name__)
 
 
+# ---------------------------------------------------------------------------
+# TYPE_CHECKING Protocols — 避免 Any，描述运行时 lazy import 的对象
+# ---------------------------------------------------------------------------
+
+
+if TYPE_CHECKING:
+
+    class _QuotaServiceLike(Protocol):
+        """Quota service 最小契约（仅覆盖 gateway._enforce_quota 用到的方法）。"""
+
+        async def is_model_allowed(
+            self, db: Path, department_id: str, model: str
+        ) -> tuple[bool, str]: ...
+
+        async def check_quota(
+            self,
+            db: Path,
+            department_id: str,
+            quota_type: str,
+            period: str,
+            current: float,
+        ) -> tuple[bool, str]: ...
+
+        async def get_quota(
+            self, db: Path, department_id: str, quota_type: str, period: str
+        ) -> dict[str, object] | None: ...
+
+
 class QuotaExceededError(Exception):
     """Raised when a department's LLM quota is exceeded.
 
@@ -29,8 +60,8 @@ class QuotaExceededError(Exception):
         department_id: str,
         quota_type: str,
         period: str,
-        limit: Any,
-        current: Any,
+        limit: object,
+        current: object,
     ) -> None:
         self.department_id = department_id
         self.quota_type = quota_type
@@ -46,13 +77,13 @@ class QuotaExceededError(Exception):
 class LLMGateway:
     """LLM 网关 - Provider 注册、模型别名解析、Fallback、Usage 追踪、Cache"""
 
-    def __init__(self, config: LLMConfig | None = None, usage_store: Any = None):
+    def __init__(self, config: LLMConfig | None = None, usage_store: object | None = None):
         self._providers: dict[str, LLMProvider] = {}
         self._usage_tracker = UsageTracker(store=usage_store) if usage_store else UsageTracker()
         self._config = config or LLMConfig()
 
         # Cache (U17 — LiteLLM 缓存管理器，opt-in，默认禁用)
-        self._cache_manager: Any = None  # LitellmCacheManager | None
+        self._cache_manager: "LitellmCacheManager | None" = None
         if self._config.cache and self._config.cache.enabled:
             from agentkit.llm.cache import LitellmCacheConfig, LitellmCacheManager
 
@@ -601,7 +632,7 @@ class LLMGateway:
 
     async def _check_quota_value(
         self,
-        quota_service: Any,
+        quota_service: _QuotaServiceLike,
         db: Path,
         dept_id: str,
         period: str,

src/agentkit/llm/migration.py

@@ -27,12 +27,11 @@ from __future__ import annotations
 
 import logging
 from pathlib import Path
-from typing import Any
 
 logger = logging.getLogger(__name__)
 
 
-def migrate_api_keys_to_secrets(config_path: Path | str) -> dict[str, dict[str, Any]]:
+def migrate_api_keys_to_secrets(config_path: Path | str) -> dict[str, dict[str, object]]:
     """把 agentkit.yaml 中的 plaintext API Key 迁移到 SecretsStore。
 
     流程：
@@ -63,8 +62,8 @@ def migrate_api_keys_to_secrets(config_path: Path | str) -> dict[str, dict[str,
 
     store = SecretsStore()  # master key 从 env 加载
 
-    async def _run() -> dict[str, dict[str, Any]]:
-        report: dict[str, dict[str, Any]] = {}
+    async def _run() -> dict[str, dict[str, object]]:
+        report: dict[str, dict[str, object]] = {}
         for name, pconf in llm_config.providers.items():
             if pconf.api_key_source == "secrets_store" and not pconf.api_key:
                 report[name] = {"status": "skipped", "source": pconf.api_key_source}
@@ -93,9 +92,9 @@ def migrate_api_keys_to_secrets(config_path: Path | str) -> dict[str, dict[str,
     report = asyncio.run(_run())
 
     # 写回 YAML：更新 llm.providers 段，保留其它段
-    providers_out: dict[str, dict[str, Any]] = {}
+    providers_out: dict[str, dict[str, object]] = {}
     for name, pconf in llm_config.providers.items():
-        entry: dict[str, Any] = {
+        entry: dict[str, object] = {
             "type": pconf.type,
             "base_url": pconf.base_url,
             "models": pconf.models,

src/agentkit/llm/protocol.py

@@ -2,7 +2,6 @@
 
 from abc import ABC, abstractmethod
 from dataclasses import dataclass, field
-from typing import Any
 
 
 @dataclass
@@ -23,7 +22,7 @@ class ToolCall:
 
     id: str
     name: str
-    arguments: dict[str, Any]
+    arguments: dict[str, object]
 
 
 @dataclass
@@ -32,7 +31,7 @@ class LLMRequest:
 
     messages: list[dict[str, str]]
     model: str
-    tools: list[dict[str, Any]] | None = None
+    tools: list[dict[str, object]] | None = None
     tool_choice: str = "auto"
     temperature: float = 0.7
     max_tokens: int = 2000
@@ -42,13 +41,13 @@ class LLMRequest:
         self,
         messages: list[dict[str, str]],
         model: str,
-        tools: list[dict[str, Any]] | None = None,
+        tools: list[dict[str, object]] | None = None,
         tool_choice: str = "auto",
         temperature: float = 0.7,
         max_tokens: int = 2000,
         timeout: float | None = None,
-        cache: dict[str, Any] | None = None,
-        **kwargs: Any,
+        cache: dict[str, object] | None = None,
+        **kwargs: object,
     ):
         self.messages = messages
         self.model = model
@@ -59,7 +58,7 @@ class LLMRequest:
         self.timeout = timeout
         self._extra = kwargs
         # U17 — LiteLLM cache 参数（cache_key 或 no-cache），透传到 litellm.acompletion
-        self._cache: dict[str, Any] | None = cache
+        self._cache: dict[str, object] | None = cache
 
 
 @dataclass

src/agentkit/llm/providers/anthropic.py

@@ -3,7 +3,6 @@
 import json
 import logging
 import time
-from typing import Any
 
 import httpx
 
@@ -99,7 +98,9 @@ class AnthropicProvider(LLMProvider):
             "content-type": "application/json",
         }
 
-    def _convert_messages(self, messages: list[dict[str, str]]) -> tuple[str | list[dict[str, Any]] | None, list[dict[str, Any]]]:
+    def _convert_messages(
+        self, messages: list[dict[str, str]]
+    ) -> tuple[str | list[dict[str, object]] | None, list[dict[str, object]]]:
         """将 OpenAI 风格消息转换为 Anthropic 格式
 
         Returns:
@@ -110,8 +111,8 @@ class AnthropicProvider(LLMProvider):
           - list[dict]: Anthropic content blocks(支持 cache_control,U2/G2)
           - None: 无 system 消息
         """
-        system_prompt: str | list[dict[str, Any]] | None = None
-        anthropic_messages: list[dict[str, Any]] = []
+        system_prompt: str | list[dict[str, object]] | None = None
+        anthropic_messages: list[dict[str, object]] = []
 
         for msg in messages:
             role = msg.get("role", "")
@@ -127,7 +128,7 @@ class AnthropicProvider(LLMProvider):
                 # 检查是否有 tool_calls (OpenAI 格式)
                 tool_calls = msg.get("tool_calls")
                 if tool_calls:
-                    blocks: list[dict[str, Any]] = []
+                    blocks: list[dict[str, object]] = []
                     # 如果有文本内容，先添加文本块
                     if content:
                         blocks.append({"type": "text", "text": content})
@@ -139,25 +140,29 @@ class AnthropicProvider(LLMProvider):
                                 arguments = json.loads(arguments)
                             except json.JSONDecodeError:
                                 arguments = {"raw": arguments}
-                        blocks.append({
+                        blocks.append(
+                            {
                                 "type": "tool_use",
                                 "id": tc.get("id", ""),
                                 "name": func.get("name", ""),
                                 "input": arguments,
-                        })
+                            }
+                        )
                     anthropic_messages.append({"role": "assistant", "content": blocks})
                 else:
-                    anthropic_messages.append({
+                    anthropic_messages.append(
+                        {
                             "role": "assistant",
                             "content": [{"type": "text", "text": content}],
-                    })
+                        }
+                    )
                 continue
 
             if role == "user":
                 # 检查是否是 tool_result 消息 (OpenAI 格式中 tool 角色的结果)
                 # OpenAI 格式: {"role": "tool", "tool_call_id": "...", "content": "..."}
                 if msg.get("tool_call_id"):
-                    tool_result_blocks: list[dict[str, Any]] = []
+                    tool_result_blocks: list[dict[str, object]] = []
                     tool_content = msg.get("content", "")
                     # tool_result 的 content 可以是字符串或内容块列表
                     if isinstance(tool_content, str):
@@ -167,56 +172,72 @@ class AnthropicProvider(LLMProvider):
                     else:
                         tool_result_blocks.append({"type": "text", "text": str(tool_content)})
 
-                    anthropic_messages.append({
+                    anthropic_messages.append(
+                        {
                             "role": "user",
-                        "content": [{
+                            "content": [
+                                {
                                     "type": "tool_result",
                                     "tool_use_id": msg.get("tool_call_id", ""),
                                     "content": tool_result_blocks,
-                        }],
-                    })
+                                }
+                            ],
+                        }
+                    )
                 else:
-                    anthropic_messages.append({
+                    anthropic_messages.append(
+                        {
                             "role": "user",
                             "content": [{"type": "text", "text": content}],
-                    })
+                        }
+                    )
                 continue
 
             if role == "tool":
                 # OpenAI 格式中独立的 tool 消息
                 tool_content = msg.get("content", "")
                 if isinstance(tool_content, str):
-                    result_content: list[dict[str, Any]] | str = [{"type": "text", "text": tool_content}]
+                    result_content: list[dict[str, object]] | str = [
+                        {"type": "text", "text": tool_content}
+                    ]
                 elif isinstance(tool_content, list):
                     result_content = tool_content
                 else:
                     result_content = [{"type": "text", "text": str(tool_content)}]
 
-                anthropic_messages.append({
+                anthropic_messages.append(
+                    {
                         "role": "user",
-                    "content": [{
+                        "content": [
+                            {
                                 "type": "tool_result",
                                 "tool_use_id": msg.get("tool_call_id", ""),
                                 "content": result_content,
-                    }],
-                })
+                            }
+                        ],
+                    }
+                )
 
         return system_prompt, anthropic_messages
 
-    def _convert_tools(self, tools: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    def _convert_tools(self, tools: list[dict[str, object]]) -> list[dict[str, object]]:
         """将 OpenAI function 格式转换为 Anthropic tool 格式"""
         anthropic_tools = []
         for tool in tools:
             if tool.get("type") == "function":
                 func = tool.get("function", {})
-                anthropic_tools.append({
+                anthropic_tools.append(
+                    {
                         "name": func.get("name", ""),
                         "description": func.get("description", ""),
-                    "input_schema": func.get("parameters", {"type": "object", "properties": {}}),
-                })
+                        "input_schema": func.get(
+                            "parameters", {"type": "object", "properties": {}}
+                        ),
+                    }
+                )
         return anthropic_tools
 
-    def _convert_tool_choice(self, tool_choice: str) -> dict[str, Any] | None:
+    def _convert_tool_choice(self, tool_choice: str) -> dict[str, object] | None:
         """将 OpenAI tool_choice 格式转换为 Anthropic 格式"""
         if tool_choice == "auto":
             return {"type": "auto"}
@@ -227,7 +248,7 @@ class AnthropicProvider(LLMProvider):
             return {"type": "tool", "name": tool_choice}
         return None
 
-    def _parse_response(self, data: dict[str, Any], model: str) -> LLMResponse:
+    def _parse_response(self, data: dict[str, object], model: str) -> LLMResponse:
         """将 Anthropic 响应转换为 LLMResponse"""
         content_blocks = data.get("content", [])
         text_parts: list[str] = []
@@ -238,11 +259,13 @@ class AnthropicProvider(LLMProvider):
             if block_type == "text":
                 text_parts.append(block.get("text", ""))
             elif block_type == "tool_use":
-                tool_calls.append(ToolCall(
+                tool_calls.append(
+                    ToolCall(
                         id=block.get("id", ""),
                         name=block.get("name", ""),
                         arguments=block.get("input", {}),
-                ))
+                    )
+                )
 
         usage_data = data.get("usage", {})
         usage = TokenUsage(
@@ -287,7 +310,7 @@ class AnthropicProvider(LLMProvider):
 
         system_prompt, anthropic_messages = self._convert_messages(request.messages)
 
-        payload: dict[str, Any] = {
+        payload: dict[str, object] = {
             "model": request.model,
             "max_tokens": request.max_tokens or self._max_tokens,
             "messages": anthropic_messages,
@@ -346,7 +369,7 @@ class AnthropicProvider(LLMProvider):
 
         system_prompt, anthropic_messages = self._convert_messages(request.messages)
 
-        payload: dict[str, Any] = {
+        payload: dict[str, object] = {
             "model": request.model,
             "max_tokens": request.max_tokens or self._max_tokens,
             "messages": anthropic_messages,
@@ -375,7 +398,7 @@ class AnthropicProvider(LLMProvider):
     async def _iterate_stream(self, response, request: LLMRequest):
         """Iterate over an already-open SSE stream and yield StreamChunks."""
         # Accumulated tool calls: tool_use_id -> {id, name, input_json_str}
-        accumulated_tool_calls: dict[str, dict[str, Any]] = {}
+        accumulated_tool_calls: dict[str, dict[str, object]] = {}
         current_tool_id: str | None = None
         current_tool_name: str | None = None
         current_tool_input_json: str = ""
@@ -433,7 +456,9 @@ class AnthropicProvider(LLMProvider):
                 # Finalize current tool call if any
                 if current_tool_id is not None:
                     try:
-                        arguments = json.loads(current_tool_input_json) if current_tool_input_json else {}
+                        arguments = (
+                            json.loads(current_tool_input_json) if current_tool_input_json else {}
+                        )
                     except json.JSONDecodeError:
                         arguments = {"raw": current_tool_input_json}
 
@@ -510,7 +535,7 @@ class AnthropicProvider(LLMProvider):
                 error_msg = error_info.get("message", "Stream error")
                 raise LLMProviderError("anthropic", error_msg)
 
-    def get_model_info(self) -> dict[str, Any]:
+    def get_model_info(self) -> dict[str, object]:
         """返回 Provider 和模型信息"""
         return {
             "provider": "anthropic",

src/agentkit/llm/providers/doubao.py

@@ -8,7 +8,6 @@ API：火山引擎 OpenAI 兼容接口
 from __future__ import annotations
 
 import logging
-from typing import Any
 
 from agentkit.llm.providers.openai import OpenAICompatibleProvider
 
@@ -48,7 +47,7 @@ class DoubaoProvider(OpenAICompatibleProvider):
         api_key: str,
         base_url: str = DOUBAO_DEFAULT_BASE_URL,
         default_model: str = "doubao-pro-32k",
-        **kwargs: Any,
+        **kwargs: object,
     ):
         super().__init__(
             api_key=api_key,

src/agentkit/llm/providers/gemini.py

@@ -3,7 +3,6 @@
 import json
 import logging
 import time
-from typing import Any
 
 import httpx
 
@@ -90,14 +89,14 @@ class GeminiProvider(LLMProvider):
 
     def _convert_messages(
         self, messages: list[dict[str, str]]
-    ) -> tuple[dict[str, Any] | None, list[dict[str, Any]]]:
+    ) -> tuple[dict[str, object] | None, list[dict[str, object]]]:
         """将 OpenAI 风格消息转换为 Gemini 格式
 
         Returns:
             (system_instruction, contents)
         """
-        system_instruction: dict[str, Any] | None = None
-        contents: list[dict[str, Any]] = []
+        system_instruction: dict[str, object] | None = None
+        contents: list[dict[str, object]] = []
 
         for msg in messages:
             role = msg.get("role", "")
@@ -119,28 +118,34 @@ class GeminiProvider(LLMProvider):
                             tool_name = parsed.get("name", "")
                         except (json.JSONDecodeError, AttributeError):
                             pass
-                    contents.append({
+                    contents.append(
+                        {
                             "role": "user",
-                        "parts": [{
+                            "parts": [
+                                {
                                     "functionResponse": {
                                         "name": tool_name,
                                         "response": {
                                             "content": content,
                                         },
                                     },
-                        }],
-                    })
+                                }
+                            ],
+                        }
+                    )
                 else:
-                    contents.append({
+                    contents.append(
+                        {
                             "role": "user",
                             "parts": [{"text": content}],
-                    })
+                        }
+                    )
                 continue
 
             if role == "assistant":
                 tool_calls = msg.get("tool_calls")
                 if tool_calls:
-                    parts: list[dict[str, Any]] = []
+                    parts: list[dict[str, object]] = []
                     if content:
                         parts.append({"text": content})
                     for tc in tool_calls:
@@ -151,54 +156,64 @@ class GeminiProvider(LLMProvider):
                                 arguments = json.loads(arguments)
                             except json.JSONDecodeError:
                                 arguments = {"raw": arguments}
-                        parts.append({
+                        parts.append(
+                            {
                                 "functionCall": {
                                     "name": func.get("name", ""),
                                     "args": arguments,
                                 },
-                        })
+                            }
+                        )
                     contents.append({"role": "model", "parts": parts})
                 else:
-                    contents.append({
+                    contents.append(
+                        {
                             "role": "model",
                             "parts": [{"text": content}],
-                    })
+                        }
+                    )
                 continue
 
             if role == "tool":
                 # OpenAI format: {"role": "tool", "tool_call_id": "...", "content": "..."}
                 tool_name = msg.get("name", "")
                 tool_content = msg.get("content", "")
-                contents.append({
+                contents.append(
+                    {
                         "role": "user",
-                    "parts": [{
+                        "parts": [
+                            {
                                 "functionResponse": {
                                     "name": tool_name,
                                     "response": {
                                         "content": tool_content,
                                     },
                                 },
-                    }],
-                })
+                            }
+                        ],
+                    }
+                )
 
         return system_instruction, contents
 
-    def _convert_tools(self, tools: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    def _convert_tools(self, tools: list[dict[str, object]]) -> list[dict[str, object]]:
         """将 OpenAI function 格式转换为 Gemini functionDeclarations"""
         declarations = []
         for tool in tools:
             if tool.get("type") == "function":
                 func = tool.get("function", {})
-                declarations.append({
+                declarations.append(
+                    {
                         "name": func.get("name", ""),
                         "description": func.get("description", ""),
                         "parameters": func.get("parameters", {"type": "object", "properties": {}}),
-                })
+                    }
+                )
         if not declarations:
             return []
         return [{"functionDeclarations": declarations}]
 
-    def _convert_tool_choice(self, tool_choice: str) -> dict[str, Any] | None:
+    def _convert_tool_choice(self, tool_choice: str) -> dict[str, object] | None:
         """将 OpenAI tool_choice 格式转换为 Gemini toolConfig"""
         if tool_choice == "auto":
             return {"functionCallingConfig": {"mode": "AUTO"}}
@@ -210,7 +225,7 @@ class GeminiProvider(LLMProvider):
             return {"functionCallingConfig": {"mode": "NONE"}}
         return None
 
-    def _parse_response(self, data: dict[str, Any], model: str) -> LLMResponse:
+    def _parse_response(self, data: dict[str, object], model: str) -> LLMResponse:
         """将 Gemini 响应转换为 LLMResponse"""
         candidates = data.get("candidates", [])
         text_parts: list[str] = []
@@ -225,11 +240,13 @@ class GeminiProvider(LLMProvider):
                     text_parts.append(part["text"])
                 elif "functionCall" in part:
                     fc = part["functionCall"]
-                    tool_calls.append(ToolCall(
+                    tool_calls.append(
+                        ToolCall(
                             id=f"call_{tool_call_index}",
                             name=fc.get("name", ""),
                             arguments=fc.get("args", {}),
-                    ))
+                        )
+                    )
                     tool_call_index += 1
 
         usage_metadata = data.get("usageMetadata", {})
@@ -275,7 +292,7 @@ class GeminiProvider(LLMProvider):
 
         system_instruction, contents = self._convert_messages(request.messages)
 
-        payload: dict[str, Any] = {
+        payload: dict[str, object] = {
             "contents": contents,
             "generationConfig": {
                 "temperature": request.temperature,
@@ -340,7 +357,7 @@ class GeminiProvider(LLMProvider):
 
         system_instruction, contents = self._convert_messages(request.messages)
 
-        payload: dict[str, Any] = {
+        payload: dict[str, object] = {
             "contents": contents,
             "generationConfig": {
                 "temperature": request.temperature,
@@ -374,7 +391,7 @@ class GeminiProvider(LLMProvider):
 
     async def _iterate_stream(self, response, request: LLMRequest):
         """Iterate over an already-open SSE stream and yield StreamChunks."""
-        accumulated_tool_calls: list[dict[str, Any]] = []
+        accumulated_tool_calls: list[dict[str, object]] = []
         model = request.model or self._model
 
         async for line in response.aiter_lines():
@@ -436,11 +453,13 @@ class GeminiProvider(LLMProvider):
                         )
                 elif "functionCall" in part:
                     fc = part["functionCall"]
-                    accumulated_tool_calls.append({
+                    accumulated_tool_calls.append(
+                        {
                             "id": f"call_{len(accumulated_tool_calls)}",
                             "name": fc.get("name", ""),
                             "arguments": fc.get("args", {}),
-                    })
+                        }
+                    )
 
             # Check for finish reason
             finish_reason = candidates[0].get("finishReason", "")
@@ -461,7 +480,7 @@ class GeminiProvider(LLMProvider):
                 )
                 accumulated_tool_calls = []
 
-    def get_model_info(self) -> dict[str, Any]:
+    def get_model_info(self) -> dict[str, object]:
         """返回 Provider 和模型信息"""
         return {
             "provider": "gemini",

src/agentkit/llm/providers/litellm_provider.py

@@ -26,8 +26,7 @@ import inspect
 import json
 import logging
 import time
-from collections.abc import AsyncGenerator
-from typing import Any
+from collections.abc import AsyncGenerator, Iterable
 
 from agentkit.core.exceptions import LLMProviderError
 from agentkit.llm.protocol import (
@@ -81,13 +80,13 @@ class LitellmProvider(LLMProvider):
         api_key: str,
         base_url: str | None = None,
         provider_type: str = "openai",
-        **default_kwargs: Any,
+        **default_kwargs: object,
     ) -> None:
         self._model_prefix = model_prefix
         self._api_key = api_key
         self._base_url = base_url or None  # 空字符串视作未设置
         self._provider_type = provider_type
-        self._default_kwargs: dict[str, Any] = dict(default_kwargs)
+        self._default_kwargs: dict[str, object] = dict(default_kwargs)
 
     async def chat(self, request: LLMRequest) -> LLMResponse:
         """非流式 chat — 调用 ``litellm.acompletion`` 并翻译响应。"""
@@ -116,7 +115,7 @@ class LitellmProvider(LLMProvider):
 
         kwargs = self._build_kwargs(request, stream=True)
 
-        accumulated_tool_calls: dict[int, dict[str, Any]] = {}
+        accumulated_tool_calls: dict[int, dict[str, object]] = {}
         final_usage: TokenUsage | None = None
         final_model: str = request.model
 
@@ -158,9 +157,9 @@ class LitellmProvider(LLMProvider):
     # 内部辅助
     # ------------------------------------------------------------------
 
-    def _build_kwargs(self, request: LLMRequest, *, stream: bool) -> dict[str, Any]:
+    def _build_kwargs(self, request: LLMRequest, *, stream: bool) -> dict[str, object]:
         """从 LLMRequest 构造 litellm.acompletion kwargs。"""
-        kwargs: dict[str, Any] = {
+        kwargs: dict[str, object] = {
             "model": f"{self._model_prefix}{request.model}",
             "messages": request.messages,
             "temperature": request.temperature,
@@ -187,7 +186,7 @@ class LitellmProvider(LLMProvider):
 
     def _parse_response(
         self,
-        response: Any,
+        response: object,
         request_model: str,
         latency_ms: float,
     ) -> LLMResponse:
@@ -229,9 +228,9 @@ class LitellmProvider(LLMProvider):
 
     def _parse_stream_chunk(
         self,
-        chunk: Any,
+        chunk: object,
         request_model: str,
-        accumulated_tool_calls: dict[int, dict[str, Any]],
+        accumulated_tool_calls: dict[int, dict[str, object]],
     ) -> StreamChunk:
         """解析单个流式 chunk（非 final）。累计 tool_calls 到传入字典。"""
         choices = getattr(chunk, "choices", None) or []
@@ -262,7 +261,7 @@ class LitellmProvider(LLMProvider):
 
     def _finalize_tool_calls(
         self,
-        accumulated: dict[int, dict[str, Any]],
+        accumulated: dict[int, dict[str, object]],
     ) -> list[ToolCall]:
         """把累计的流式 tool_calls 字典转成 ToolCall 列表。"""
         tool_calls: list[ToolCall] = []
@@ -288,7 +287,7 @@ class LitellmProvider(LLMProvider):
 # ----------------------------------------------------------------------
 
 
-def _parse_tool_calls(raw_tool_calls: Any) -> list[ToolCall]:
+def _parse_tool_calls(raw_tool_calls: Iterable[object]) -> list[ToolCall]:
     """解析非流式响应的 tool_calls（OpenAI 格式 list[ChoiceMessageToolCall]）。"""
     result: list[ToolCall] = []
     for tc in raw_tool_calls:
@@ -312,7 +311,7 @@ def _parse_tool_calls(raw_tool_calls: Any) -> list[ToolCall]:
     return result
 
 
-def _parse_usage(usage_obj: Any) -> TokenUsage:
+def _parse_usage(usage_obj: object) -> TokenUsage:
     """解析 usage 对象（OpenAI CompletionUsage 或 dict）。"""
     prompt = getattr(usage_obj, "prompt_tokens", None)
     completion = getattr(usage_obj, "completion_tokens", None)
@@ -326,8 +325,8 @@ def _parse_usage(usage_obj: Any) -> TokenUsage:
 
 
 def _accumulate_stream_tool_calls(
-    raw_tool_calls: Any,
-    accumulated: dict[int, dict[str, Any]],
+    raw_tool_calls: Iterable[object],
+    accumulated: dict[int, dict[str, object]],
 ) -> None:
     """累计流式 chunk 里的 tool_calls 片段（OpenAI delta.tool_calls 格式）。
 
@@ -364,7 +363,7 @@ def create_litellm_provider(
     provider_type: str,
     api_key: str,
     base_url: str | None = None,
-    **kwargs: Any,
+    **kwargs: object,
 ) -> LitellmProvider:
     """根据 provider_type 创建 LitellmProvider 实例。
 

src/agentkit/llm/providers/usage_store.py

@@ -18,7 +18,7 @@ import json
 import logging
 from dataclasses import dataclass, field
 from datetime import datetime, timedelta, timezone
-from typing import Any, Protocol, runtime_checkable
+from typing import Protocol, runtime_checkable
 
 from agentkit.llm.protocol import TokenUsage
 
@@ -294,8 +294,8 @@ class RedisUsageStore:
 
     def __init__(self, redis_url: str = "redis://localhost:6379"):
         self._redis_url = redis_url
-        self._redis: Any = None
-        self._sync_redis: Any = None
+        self._redis: object | None = None
+        self._sync_redis: object | None = None
         self._fallback: InMemoryUsageStore | None = None
         self._degraded = False
         self._health_check_task: asyncio.Task[None] | None = None
@@ -687,7 +687,7 @@ class RedisUsageStore:
 
     @staticmethod
     def _read_list(
-        r: Any,
+        r: object,
         list_key: str,
         start: datetime,
         end: datetime,

src/agentkit/llm/providers/wenxin.py

@@ -9,7 +9,6 @@ from __future__ import annotations
 
 import logging
 import time
-from typing import Any
 
 from agentkit.llm.providers.openai import OpenAICompatibleProvider
 from agentkit.llm.protocol import LLMRequest, LLMResponse
@@ -51,7 +50,7 @@ class WenxinProvider(OpenAICompatibleProvider):
         secret_key: str | None = None,
         base_url: str = WENXIN_DEFAULT_BASE_URL,
         default_model: str = "ernie-4.5-turbo-128k",
-        **kwargs: Any,
+        **kwargs: object,
     ):
         # If AK/SK provided, use token-based auth
         self._access_key = access_key

src/agentkit/llm/providers/yuanbao.py

@@ -8,7 +8,6 @@ API：腾讯云 OpenAI 兼容接口
 from __future__ import annotations
 
 import logging
-from typing import Any
 
 from agentkit.llm.providers.openai import OpenAICompatibleProvider
 from agentkit.llm.protocol import LLMRequest, LLMResponse
@@ -48,7 +47,7 @@ class YuanbaoProvider(OpenAICompatibleProvider):
         base_url: str = YUANBAO_DEFAULT_BASE_URL,
         default_model: str = "hunyuan-turbos-latest",
         enable_enhancement: bool = False,
-        **kwargs: Any,
+        **kwargs: object,
     ):
         self._enable_enhancement = enable_enhancement
         super().__init__(

src/agentkit/llm/remote_provider.py

@@ -3,7 +3,6 @@
 import json
 import logging
 from collections.abc import AsyncIterator, Awaitable, Callable
-from typing import Any
 
 import httpx
 
@@ -66,7 +65,7 @@ class RemoteLLMProvider(LLMProvider):
             "Content-Type": "application/json",
         }
 
-    def _build_payload(self, request: LLMRequest) -> dict[str, Any]:
+    def _build_payload(self, request: LLMRequest) -> dict[str, object]:
         """Convert LLMRequest to server API payload."""
         return {
             "messages": request.messages,
@@ -91,7 +90,7 @@ class RemoteLLMProvider(LLMProvider):
                 return str(body["error"])
         return str(body)
 
-    def _parse_response(self, data: dict[str, Any], request: LLMRequest) -> LLMResponse:
+    def _parse_response(self, data: dict[str, object], request: LLMRequest) -> LLMResponse:
         """Parse server response JSON into an LLMResponse."""
         usage_data = data.get("usage") or {}
         usage = TokenUsage(
@@ -115,7 +114,7 @@ class RemoteLLMProvider(LLMProvider):
             latency_ms=data.get("latency_ms", 0.0),
         )
 
-    def _parse_chunk(self, data: dict[str, Any], request: LLMRequest) -> StreamChunk:
+    def _parse_chunk(self, data: dict[str, object], request: LLMRequest) -> StreamChunk:
         """Parse a single SSE data payload into a StreamChunk."""
         usage: TokenUsage | None = None
         usage_data = data.get("usage")
@@ -218,9 +217,7 @@ class RemoteLLMProvider(LLMProvider):
                     if response.status_code == 502:
                         await response.aread()
                         detail = self._extract_error_detail(response)
-                        raise LLMProviderError(
-                            "remote", f"Server LLM gateway error: {detail}"
-                        )
+                        raise LLMProviderError("remote", f"Server LLM gateway error: {detail}")
                     if response.status_code != 200:
                         await response.aread()
                         raise LLMProviderError(

src/agentkit/llm/retry.py

@@ -5,7 +5,7 @@ import logging
 import time
 from dataclasses import dataclass, field
 from enum import Enum
-from typing import Any, Callable
+from typing import Callable
 
 from agentkit.core.exceptions import LLMProviderError
 
@@ -20,9 +20,7 @@ class RetryConfig:
     base_delay: float = 1.0
     max_delay: float = 30.0
     exponential_base: float = 2.0
-    retryable_status_codes: set[int] = field(
-        default_factory=lambda: {429, 500, 502, 503, 529}
-    )
+    retryable_status_codes: set[int] = field(default_factory=lambda: {429, 500, 502, 503, 529})
 
 
 class CircuitState(Enum):
@@ -69,7 +67,7 @@ class RetryPolicy:
     def __init__(self, config: RetryConfig | None = None):
         self._config = config or RetryConfig()
 
-    async def execute(self, fn: Callable, *args: Any, **kwargs: Any) -> Any:
+    async def execute(self, fn: Callable, *args: object, **kwargs: object) -> object:
         """Execute fn with retry on retryable errors."""
         last_error: Exception | None = None
 
@@ -84,7 +82,7 @@ class RetryPolicy:
                     raise
 
                 delay = min(
-                    self._config.base_delay * (self._config.exponential_base ** attempt),
+                    self._config.base_delay * (self._config.exponential_base**attempt),
                     self._config.max_delay,
                 )
                 logger.warning(
@@ -142,7 +140,7 @@ class CircuitBreaker:
                 f"after {self._failure_count} failures"
             )
 
-    async def execute(self, fn: Callable, *args: Any, **kwargs: Any) -> Any:
+    async def execute(self, fn: Callable, *args: object, **kwargs: object) -> object:
         """Execute fn through the circuit breaker."""
         current_state = self.state
 
@@ -158,6 +156,6 @@ class CircuitBreaker:
             result = await fn(*args, **kwargs)
             self._on_success()
             return result
-        except Exception as e:
+        except Exception:
             self._on_failure()
             raise

src/agentkit/memory/adapters/base.py

@@ -10,7 +10,6 @@ from __future__ import annotations
 
 import logging
 from abc import ABC, abstractmethod
-from typing import Any
 
 import httpx
 
@@ -95,8 +94,7 @@ class KBAdapter(ABC):
     async def delete_by_id(self, id: str) -> bool:
         """按文档 ID 删除（子类可覆盖）"""
         logger.warning(
-            f"{self.__class__.__name__} does not support delete_by_id; "
-            f"id '{id}' skipped"
+            f"{self.__class__.__name__} does not support delete_by_id; id '{id}' skipped"
         )
         return False
 
@@ -127,8 +125,7 @@ class KBAdapter(ABC):
     async def get_document(self, doc_id: str) -> Document | None:
         """按 ID 获取单个文档（子类可覆盖）"""
         logger.warning(
-            f"{self.__class__.__name__} does not support get_document; "
-            f"doc_id '{doc_id}' not found"
+            f"{self.__class__.__name__} does not support get_document; doc_id '{doc_id}' not found"
         )
         return None
 
@@ -156,5 +153,5 @@ class KBAdapter(ABC):
     async def __aenter__(self) -> KBAdapter:
         return self
 
-    async def __aexit__(self, *args: Any) -> None:
+    async def __aexit__(self, *args: object) -> None:
         await self.close()

src/agentkit/memory/adapters/confluence.py

@@ -7,7 +7,6 @@
 from __future__ import annotations
 
 import logging
-from typing import Any
 
 import httpx
 
@@ -56,7 +55,9 @@ class ConfluenceAdapter(KBAdapter):
         )
         self._base_url = base_url.rstrip("/")
         if not is_safe_url(self._base_url):
-            raise ValueError(f"Unsafe base_url: {self._base_url}. Private/internal URLs are not allowed.")
+            raise ValueError(
+                f"Unsafe base_url: {self._base_url}. Private/internal URLs are not allowed."
+            )
         self._username = username
         self._api_token = api_token
         self._space_keys = space_keys or []
@@ -65,9 +66,7 @@ class ConfluenceAdapter(KBAdapter):
         """创建 Confluence API HTTP 客户端"""
         import base64
 
-        credentials = base64.b64encode(
-            f"{self._username}:{self._api_token}".encode()
-        ).decode()
+        credentials = base64.b64encode(f"{self._username}:{self._api_token}".encode()).decode()
         return httpx.AsyncClient(
             base_url=self._base_url,
             headers={
@@ -101,7 +100,7 @@ class ConfluenceAdapter(KBAdapter):
                 space_filter = " OR ".join(
                     f'space = "{_escape_cql(key)}"' for key in self._space_keys
                 )
-                cql = f'{cql} AND ({space_filter})'
+                cql = f"{cql} AND ({space_filter})"
 
             resp = await client.get(
                 "/rest/api/content/search",
@@ -115,6 +114,7 @@ class ConfluenceAdapter(KBAdapter):
                 body = page.get("body", {}).get("storage", {}).get("value", "")
                 # Strip HTML tags for plain text content
                 import re
+
                 content = re.sub(r"<[^>]+>", "", body) if body else page.get("title", "")
 
                 results.append(
@@ -136,8 +136,7 @@ class ConfluenceAdapter(KBAdapter):
 
         except httpx.HTTPStatusError as e:
             logger.error(
-                f"Confluence search HTTP error: {e.response.status_code} — "
-                f"{e.response.text[:200]}"
+                f"Confluence search HTTP error: {e.response.status_code} — {e.response.text[:200]}"
             )
             return []
         except Exception as e:
@@ -157,6 +156,7 @@ class ConfluenceAdapter(KBAdapter):
 
             body = page.get("body", {}).get("storage", {}).get("value", "")
             import re
+
             content = re.sub(r"<[^>]+>", "", body) if body else ""
 
             return Document(
@@ -191,13 +191,17 @@ class ConfluenceAdapter(KBAdapter):
                         source_type="confluence",
                     )
                 )
-            return sources if sources else [
+            return (
+                sources
+                if sources
+                else [
                     SourceInfo(
                         source_id=self._source_id,
                         source_name=self._source_name,
                         source_type=self._source_type,
                     )
                 ]
+            )
         except Exception as e:
             logger.error(f"Confluence list_sources error: {e}")
             return [

src/agentkit/memory/adapters/feishu.py

@@ -8,7 +8,7 @@ from __future__ import annotations
 
 import logging
 import time
-from typing import Any
+from typing import TypeAlias
 
 import httpx
 
@@ -18,6 +18,9 @@ from agentkit.utils.security import is_safe_url
 
 logger = logging.getLogger(__name__)
 
+# 飞书搜索请求 payload：search_key/page_size/wiki_space_ids — 值为 str|int|list[str]。
+FeishuSearchPayload: TypeAlias = dict[str, object]
+
 
 class FeishuKBAdapter(KBAdapter):
     """飞书知识库适配器
@@ -54,7 +57,9 @@ class FeishuKBAdapter(KBAdapter):
         self._app_secret = app_secret
         self._base_url = base_url.rstrip("/")
         if not is_safe_url(self._base_url):
-            raise ValueError(f"Unsafe base_url: {self._base_url}. Private/internal URLs are not allowed.")
+            raise ValueError(
+                f"Unsafe base_url: {self._base_url}. Private/internal URLs are not allowed."
+            )
         self._space_ids = space_ids or []
         self._access_token: str | None = None
         self._token_expiry: float = 0.0
@@ -94,10 +99,7 @@ class FeishuKBAdapter(KBAdapter):
                 self._client = None
                 return self._access_token
             else:
-                logger.error(
-                    f"Feishu auth failed: code={data.get('code')}, "
-                    f"msg={data.get('msg')}"
-                )
+                logger.error(f"Feishu auth failed: code={data.get('code')}, msg={data.get('msg')}")
                 return None
         except Exception as e:
             logger.error(f"Feishu auth error: {e}")
@@ -121,7 +123,7 @@ class FeishuKBAdapter(KBAdapter):
 
         client = self._get_client()
         try:
-            payload: dict[str, Any] = {
+            payload: FeishuSearchPayload = {
                 "search_key": query,
                 "page_size": top_k,
             }
@@ -137,8 +139,7 @@ class FeishuKBAdapter(KBAdapter):
 
             if data.get("code") != 0:
                 logger.error(
-                    f"Feishu search failed: code={data.get('code')}, "
-                    f"msg={data.get('msg')}"
+                    f"Feishu search failed: code={data.get('code')}, msg={data.get('msg')}"
                 )
                 return []
 
@@ -162,8 +163,7 @@ class FeishuKBAdapter(KBAdapter):
 
         except httpx.HTTPStatusError as e:
             logger.error(
-                f"Feishu search HTTP error: {e.response.status_code} — "
-                f"{e.response.text[:200]}"
+                f"Feishu search HTTP error: {e.response.status_code} — {e.response.text[:200]}"
             )
             return []
         except Exception as e:
@@ -179,7 +179,7 @@ class FeishuKBAdapter(KBAdapter):
         client = self._get_client()
         try:
             resp = await client.get(
-                f"/wiki/v2/spaces/get_node",
+                "/wiki/v2/spaces/get_node",
                 params={"token": doc_id},
             )
             resp.raise_for_status()
@@ -230,13 +230,17 @@ class FeishuKBAdapter(KBAdapter):
                         source_type="feishu",
                     )
                 )
-            return sources if sources else [
+            return (
+                sources
+                if sources
+                else [
                     SourceInfo(
                         source_id=self._source_id,
                         source_name=self._source_name,
                         source_type=self._source_type,
                     )
                 ]
+            )
         except Exception as e:
             logger.error(f"Feishu list_sources error: {e}")
             return [

src/agentkit/memory/adapters/generic_http.py

@@ -7,7 +7,6 @@
 from __future__ import annotations
 
 import logging
-from typing import Any
 
 import httpx
 
@@ -55,7 +54,9 @@ class GenericHTTPAdapter(KBAdapter):
         )
         self._endpoint_url = endpoint_url.rstrip("/")
         if not is_safe_url(self._endpoint_url):
-            raise ValueError(f"Unsafe endpoint_url: {self._endpoint_url}. Private/internal URLs are not allowed.")
+            raise ValueError(
+                f"Unsafe endpoint_url: {self._endpoint_url}. Private/internal URLs are not allowed."
+            )
         self._auth_config = auth_config or {}
         self._extra_headers = headers or {}
 
@@ -74,12 +75,11 @@ class GenericHTTPAdapter(KBAdapter):
                 headers["Authorization"] = f"Bearer {token}"
         elif auth_type == "basic":
             import base64
+
             username = self._auth_config.get("username", "")
             password = self._auth_config.get("password", "")
             if username and password:
-                credentials = base64.b64encode(
-                    f"{username}:{password}".encode()
-                ).decode()
+                credentials = base64.b64encode(f"{username}:{password}".encode()).decode()
                 headers["Authorization"] = f"Basic {credentials}"
         elif auth_type == "api_key":
             key_name = self._auth_config.get("header_name", "X-API-Key")
@@ -135,8 +135,7 @@ class GenericHTTPAdapter(KBAdapter):
 
         except httpx.HTTPStatusError as e:
             logger.error(
-                f"GenericHTTP search HTTP error: {e.response.status_code} — "
-                f"{e.response.text[:200]}"
+                f"GenericHTTP search HTTP error: {e.response.status_code} — {e.response.text[:200]}"
             )
             return []
         except Exception as e:
@@ -177,8 +176,7 @@ class GenericHTTPAdapter(KBAdapter):
 
         except httpx.HTTPStatusError as e:
             logger.error(
-                f"GenericHTTP ingest HTTP error: {e.response.status_code} — "
-                f"{e.response.text[:200]}"
+                f"GenericHTTP ingest HTTP error: {e.response.status_code} — {e.response.text[:200]}"
             )
             return []
         except Exception as e:
@@ -245,13 +243,17 @@ class GenericHTTPAdapter(KBAdapter):
                                 document_count=item.get("document_count", 0),
                             )
                         )
-                return sources if sources else [
+                return (
+                    sources
+                    if sources
+                    else [
                         SourceInfo(
                             source_id=self._source_id,
                             source_name=self._source_name,
                             source_type=self._source_type,
                         )
                     ]
+                )
         except Exception as e:
             logger.debug(f"GenericHTTP list_sources error (endpoint may not exist): {e}")
 

src/agentkit/memory/base.py

@@ -3,19 +3,29 @@
 from abc import ABC, abstractmethod
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
-from typing import Any
+from typing import TypeAlias
+
+# 共享类型别名 — 跨 memory 子系统复用，避免 `Any` 残留。
+# MetadataValue 覆盖 metadata dict 中实际出现的原始类型；
+# MemoryValue 额外允许 dict/list 容器以容纳结构化负载（如 episodic 经验字典）。
+MetadataValue: TypeAlias = str | int | float | bool | None
+MetadataDict: TypeAlias = dict[str, MetadataValue]
+MemoryValue: TypeAlias = (
+    str | int | float | bool | None | dict[str, MetadataValue] | list[MetadataValue]
+)
 
 
 @dataclass
 class MemoryItem:
     """记忆条目"""
+
     key: str
-    value: Any
-    metadata: dict[str, Any] = field(default_factory=dict)
+    value: object
+    metadata: MetadataDict = field(default_factory=dict)
     score: float = 1.0
     created_at: datetime = field(default_factory=lambda: datetime.now(timezone.utc))
 
-    def to_dict(self) -> dict:
+    def to_dict(self) -> dict[str, object]:
         return {
             "key": self.key,
             "value": self.value,
@@ -35,7 +45,7 @@ class Memory(ABC):
     """
 
     @abstractmethod
-    async def store(self, key: str, value: Any, metadata: dict[str, Any] | None = None) -> None:
+    async def store(self, key: str, value: object, metadata: MetadataDict | None = None) -> None:
         """存储记忆"""
         ...
 
@@ -45,7 +55,9 @@ class Memory(ABC):
         ...
 
     @abstractmethod
-    async def search(self, query: str, top_k: int = 5, filters: dict[str, Any] | None = None) -> list[MemoryItem]:
+    async def search(
+        self, query: str, top_k: int = 5, filters: MetadataDict | None = None
+    ) -> list[MemoryItem]:
         """语义检索"""
         ...
 
@@ -54,7 +66,7 @@ class Memory(ABC):
         """删除记忆"""
         ...
 
-    async def store_batch(self, items: list[tuple[str, Any, dict | None]]) -> None:
+    async def store_batch(self, items: list[tuple[str, object, MetadataDict | None]]) -> None:
         """批量存储"""
         for key, value, metadata in items:
             await self.store(key, value, metadata)

src/agentkit/memory/chunking.py

@@ -11,10 +11,18 @@ import logging
 import re
 import uuid
 from dataclasses import dataclass, field
-from typing import Any
+from typing import TypeAlias
+
+from agentkit.memory.base import MetadataDict
 
 logger = logging.getLogger(__name__)
 
+# 分块元数据：source_doc/position/char_count/chunking_strategy/heading/heading_level
+# — 全部为原始标量（str/int）。
+ChunkMetadata: TypeAlias = MetadataDict
+# _split_by_headings 返回的节段结构。
+SectionInfo: TypeAlias = dict[str, str | int]
+
 
 @dataclass
 class Chunk:
@@ -22,7 +30,7 @@ class Chunk:
 
     chunk_id: str
     content: str
-    metadata: dict[str, Any] = field(default_factory=dict)
+    metadata: ChunkMetadata = field(default_factory=dict)
 
     def __post_init__(self) -> None:
         if "source_doc" not in self.metadata:
@@ -30,7 +38,7 @@ class Chunk:
         if "position" not in self.metadata:
             self.metadata["position"] = 0
 
-    def to_dict(self) -> dict[str, Any]:
+    def to_dict(self) -> dict[str, object]:
         return {
             "chunk_id": self.chunk_id,
             "content": self.content,
@@ -57,7 +65,9 @@ class TextChunker:
             separator: 优先分割符
         """
         if chunk_overlap >= chunk_size:
-            raise ValueError(f"chunk_overlap ({chunk_overlap}) must be less than chunk_size ({chunk_size})")
+            raise ValueError(
+                f"chunk_overlap ({chunk_overlap}) must be less than chunk_size ({chunk_size})"
+            )
         self._chunk_size = chunk_size
         self._chunk_overlap = chunk_overlap
         self._separator = separator
@@ -66,7 +76,7 @@ class TextChunker:
         self,
         text: str,
         source_doc_id: str = "",
-        metadata: dict[str, Any] | None = None,
+        metadata: ChunkMetadata | None = None,
     ) -> list[Chunk]:
         """将文本分块
 
@@ -96,11 +106,13 @@ class TextChunker:
             chunk_meta = dict(base_meta)
             chunk_meta["position"] = i
             chunk_meta["char_count"] = len(chunk_text)
-            chunks.append(Chunk(
+            chunks.append(
+                Chunk(
                     chunk_id=str(uuid.uuid4()),
                     content=chunk_text,
                     metadata=chunk_meta,
-            ))
+                )
+            )
 
         return chunks
 
@@ -142,7 +154,9 @@ class TextChunker:
                     overlap_text[overlap_start:], segments
                 )
                 current = overlap_segments
-                current_len = sum(len(s) for s in current) + len(self._separator) * max(0, len(current) - 1)
+                current_len = sum(len(s) for s in current) + len(self._separator) * max(
+                    0, len(current) - 1
+                )
 
             current.append(segment)
             current_len += seg_len + len(self._separator)
@@ -214,7 +228,7 @@ class StructuralChunker:
         self,
         text: str,
         source_doc_id: str = "",
-        metadata: dict[str, Any] | None = None,
+        metadata: ChunkMetadata | None = None,
     ) -> list[Chunk]:
         """将文本按结构分块
 
@@ -266,23 +280,25 @@ class StructuralChunker:
                 chunk_meta["heading"] = heading
                 chunk_meta["heading_level"] = level
                 chunk_meta["char_count"] = len(content)
-                chunks.append(Chunk(
+                chunks.append(
+                    Chunk(
                         chunk_id=str(uuid.uuid4()),
                         content=content,
                         metadata=chunk_meta,
-                ))
+                    )
+                )
                 position += 1
 
         return chunks
 
-    def _split_by_headings(self, text: str) -> list[dict[str, Any]]:
+    def _split_by_headings(self, text: str) -> list[SectionInfo]:
         """按标题分割 Markdown 文本
 
         Returns:
             列表，每项包含 heading, content, level
         """
         lines = text.split("\n")
-        sections: list[dict[str, Any]] = []
+        sections: list[SectionInfo] = []
         current_heading = ""
         current_level = 0
         current_lines: list[str] = []
@@ -296,11 +312,13 @@ class StructuralChunker:
                 if current_lines:
                     content = "\n".join(current_lines).strip()
                     if content:
-                        sections.append({
+                        sections.append(
+                            {
                                 "heading": current_heading,
                                 "content": content,
                                 "level": current_level,
-                        })
+                            }
+                        )
 
                 # 开始新节
                 current_heading = match.group(2).strip()
@@ -313,18 +331,22 @@ class StructuralChunker:
         if current_lines:
             content = "\n".join(current_lines).strip()
             if content:
-                sections.append({
+                sections.append(
+                    {
                         "heading": current_heading,
                         "content": content,
                         "level": current_level,
-                })
+                    }
+                )
 
         # 如果没有标题结构，整体作为一个块
         if not sections:
-            sections.append({
+            sections.append(
+                {
                     "heading": "",
                     "content": text.strip(),
                     "level": 0,
-            })
+                }
+            )
 
         return sections

src/agentkit/memory/contextual_retrieval.py

@@ -9,10 +9,14 @@ from __future__ import annotations
 import hashlib
 import logging
 from dataclasses import dataclass
-from typing import Any
+from typing import TYPE_CHECKING
 
+from agentkit.memory.base import MetadataDict
 from agentkit.memory.embedder import EmbeddingCache
 
+if TYPE_CHECKING:
+    from agentkit.llm.gateway import LLMGateway
+
 logger = logging.getLogger(__name__)
 
 
@@ -24,7 +28,7 @@ class ContextualChunk:
     context_prefix: str
     enhanced_content: str
     chunk_index: int
-    metadata: dict[str, Any]
+    metadata: MetadataDict
 
     @property
     def content(self) -> str:
@@ -65,7 +69,7 @@ class ContextualChunker:
 
     def __init__(
         self,
-        llm_gateway: Any = None,
+        llm_gateway: LLMGateway | None = None,
         cache: EmbeddingCache | None = None,
         batch_size: int = 8,
         max_context_length: int = 200,
@@ -90,7 +94,7 @@ class ContextualChunker:
         self,
         document: str,
         chunks: list[str],
-        metadata: dict[str, Any] | None = None,
+        metadata: MetadataDict | None = None,
     ) -> list[ContextualChunk]:
         """为文档块添加上下文前缀
 
@@ -134,7 +138,7 @@ class ContextualChunker:
         document: str,
         chunks: list[str],
         start_index: int,
-        metadata: dict[str, Any] | None,
+        metadata: MetadataDict | None,
     ) -> list[ContextualChunk]:
         """处理一批文档块"""
         results: list[ContextualChunk] = []

src/agentkit/memory/document_loader.py

@@ -12,7 +12,9 @@ import uuid
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Any
+from typing import TypeAlias
+
+from agentkit.memory.base import MetadataDict
 
 logger = logging.getLogger(__name__)
 
@@ -23,6 +25,11 @@ MAX_CONTENT_SIZE = 100 * 1024 * 1024  # 100MB
 MAX_ROWS_PER_SHEET = 10_000
 MAX_CELL_CHARS = 10_000
 
+# 文档元数据：source/format/parser/page_count/table_count/sheet_count/row_count/
+# heading_count/created_at/title/truncated — 全部为原始标量。
+DocumentMetadata: TypeAlias = MetadataDict
+ParseResult: TypeAlias = tuple[str, DocumentMetadata]
+
 
 @dataclass
 class Document:
@@ -31,7 +38,7 @@ class Document:
     doc_id: str
     title: str
     content: str
-    metadata: dict[str, Any] = field(default_factory=dict)
+    metadata: DocumentMetadata = field(default_factory=dict)
 
     def __post_init__(self) -> None:
         if "source" not in self.metadata:
@@ -43,7 +50,7 @@ class Document:
         if "created_at" not in self.metadata:
             self.metadata["created_at"] = datetime.now(timezone.utc).isoformat()
 
-    def to_dict(self) -> dict[str, Any]:
+    def to_dict(self) -> dict[str, object]:
         return {
             "doc_id": self.doc_id,
             "title": self.title,
@@ -136,12 +143,14 @@ class DocumentLoader:
 
         parser = parsers.get(doc_format)
         if parser is None:
-            logger.warning(f"Unsupported format '{doc_format}' for {filename}, falling back to text")
+            logger.warning(
+                f"Unsupported format '{doc_format}' for {filename}, falling back to text"
+            )
             parser = self._parse_text
 
         text, extra_meta = parser(content, filename)
 
-        metadata: dict[str, Any] = {
+        metadata: DocumentMetadata = {
             "source": filename,
             "format": doc_format,
             "created_at": datetime.now(timezone.utc).isoformat(),
@@ -159,7 +168,7 @@ class DocumentLoader:
             metadata=metadata,
         )
 
-    def _parse_pdf(self, content: bytes, filename: str) -> tuple[str, dict[str, Any]]:
+    def _parse_pdf(self, content: bytes, filename: str) -> ParseResult:
         """解析 PDF 文件
 
         优先使用 PyMuPDF (fitz)，回退到 pdfplumber，最终回退到纯文本。
@@ -215,7 +224,7 @@ class DocumentLoader:
         logger.warning(f"No PDF parser available for {filename}, falling back to text extraction")
         return self._parse_text(content, filename)
 
-    def _parse_docx(self, content: bytes, filename: str) -> tuple[str, dict[str, Any]]:
+    def _parse_docx(self, content: bytes, filename: str) -> ParseResult:
         """解析 Word 文件
 
         使用 python-docx，回退到纯文本。
@@ -259,7 +268,7 @@ class DocumentLoader:
             logger.warning(f"python-docx parsing failed for {filename}: {e}")
             return self._parse_text(content, filename)
 
-    def _parse_xlsx(self, content: bytes, filename: str) -> tuple[str, dict[str, Any]]:
+    def _parse_xlsx(self, content: bytes, filename: str) -> ParseResult:
         """解析 Excel 文件
 
         使用 openpyxl，回退到纯文本。每个 sheet 转为 Markdown 表格，
@@ -313,7 +322,7 @@ class DocumentLoader:
             finally:
                 wb.close()
             text = "\n".join(sections).strip()
-            meta: dict[str, Any] = {
+            meta: DocumentMetadata = {
                 "parser": "openpyxl",
                 "sheet_count": sheet_count,
                 "row_count": total_rows,
@@ -328,7 +337,7 @@ class DocumentLoader:
             logger.warning(f"openpyxl parsing failed for {filename}: {e}")
             return self._parse_text(content, filename)
 
-    def _parse_markdown(self, content: bytes, filename: str) -> tuple[str, dict[str, Any]]:
+    def _parse_markdown(self, content: bytes, filename: str) -> ParseResult:
         """解析 Markdown 文件
 
         使用 mistune（如果可用），否则直接读取文本。
@@ -347,7 +356,7 @@ class DocumentLoader:
                 title = line_stripped.lstrip("#").strip()
                 break
 
-        meta: dict[str, Any] = {
+        meta: DocumentMetadata = {
             "parser": "markdown",
         }
         if title:
@@ -362,7 +371,7 @@ class DocumentLoader:
 
         return text, meta
 
-    def _parse_html(self, content: bytes, filename: str) -> tuple[str, dict[str, Any]]:
+    def _parse_html(self, content: bytes, filename: str) -> ParseResult:
         """解析 HTML 文件
 
         使用 BeautifulSoup 提取文本，回退到纯文本。
@@ -388,7 +397,7 @@ class DocumentLoader:
             if soup.title and soup.title.string:
                 title = soup.title.string.strip()
 
-            meta: dict[str, Any] = {
+            meta: DocumentMetadata = {
                 "parser": "beautifulsoup",
             }
             if title:
@@ -402,7 +411,7 @@ class DocumentLoader:
             logger.warning(f"BeautifulSoup parsing failed for {filename}: {e}")
             return self._parse_text(content, filename)
 
-    def _parse_text(self, content: bytes, filename: str) -> tuple[str, dict[str, Any]]:
+    def _parse_text(self, content: bytes, filename: str) -> ParseResult:
         """解析纯文本文件"""
         try:
             text = content.decode("utf-8")

src/agentkit/memory/embedder.py

@@ -1,12 +1,17 @@
 """Embedder 接口与实现 - 文本向量化"""
 
+from __future__ import annotations
+
 import hashlib
 import logging
 import os
 import time
 from abc import ABC, abstractmethod
 from collections import OrderedDict
-from typing import Any
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    import httpx
 
 logger = logging.getLogger(__name__)
 
@@ -97,13 +102,14 @@ class OpenAIEmbedder(Embedder):
         self._model = model
         self._base_url = base_url
         self._dimension = 1536  # text-embedding-3-small 默认维度
-        self._client: Any = None
+        self._client: httpx.AsyncClient | None = None
         self._cache = cache
 
-    def _get_client(self):
+    def _get_client(self) -> httpx.AsyncClient:
         """Lazily create and reuse a single httpx.AsyncClient."""
         if self._client is None:
             import httpx
+
             self._client = httpx.AsyncClient(timeout=30.0)
         return self._client
 

src/agentkit/memory/episodic.py

@@ -1,17 +1,22 @@
 """Episodic Memory - 基于 pgvector + PostgreSQL 的任务经验记忆"""
 
+from __future__ import annotations
+
 import json
 import logging
 import math
 from datetime import datetime, timezone
-from typing import Any
+from typing import TYPE_CHECKING
 
 from sqlalchemy import text
 
-from agentkit.memory.base import Memory, MemoryItem
+from agentkit.memory.base import Memory, MemoryItem, MetadataDict
 from agentkit.memory.embedder import Embedder
 from agentkit.utils.vector_math import compute_cosine_similarity
 
+if TYPE_CHECKING:
+    from sqlalchemy.ext.asyncio import AsyncSession
+
 logger = logging.getLogger(__name__)
 
 
@@ -28,8 +33,8 @@ class EpisodicMemory(Memory):
 
     def __init__(
         self,
-        session_factory: Any,
-        episodic_model: Any,
+        session_factory: object,
+        episodic_model: object,
         embedder: Embedder | None = None,
         decay_rate: float = 0.01,
         alpha: float = 0.7,
@@ -57,7 +62,7 @@ class EpisodicMemory(Memory):
         self._pgvector_enabled = pgvector_enabled
         self._table_name = table_name
 
-    async def store(self, key: str, value: Any, metadata: dict[str, Any] | None = None) -> None:
+    async def store(self, key: str, value: object, metadata: MetadataDict | None = None) -> None:
         """存储任务经验"""
         async with self._session_factory() as db:
             try:
@@ -68,7 +73,11 @@ class EpisodicMemory(Memory):
                 embedding = None
                 if self._embedder:
                     if isinstance(value, dict):
-                        text = value.get("output_summary", "") or value.get("input_summary", "") or json.dumps(value, ensure_ascii=False)[:500]
+                        text = (
+                            value.get("output_summary", "")
+                            or value.get("input_summary", "")
+                            or json.dumps(value, ensure_ascii=False)[:500]
+                        )
                     else:
                         text = str(value)
                     embedding = await self._embedder.embed(text)
@@ -106,13 +115,11 @@ class EpisodicMemory(Memory):
                 logger.error(f"Failed to retrieve episodic memory: {e}")
                 return None
 
-    async def _retrieve_pgvector(self, db: Any, query_embedding: list[float]) -> MemoryItem | None:
+    async def _retrieve_pgvector(
+        self, db: AsyncSession, query_embedding: list[float]
+    ) -> MemoryItem | None:
         """使用 pgvector ``<=>`` 算符检索最相似条目"""
-        sql = text(
-            f"SELECT * FROM {self._table_name} "
-            f"ORDER BY embedding <=> :query_vec "
-            f"LIMIT :lim"
-        )
+        sql = text(f"SELECT * FROM {self._table_name} ORDER BY embedding <=> :query_vec LIMIT :lim")
         result = await db.execute(sql, {"query_vec": str(query_embedding), "lim": 1})
         row = result.mappings().first()
 
@@ -147,7 +154,9 @@ class EpisodicMemory(Memory):
             created_at=row.get("created_at") or datetime.now(timezone.utc),
         )
 
-    async def _retrieve_client_side(self, db: Any, query_embedding: list[float]) -> MemoryItem | None:
+    async def _retrieve_client_side(
+        self, db: AsyncSession, query_embedding: list[float]
+    ) -> MemoryItem | None:
         """客户端 O(N) cosine similarity 检索（回退路径）"""
         Model = self._episodic_model
         from sqlalchemy import select
@@ -193,7 +202,13 @@ class EpisodicMemory(Memory):
             created_at=best_item.created_at or datetime.now(timezone.utc),
         )
 
-    async def search(self, query: str, top_k: int = 5, filters: dict[str, Any] | None = None, search_multiplier: int = 5) -> list[MemoryItem]:
+    async def search(
+        self,
+        query: str,
+        top_k: int = 5,
+        filters: MetadataDict | None = None,
+        search_multiplier: int = 5,
+    ) -> list[MemoryItem]:
         """语义检索相似历史案例
 
         Args:
@@ -214,10 +229,10 @@ class EpisodicMemory(Memory):
 
     async def _search_pgvector(
         self,
-        db: Any,
+        db: AsyncSession,
         query: str,
         top_k: int,
-        filters: dict[str, Any] | None,
+        filters: MetadataDict | None,
         search_multiplier: int,
     ) -> list[MemoryItem]:
         """使用 pgvector ``<=>`` 算符检索，再 Python 侧 time_decay 重排"""
@@ -225,7 +240,7 @@ class EpisodicMemory(Memory):
         fetch_limit = top_k * search_multiplier
 
         where_clauses = []
-        params: dict[str, Any] = {"query_vec": str(query_embedding), "lim": fetch_limit}
+        params: dict[str, object] = {"query_vec": str(query_embedding), "lim": fetch_limit}
 
         filters = filters or {}
         if filters.get("agent_name"):
@@ -256,7 +271,11 @@ class EpisodicMemory(Memory):
         items = []
         for row in rows:
             row_embedding = row.get("embedding")
-            age_hours = (datetime.now(timezone.utc) - row["created_at"]).total_seconds() / 3600 if row.get("created_at") else 0
+            age_hours = (
+                (datetime.now(timezone.utc) - row["created_at"]).total_seconds() / 3600
+                if row.get("created_at")
+                else 0
+            )
             decay = math.exp(-self._decay_rate * age_hours)
             time_decay_score = (row.get("quality_score") or 0.5) * decay
 
@@ -266,7 +285,8 @@ class EpisodicMemory(Memory):
             else:
                 score = time_decay_score
 
-            items.append(MemoryItem(
+            items.append(
+                MemoryItem(
                     key=str(row.get("id", "")),
                     value={
                         "input_summary": row.get("input_summary", ""),
@@ -278,21 +298,24 @@ class EpisodicMemory(Memory):
                     metadata={
                         "agent_name": row.get("agent_name", ""),
                         "task_type": row.get("task_type", ""),
-                    "created_at": row["created_at"].isoformat() if row.get("created_at") else None,
+                        "created_at": row["created_at"].isoformat()
+                        if row.get("created_at")
+                        else None,
                     },
                     score=score,
                     created_at=row.get("created_at") or datetime.now(timezone.utc),
-            ))
+                )
+            )
 
         items.sort(key=lambda x: x.score, reverse=True)
         return items[:top_k]
 
     async def _search_client_side(
         self,
-        db: Any,
+        db: AsyncSession,
         query: str,
         top_k: int,
-        filters: dict[str, Any] | None,
+        filters: MetadataDict | None,
         search_multiplier: int,
     ) -> list[MemoryItem]:
         """客户端 O(N) cosine similarity 检索（回退路径）"""
@@ -300,6 +323,7 @@ class EpisodicMemory(Memory):
         filters = filters or {}
 
         from sqlalchemy import select
+
         stmt = select(Model)
 
         if filters.get("agent_name"):
@@ -322,7 +346,11 @@ class EpisodicMemory(Memory):
         # 计算得分并构建 MemoryItem
         items = []
         for entry in entries:
-            age_hours = (datetime.now(timezone.utc) - entry.created_at).total_seconds() / 3600 if entry.created_at else 0
+            age_hours = (
+                (datetime.now(timezone.utc) - entry.created_at).total_seconds() / 3600
+                if entry.created_at
+                else 0
+            )
             decay = math.exp(-self._decay_rate * age_hours)
             time_decay_score = (entry.quality_score or 0.5) * decay
 
@@ -333,7 +361,8 @@ class EpisodicMemory(Memory):
             else:
                 score = time_decay_score
 
-            items.append(MemoryItem(
+            items.append(
+                MemoryItem(
                     key=str(entry.id),
                     value={
                         "input_summary": entry.input_summary,
@@ -349,14 +378,17 @@ class EpisodicMemory(Memory):
                     },
                     score=score,
                     created_at=entry.created_at or datetime.now(timezone.utc),
-            ))
+                )
+            )
 
         items.sort(key=lambda x: x.score, reverse=True)
         if len(items) < top_k:
             logger.warning(
                 "EpisodicMemory.search returned %d results after scoring (top_k=%d). "
                 "Consider increasing search_multiplier (current=%d) to avoid missing relevant entries.",
-                len(items), top_k, search_multiplier,
+                len(items),
+                top_k,
+                search_multiplier,
             )
         return items[:top_k]
 
@@ -364,8 +396,9 @@ class EpisodicMemory(Memory):
         """删除指定经验"""
         async with self._session_factory() as db:
             try:
-                from sqlalchemy import select, delete as sql_delete
+                from sqlalchemy import delete as sql_delete
                 import uuid
+
                 Model = self._episodic_model
 
                 stmt = sql_delete(Model).where(Model.id == uuid.UUID(key))

src/agentkit/memory/http_rag.py

@@ -3,13 +3,26 @@
 配置驱动，不直接依赖业务系统代码，通过 base_url + api_key 连接。
 """
 
+from __future__ import annotations
+
 import logging
-from typing import Any
+from typing import TYPE_CHECKING, TypeAlias
 
 import httpx
 
+from agentkit.memory.base import MetadataDict
+
+if TYPE_CHECKING:
+    from agentkit.llm.gateway import LLMGateway
+
 logger = logging.getLogger(__name__)
 
+# 标准化检索结果：id/content/score/source/document_id/document_title/
+# knowledge_base_id/metadata — 值为原始标量或嵌套 dict。
+RAGSearchResult: TypeAlias = dict[str, object]
+# ingest() 写入的文档负载：title/content/source_type/metadata。
+RAGIngestPayload: TypeAlias = dict[str, object]
+
 
 class HttpRAGService:
     """HTTP 客户端，调用业务系统的知识库检索 API
@@ -39,7 +52,7 @@ class HttpRAGService:
         knowledge_base_ids: list[str] | None = None,
         timeout: int = 30,
         contextual_chunking: bool = False,
-        llm_gateway: Any = None,
+        llm_gateway: LLMGateway | None = None,
     ):
         """
         Args:
@@ -74,7 +87,7 @@ class HttpRAGService:
         query: str,
         knowledge_base_ids: list[str] | None = None,
         top_k: int = 5,
-    ) -> list[dict[str, Any]]:
+    ) -> list[RAGSearchResult]:
         """语义检索知识库
 
         Args:
@@ -113,7 +126,8 @@ class HttpRAGService:
             normalized = []
             for r in results:
                 if isinstance(r, dict):
-                    normalized.append({
+                    normalized.append(
+                        {
                             "id": r.get("chunk_id", r.get("id", "")),
                             "content": r.get("content", ""),
                             "score": float(r.get("score", 0.0)),
@@ -121,11 +135,14 @@ class HttpRAGService:
                             "document_id": r.get("document_id", ""),
                             "document_title": r.get("document_title", ""),
                             "metadata": r.get("metadata", {}),
-                    })
+                        }
+                    )
             return normalized
 
         except httpx.HTTPStatusError as e:
-            logger.error(f"RAG search HTTP error: {e.response.status_code} — {e.response.text[:200]}")
+            logger.error(
+                f"RAG search HTTP error: {e.response.status_code} — {e.response.text[:200]}"
+            )
             return []
         except httpx.RequestError as e:
             logger.error(f"RAG search request error: {e}")
@@ -141,7 +158,7 @@ class HttpRAGService:
         top_k: int = 5,
         use_rerank: bool = True,
         use_compression: bool = False,
-    ) -> list[dict[str, Any]]:
+    ) -> list[RAGSearchResult]:
         """增强语义检索知识库（支持 rerank 和 compression）
 
         对每个知识库分别调用 /bases/{kb_id}/retrieve 接口，
@@ -169,7 +186,7 @@ class HttpRAGService:
         }
 
         client = self._get_client()
-        all_results: list[dict[str, Any]] = []
+        all_results: list[RAGSearchResult] = []
 
         for kb_id in kb_ids:
             try:
@@ -189,7 +206,8 @@ class HttpRAGService:
                 # 标准化
                 for r in results:
                     if isinstance(r, dict):
-                        all_results.append({
+                        all_results.append(
+                            {
                                 "id": r.get("chunk_id", r.get("id", "")),
                                 "content": r.get("content", ""),
                                 "score": float(r.get("score", 0.0)),
@@ -198,19 +216,17 @@ class HttpRAGService:
                                 "document_title": r.get("document_title", ""),
                                 "knowledge_base_id": kb_id,
                                 "metadata": r.get("metadata", {}),
-                        })
+                            }
+                        )
 
             except httpx.HTTPStatusError as e:
                 if e.response.status_code == 404:
                     # This KB doesn't support enhanced search — fall back to
                     # standard search for THIS KB only, not all KBs.
                     logger.info(
-                        f"Enhanced search not available for KB {kb_id}, "
-                        f"using standard search"
-                    )
-                    std_result = await self.search(
-                        query, knowledge_base_ids=[kb_id], top_k=top_k
+                        f"Enhanced search not available for KB {kb_id}, using standard search"
                     )
+                    std_result = await self.search(query, knowledge_base_ids=[kb_id], top_k=top_k)
                     all_results.extend(std_result)
                 else:
                     logger.error(
@@ -232,9 +248,9 @@ class HttpRAGService:
     async def ingest(
         self,
         key: str,
-        value: Any,
-        metadata: dict[str, Any] | None = None,
-    ) -> dict[str, Any] | None:
+        value: object,
+        metadata: MetadataDict | None = None,
+    ) -> dict[str, object] | None:
         """写入文档到知识库（可选操作）
 
         When contextual_chunking is enabled and llm_gateway is configured,
@@ -308,5 +324,5 @@ class HttpRAGService:
     async def __aenter__(self) -> "HttpRAGService":
         return self
 
-    async def __aexit__(self, *args: Any) -> None:
+    async def __aexit__(self, *args: object) -> None:
         await self.close()

src/agentkit/memory/knowledge_base.py

@@ -11,7 +11,11 @@ from __future__ import annotations
 
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
-from typing import Any, Protocol, runtime_checkable
+from typing import Protocol, TypeAlias, runtime_checkable
+
+from agentkit.memory.base import MetadataDict
+
+KBMetadata: TypeAlias = MetadataDict
 
 
 @dataclass
@@ -22,7 +26,7 @@ class Document:
     content: str
     title: str = ""
     source_id: str = ""
-    metadata: dict[str, Any] = field(default_factory=dict)
+    metadata: KBMetadata = field(default_factory=dict)
     created_at: datetime = field(default_factory=lambda: datetime.now(timezone.utc))
 
 
@@ -34,7 +38,7 @@ class QueryResult:
     source_id: str
     source_name: str
     score: float
-    metadata: dict[str, Any] = field(default_factory=dict)
+    metadata: KBMetadata = field(default_factory=dict)
     doc_id: str = ""
     title: str = ""
 

src/agentkit/memory/local_rag.py

@@ -11,25 +11,32 @@ from __future__ import annotations
 import json
 import logging
 import re
-import uuid
 from datetime import datetime, timezone
-from typing import Any
-
-_SAFE_TABLE_NAME_PATTERN = re.compile(r'^[a-zA-Z_][a-zA-Z0-9_]*$')
+from typing import TYPE_CHECKING, TypeAlias
 
 from agentkit.memory.chunking import Chunk, StructuralChunker, TextChunker
 from agentkit.memory.document_loader import Document as LoaderDocument
 from agentkit.memory.embedder import Embedder
 from agentkit.memory.knowledge_base import (
     Document,
-    KnowledgeBase,
     QueryResult,
     SourceInfo,
 )
 from agentkit.utils.vector_math import compute_cosine_similarity
 
+if TYPE_CHECKING:
+    from sqlalchemy.ext.asyncio import AsyncSession
+
+_SAFE_TABLE_NAME_PATTERN = re.compile(r"^[a-zA-Z_][a-zA-Z0-9_]*$")
+
 logger = logging.getLogger(__name__)
 
+# InMemoryLocalRAGService 内部存储的文档元信息结构。
+# 字段：title/source_id/format/chunk_ids/metadata/created_at — 值为标量或 list[str]。
+InMemoryDocInfo: TypeAlias = dict[str, object]
+# 内部 chunk 存储结构：content/embedding/metadata/source_doc_id。
+InMemoryChunkInfo: TypeAlias = dict[str, object]
+
 
 def _loader_doc_to_kb_doc(loader_doc: LoaderDocument) -> Document:
     """将 document_loader.Document 转换为 knowledge_base.Document"""
@@ -53,7 +60,7 @@ class LocalRAGService:
 
     def __init__(
         self,
-        session_factory: Any,
+        session_factory: object,
         embedder: Embedder,
         chunk_size: int = 1000,
         chunk_overlap: int = 200,
@@ -75,10 +82,14 @@ class LocalRAGService:
         self._chunk_overlap = chunk_overlap
         self._table_name = table_name
         if not _SAFE_TABLE_NAME_PATTERN.match(self._table_name):
-            raise ValueError(f"Invalid table_name: {self._table_name}. Must match [a-zA-Z_][a-zA-Z0-9_]*")
+            raise ValueError(
+                f"Invalid table_name: {self._table_name}. Must match [a-zA-Z_][a-zA-Z0-9_]*"
+            )
         self._pgvector_enabled = pgvector_enabled
         self._text_chunker = TextChunker(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
-        self._structural_chunker = StructuralChunker(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+        self._structural_chunker = StructuralChunker(
+            chunk_size=chunk_size, chunk_overlap=chunk_overlap
+        )
 
     async def ingest(self, documents: list[Document]) -> list[str]:
         """摄取文档列表
@@ -136,9 +147,7 @@ class LocalRAGService:
             try:
                 from sqlalchemy import text as sql_text
 
-                sql = sql_text(
-                    f"DELETE FROM {self._table_name} WHERE source_doc_id = :doc_id"
-                )
+                sql = sql_text(f"DELETE FROM {self._table_name} WHERE source_doc_id = :doc_id")
                 await db.execute(sql, {"doc_id": id})
                 await db.commit()
                 return True
@@ -171,20 +180,15 @@ class LocalRAGService:
 
                 sources = []
                 for row in rows:
-                    meta = {}
-                    if row.get("doc_metadata"):
-                        try:
-                            meta = json.loads(row["doc_metadata"])
-                        except (json.JSONDecodeError, TypeError):
-                            pass
-
-                    sources.append(SourceInfo(
+                    sources.append(
+                        SourceInfo(
                             source_id=row["source_doc_id"],
                             source_name=row.get("source_title", ""),
                             source_type=row.get("doc_format", "local"),
                             document_count=row.get("chunk_count", 0),
                             last_updated=row["created_at"] if row.get("created_at") else None,
-                    ))
+                        )
+                    )
                 return sources
             except Exception as e:
                 logger.error(f"Failed to list sources: {e}")
@@ -271,7 +275,7 @@ class LocalRAGService:
 
     async def _query_pgvector(
         self,
-        db: Any,
+        db: AsyncSession,
         query_embedding: list[float],
         top_k: int,
     ) -> list[QueryResult]:
@@ -286,10 +290,13 @@ class LocalRAGService:
             f"LIMIT :lim"
         )
 
-        result = await db.execute(sql, {
+        result = await db.execute(
+            sql,
+            {
                 "query_vec": str(query_embedding),
                 "lim": top_k,
-        })
+            },
+        )
         rows = result.mappings().all()
 
         results = []
@@ -306,7 +313,8 @@ class LocalRAGService:
                 except (json.JSONDecodeError, TypeError):
                     pass
 
-            results.append(QueryResult(
+            results.append(
+                QueryResult(
                     content=row["content"],
                     source_id=row["source_doc_id"],
                     source_name=row.get("source_title", ""),
@@ -314,13 +322,14 @@ class LocalRAGService:
                     metadata=chunk_meta,
                     doc_id=row["source_doc_id"],
                     title=row.get("source_title", ""),
-            ))
+                )
+            )
 
         return results
 
     async def _query_client_side(
         self,
-        db: Any,
+        db: AsyncSession,
         query_embedding: list[float],
         top_k: int,
     ) -> list[QueryResult]:
@@ -363,7 +372,8 @@ class LocalRAGService:
                 except (json.JSONDecodeError, TypeError):
                     pass
 
-            candidates.append(QueryResult(
+            candidates.append(
+                QueryResult(
                     content=row["content"],
                     source_id=row["source_doc_id"],
                     source_name=row.get("source_title", ""),
@@ -371,7 +381,8 @@ class LocalRAGService:
                     metadata=chunk_meta,
                     doc_id=row["source_doc_id"],
                     title=row.get("source_title", ""),
-            ))
+                )
+            )
 
         candidates.sort(key=lambda x: x.score, reverse=True)
         return candidates[:top_k]
@@ -398,11 +409,15 @@ class InMemoryLocalRAGService:
         """
         self._embedder = embedder
         self._text_chunker = TextChunker(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
-        self._structural_chunker = StructuralChunker(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+        self._structural_chunker = StructuralChunker(
+            chunk_size=chunk_size, chunk_overlap=chunk_overlap
+        )
 
         # 内存存储
-        self._chunks: dict[str, dict[str, Any]] = {}  # chunk_id → {content, embedding, metadata}
-        self._documents: dict[str, dict[str, Any]] = {}  # doc_id → {title, format, chunk_ids, metadata, created_at}
+        self._chunks: dict[str, InMemoryChunkInfo] = {}  # chunk_id → {content, embedding, metadata}
+        self._documents: dict[
+            str, InMemoryDocInfo
+        ] = {}  # doc_id → {title, format, chunk_ids, metadata, created_at}
 
     async def ingest(self, documents: list[Document]) -> list[str]:
         """摄取文档列表
@@ -459,7 +474,8 @@ class InMemoryLocalRAGService:
             source_doc_id = chunk_data["source_doc_id"]
             doc_info = self._documents.get(source_doc_id, {})
 
-            candidates.append(QueryResult(
+            candidates.append(
+                QueryResult(
                     content=chunk_data["content"],
                     source_id=source_doc_id,
                     source_name=doc_info.get("title", ""),
@@ -467,7 +483,8 @@ class InMemoryLocalRAGService:
                     metadata=chunk_data.get("metadata", {}),
                     doc_id=source_doc_id,
                     title=doc_info.get("title", ""),
-            ))
+                )
+            )
 
         candidates.sort(key=lambda x: x.score, reverse=True)
         return candidates[:top_k]
@@ -488,13 +505,15 @@ class InMemoryLocalRAGService:
         """列出已摄取的文档"""
         sources = []
         for doc_id, doc_info in self._documents.items():
-            sources.append(SourceInfo(
+            sources.append(
+                SourceInfo(
                     source_id=doc_id,
                     source_name=doc_info["title"],
                     source_type=doc_info.get("format", "local"),
                     document_count=len(doc_info.get("chunk_ids", [])),
                     last_updated=doc_info.get("created_at"),
-            ))
+                )
+            )
         return sources
 
     async def health_check(self) -> bool:

src/agentkit/memory/multi_source_retriever.py

@@ -13,7 +13,6 @@ import asyncio
 import hashlib
 import logging
 from dataclasses import replace
-from typing import Any
 
 from agentkit.memory.knowledge_base import KnowledgeBase, QueryResult, SourceInfo
 
@@ -186,15 +185,13 @@ class MultiSourceRetriever:
         Returns:
             所有源的检索结果列表（已应用权重）
         """
+
         async def _query_one(name: str, kb: KnowledgeBase) -> list[QueryResult]:
             try:
                 results = await kb.query(query, top_k=top_k)
                 # 应用权重
                 weight = (weights or {}).get(name, 1.0)
-                return [
-                    replace(r, score=r.score * weight, source_name=name)
-                    for r in results
-                ]
+                return [replace(r, score=r.score * weight, source_name=name) for r in results]
             except Exception as e:
                 logger.error(f"Query failed for source '{name}': {e}")
                 return []

src/agentkit/memory/profile.py

@@ -7,10 +7,10 @@
 from __future__ import annotations
 
 import re
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Any, Callable
+from typing import Callable
 
 
 class MemoryFile:
@@ -26,8 +26,9 @@ class MemoryFile:
 
     """
 
-    def __init__(self, path: Path, char_budget: int | None = None,
-                 protected_sections: set[str] | None = None):
+    def __init__(
+        self, path: Path, char_budget: int | None = None, protected_sections: set[str] | None = None
+    ):
         self.path = Path(path)
         self.char_budget = char_budget
         self._protected_sections = protected_sections or set()
@@ -138,7 +139,7 @@ class MemoryFile:
         for match in re.finditer(r"^## (.+)$", content, re.MULTILINE):
             name = match.group(1).strip()
             start = match.start()
-            next_match = re.search(r"^## ", content[match.end():], re.MULTILINE)
+            next_match = re.search(r"^## ", content[match.end() :], re.MULTILINE)
             if next_match:
                 end = match.end() + next_match.start()
             else:
@@ -146,7 +147,7 @@ class MemoryFile:
             sections.append((name, start, end))
 
         if not sections:
-            return content[:self.char_budget]
+            return content[: self.char_budget]
 
         # 保持原始顺序，标记每个 section 是否受保护
         ordered: list[tuple[str, str, bool]] = []  # (name, text, is_protected)
@@ -222,8 +223,9 @@ class MemoryStore:
 
     """
 
-    def __init__(self, base_dir: Path | str | None = None,
-                 on_change: Callable[[str], None] | None = None):
+    def __init__(
+        self, base_dir: Path | str | None = None, on_change: Callable[[str], None] | None = None
+    ):
         if base_dir is None:
             base_dir = Path.home() / ".agentkit"
         self.base_dir = Path(base_dir)
@@ -238,7 +240,9 @@ class MemoryStore:
             protected_sections={"版本", "更新历史"},
         )
         self._user = MemoryFile(self.base_dir / "memories" / "USER.md", char_budget=USER_BUDGET)
-        self._memory = MemoryFile(self.base_dir / "memories" / "MEMORY.md", char_budget=MEMORY_BUDGET)
+        self._memory = MemoryFile(
+            self.base_dir / "memories" / "MEMORY.md", char_budget=MEMORY_BUDGET
+        )
         self._daily_dir = self.base_dir / "memories" / "daily"
         self._daily_dir.mkdir(parents=True, exist_ok=True)
 
@@ -376,4 +380,5 @@ class MemoryStore:
             self._on_change(new_prompt)
         except Exception:
             import logging
+
             logging.getLogger(__name__).warning("Memory notify_change failed", exc_info=True)

src/agentkit/memory/query_transformer.py

@@ -87,10 +87,22 @@ class RuleQueryTransformer(QueryTransformerBase):
     """
 
     _FILLER_WORDS_CN: list[str] = [
-        "帮我", "请", "一下", "分析", "看看", "告诉我", "想知道", "请问",
+        "帮我",
+        "请",
+        "一下",
+        "分析",
+        "看看",
+        "告诉我",
+        "想知道",
+        "请问",
     ]
     _FILLER_WORDS_EN: list[str] = [
-        "please", "can you", "help me", "could you", "i want to", "i need to",
+        "please",
+        "can you",
+        "help me",
+        "could you",
+        "i want to",
+        "i need to",
     ]
 
     def __init__(
@@ -101,9 +113,7 @@ class RuleQueryTransformer(QueryTransformerBase):
         self._synonyms = synonyms or {}
         self._max_sub_queries = max_sub_queries
         # Pre-compile filler patterns
-        self._filler_patterns_cn = [
-            re.compile(re.escape(w)) for w in self._FILLER_WORDS_CN
-        ]
+        self._filler_patterns_cn = [re.compile(re.escape(w)) for w in self._FILLER_WORDS_CN]
         self._filler_patterns_en = [
             re.compile(re.escape(w), re.IGNORECASE) for w in self._FILLER_WORDS_EN
         ]
@@ -166,7 +176,9 @@ def create_query_transformer(
     """工厂函数：根据策略创建查询改写器"""
     if strategy == "llm":
         if llm_gateway is None:
-            logger.warning("LLM strategy requested but no llm_gateway provided, falling back to NoOp")
+            logger.warning(
+                "LLM strategy requested but no llm_gateway provided, falling back to NoOp"
+            )
             return NoOpQueryTransformer()
         return LLMQueryTransformer(llm_gateway, max_sub_queries=max_sub_queries)
     elif strategy == "rule":

src/agentkit/memory/rag_loop.py

@@ -7,11 +7,11 @@
 from __future__ import annotations
 
 import logging
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 from enum import Enum
-from typing import Any
+from typing import TYPE_CHECKING
 
-from agentkit.memory.base import MemoryItem
+from agentkit.memory.base import MemoryItem, MetadataDict
 from agentkit.memory.query_transformer import QueryTransformerBase, NoOpQueryTransformer
 from agentkit.memory.relevance_scorer import (
     RelevanceScorer,
@@ -19,6 +19,10 @@ from agentkit.memory.relevance_scorer import (
     RetrievalEvaluation,
 )
 
+if TYPE_CHECKING:
+    # 避免与 retriever.py 形成运行时循环导入。
+    from agentkit.memory.retriever import MemoryRetriever
+
 logger = logging.getLogger(__name__)
 
 
@@ -70,7 +74,7 @@ class RAGSelfCorrectionLoop:
 
     def __init__(
         self,
-        retriever: Any,  # MemoryRetriever
+        retriever: MemoryRetriever,
         scorer: RelevanceScorer | None = None,
         query_transformer: QueryTransformerBase | None = None,
         max_retries: int = 3,
@@ -87,7 +91,7 @@ class RAGSelfCorrectionLoop:
         query: str,
         top_k: int = 5,
         token_budget: int = 3000,
-        filters: dict[str, Any] | None = None,
+        filters: MetadataDict | None = None,
     ) -> RAGLoopResult:
         """执行带自纠正的检索
 
@@ -107,8 +111,11 @@ class RAGSelfCorrectionLoop:
         while retry_count <= self._max_retries:
             # RETRIEVE
             items = await self._retriever.retrieve(
-                current_query, top_k=top_k, token_budget=token_budget,
-                filters=filters, _skip_correction=True,
+                current_query,
+                top_k=top_k,
+                token_budget=token_budget,
+                filters=filters,
+                _skip_correction=True,
             )
 
             # EVALUATE
@@ -144,9 +151,7 @@ class RAGSelfCorrectionLoop:
             # CORRECT — rewrite query and retry
             retry_count += 1
             if retry_count <= self._max_retries:
-                current_query = await self._rewrite_query(
-                    query, current_query, evaluation
-                )
+                current_query = await self._rewrite_query(query, current_query, evaluation)
                 continue
 
             # DEGRADE — exceeded max retries
@@ -154,9 +159,7 @@ class RAGSelfCorrectionLoop:
 
         # Degraded result: filter to relevant items and mark low confidence
         relevant_items = [
-            s.item
-            for s in evaluation.scores
-            if s.verdict != RelevanceVerdict.INCORRECT
+            s.item for s in evaluation.scores if s.verdict != RelevanceVerdict.INCORRECT
         ]
         result_items = relevant_items if relevant_items else items
 

src/agentkit/memory/relevance_scorer.py

@@ -6,11 +6,9 @@
 from __future__ import annotations
 
 import logging
-import math
 import re
 from dataclasses import dataclass
 from enum import Enum
-from typing import Any
 
 from agentkit.memory.base import MemoryItem
 
@@ -120,9 +118,7 @@ class RelevanceScorer:
             reason=reason,
         )
 
-    def evaluate(
-        self, query: str, items: list[MemoryItem]
-    ) -> RetrievalEvaluation:
+    def evaluate(self, query: str, items: list[MemoryItem]) -> RetrievalEvaluation:
         """评估一次检索的整体质量"""
         if not items:
             return RetrievalEvaluation(
@@ -134,9 +130,7 @@ class RelevanceScorer:
             )
 
         scores = [self.score_item(query, item) for item in items]
-        relevant_count = sum(
-            1 for s in scores if s.verdict != RelevanceVerdict.INCORRECT
-        )
+        relevant_count = sum(1 for s in scores if s.verdict != RelevanceVerdict.INCORRECT)
         avg_score = sum(s.score for s in scores) / len(scores)
 
         # Overall verdict based on average score and relevant ratio

src/agentkit/memory/retriever.py

@@ -7,19 +7,16 @@ from __future__ import annotations
 
 import asyncio
 import logging
-import math
 from dataclasses import replace
-from datetime import datetime
-from typing import Any
 
-from agentkit.memory.base import Memory, MemoryItem
+from agentkit.memory.base import MemoryItem, MetadataDict
 from agentkit.memory.working import WorkingMemory
 from agentkit.memory.episodic import EpisodicMemory
 from agentkit.memory.semantic import SemanticMemory
 from agentkit.memory.query_transformer import QueryTransformerBase
 from agentkit.memory.rag_loop import RAGSelfCorrectionLoop
 from agentkit.memory.relevance_scorer import RelevanceScorer
-from agentkit.memory.knowledge_base import KnowledgeBase, QueryResult
+from agentkit.memory.knowledge_base import KnowledgeBase
 from agentkit.memory.multi_source_retriever import MultiSourceRetriever
 from agentkit.tools.base import Tool
 
@@ -32,11 +29,11 @@ def _estimate_tokens(text: str) -> int:
     Chinese characters typically use 1-2 tokens each.
     English words typically use 1 token each.
     """
-    cjk_count = sum(1 for c in text if '\u4e00' <= c <= '\u9fff')
+    cjk_count = sum(1 for c in text if "\u4e00" <= c <= "\u9fff")
     non_cjk = text
     for c in text:
-        if '\u4e00' <= c <= '\u9fff':
-            non_cjk = non_cjk.replace(c, ' ')
+        if "\u4e00" <= c <= "\u9fff":
+            non_cjk = non_cjk.replace(c, " ")
     word_count = len(non_cjk.split())
     return cjk_count * 2 + word_count
 
@@ -89,7 +86,7 @@ class MemoryRetriever:
         query: str,
         top_k: int = 5,
         token_budget: int = 3000,
-        filters: dict[str, Any] | None = None,
+        filters: MetadataDict | None = None,
         _skip_correction: bool = False,
         sources: list[str] | None = None,
         source_weights: dict[str, float] | None = None,
@@ -121,9 +118,7 @@ class MemoryRetriever:
                 query, top_k=top_k, token_budget=token_budget, filters=filters
             )
             if result.degraded:
-                logger.warning(
-                    f"RAG self-correction degraded after {result.total_retries} retries"
-                )
+                logger.warning(f"RAG self-correction degraded after {result.total_retries} retries")
             return result.items
         # Query transformation
         if self._query_transformer is not None:
@@ -139,9 +134,7 @@ class MemoryRetriever:
 
         # Sub-query search in parallel
         if sub_queries:
-            sub_tasks = [
-                self._search_layers(sq, top_k, filters) for sq in sub_queries
-            ]
+            sub_tasks = [self._search_layers(sq, top_k, filters) for sq in sub_queries]
             sub_results = await asyncio.gather(*sub_tasks, return_exceptions=True)
             for result in sub_results:
                 if isinstance(result, Exception):
@@ -178,7 +171,7 @@ class MemoryRetriever:
         self,
         query: str,
         top_k: int = 5,
-        filters: dict[str, Any] | None = None,
+        filters: MetadataDict | None = None,
     ) -> list[MemoryItem]:
         """Search all configured memory layers with a single query"""
         tasks = []
@@ -237,7 +230,8 @@ class MemoryRetriever:
         # QueryResult → MemoryItem
         items = []
         for r in kb_results:
-            items.append(MemoryItem(
+            items.append(
+                MemoryItem(
                     key=r.source_id,
                     value=r.content,
                     metadata={
@@ -248,7 +242,8 @@ class MemoryRetriever:
                         "document_title": r.title,
                     },
                     score=r.score,
-            ))
+                )
+            )
 
         # Token 预算管理
         selected = []
@@ -318,7 +313,9 @@ class MemoryRetriever:
         if source == "rag":
             kb_type = item.metadata.get("kb_type", "知识库")
             document_title = item.metadata.get("document_title", "未知文档")
-            return f"### 知识库参考 [来源: {kb_type} | 相关度: {score:.2f} | 文档: {document_title}]"
+            return (
+                f"### 知识库参考 [来源: {kb_type} | 相关度: {score:.2f} | 文档: {document_title}]"
+            )
         elif source == "graph":
             return f"### 知识图谱 [实体: {item.key} | 相关度: {score:.2f}]"
         elif source == "episodic":
@@ -330,7 +327,7 @@ class MemoryRetriever:
             return f"### 参考 [来源: {source} | 相关度: {score:.2f}]"
 
     async def store_episode(
-        self, key: str, value: Any, metadata: dict[str, Any] | None = None
+        self, key: str, value: object, metadata: MetadataDict | None = None
     ) -> None:
         """Store an episode into episodic memory if available.
 
@@ -386,12 +383,14 @@ class RetrieveKnowledgeTool(Tool):
             items = await self._retriever.retrieve(query, top_k=5)
             results = []
             for item in items:
-                results.append({
+                results.append(
+                    {
                         "content": item.value,
                         "score": item.score,
                         "source": item.metadata.get("source", "unknown"),
                         "document_title": item.metadata.get("document_title", ""),
-                })
+                    }
+                )
             return {"query": query, "results": results, "call_count": self._call_count}
         except Exception as e:
             return {"error": str(e), "results": []}

src/agentkit/memory/semantic.py

@@ -3,14 +3,45 @@
 适配器模式，对接外部 RAG 服务和知识图谱。
 """
 
-import logging
-from typing import Any
+from __future__ import annotations
 
-from agentkit.memory.base import Memory, MemoryItem
+import logging
+from typing import TYPE_CHECKING, Protocol
+
+from agentkit.memory.base import Memory, MemoryItem, MetadataDict
+
+if TYPE_CHECKING:
+    from agentkit.memory.http_rag import RAGSearchResult
 
 logger = logging.getLogger(__name__)
 
 
+class _RAGServiceLike(Protocol):
+    """RAG 检索服务最小接口契约（duck-typed）。"""
+
+    async def search(
+        self,
+        query: str,
+        knowledge_base_ids: list[str] | None = ...,
+        top_k: int = ...,
+    ) -> list[RAGSearchResult]: ...
+
+    async def enhanced_search(
+        self,
+        query: str,
+        knowledge_base_ids: list[str] | None = ...,
+        top_k: int = ...,
+        use_rerank: bool = ...,
+        use_compression: bool = ...,
+    ) -> list[RAGSearchResult]: ...
+
+
+class _GraphServiceLike(Protocol):
+    """知识图谱服务最小接口契约（duck-typed）。"""
+
+    async def query(self, query: str, depth: int = ...) -> list[dict[str, object]]: ...
+
+
 class SemanticMemory(Memory):
     """Semantic Memory - 知识库检索
 
@@ -19,8 +50,8 @@ class SemanticMemory(Memory):
 
     def __init__(
         self,
-        rag_service: Any = None,
-        graph_service: Any = None,
+        rag_service: _RAGServiceLike | None = None,
+        graph_service: _GraphServiceLike | None = None,
         knowledge_base_ids: list[str] | None = None,
         search_mode: str = "standard",
         use_rerank: bool = True,
@@ -45,9 +76,9 @@ class SemanticMemory(Memory):
         self._use_compression = use_compression
         self._kb_weights = kb_weights
 
-    async def store(self, key: str, value: Any, metadata: dict[str, Any] | None = None) -> None:
+    async def store(self, key: str, value: object, metadata: MetadataDict | None = None) -> None:
         """Semantic Memory 通常只读，写入委托给 RAG 服务的 ingest 方法"""
-        if self._rag_service and hasattr(self._rag_service, 'ingest'):
+        if self._rag_service and hasattr(self._rag_service, "ingest"):
             await self._rag_service.ingest(key, value, metadata)
         else:
             logger.warning("SemanticMemory.store: no RAG service configured for writing")
@@ -56,7 +87,9 @@ class SemanticMemory(Memory):
         """按 key 精确检索（Semantic Memory 通常不按 key 检索）"""
         return None
 
-    async def search(self, query: str, top_k: int = 5, filters: dict[str, Any] | None = None) -> list[MemoryItem]:
+    async def search(
+        self, query: str, top_k: int = 5, filters: MetadataDict | None = None
+    ) -> list[MemoryItem]:
         """语义检索知识库"""
         items = []
 
@@ -64,7 +97,9 @@ class SemanticMemory(Memory):
         if self._rag_service:
             try:
                 kb_ids = (filters or {}).get("knowledge_base_ids", self._knowledge_base_ids)
-                if self._search_mode == "enhanced" and hasattr(self._rag_service, "enhanced_search"):
+                if self._search_mode == "enhanced" and hasattr(
+                    self._rag_service, "enhanced_search"
+                ):
                     results = await self._rag_service.enhanced_search(
                         query,
                         knowledge_base_ids=kb_ids,
@@ -73,14 +108,17 @@ class SemanticMemory(Memory):
                         use_compression=self._use_compression,
                     )
                 else:
-                    results = await self._rag_service.search(query, knowledge_base_ids=kb_ids, top_k=top_k)
+                    results = await self._rag_service.search(
+                        query, knowledge_base_ids=kb_ids, top_k=top_k
+                    )
                 for r in results:
                     kb_id = r.get("knowledge_base_id", "")
                     score = r.get("score", 0.0)
                     # Apply per-KB weights
                     if self._kb_weights and kb_id in self._kb_weights:
                         score *= self._kb_weights[kb_id]
-                    items.append(MemoryItem(
+                    items.append(
+                        MemoryItem(
                             key=r.get("id", ""),
                             value=r.get("content", ""),
                             metadata={
@@ -90,7 +128,8 @@ class SemanticMemory(Memory):
                                 "knowledge_base_id": kb_id,
                             },
                             score=score,
-                    ))
+                        )
+                    )
             except Exception as e:
                 logger.error(f"RAG search failed: {e}")
 
@@ -99,7 +138,8 @@ class SemanticMemory(Memory):
             try:
                 graph_results = await self._graph_service.query(query, depth=2)
                 for r in graph_results[:top_k]:
-                    items.append(MemoryItem(
+                    items.append(
+                        MemoryItem(
                             key=r.get("id", ""),
                             value=r.get("content", ""),
                             metadata={
@@ -108,7 +148,8 @@ class SemanticMemory(Memory):
                                 "relations": r.get("relations", []),
                             },
                             score=r.get("score", 0.0),
-                    ))
+                        )
+                    )
             except Exception as e:
                 logger.error(f"Graph search failed: {e}")
 

src/agentkit/memory/working.py

@@ -3,11 +3,10 @@
 import json
 import logging
 from datetime import datetime, timezone
-from typing import Any
 
 import redis.asyncio as aioredis
 
-from agentkit.memory.base import Memory, MemoryItem
+from agentkit.memory.base import Memory, MemoryItem, MetadataDict
 
 logger = logging.getLogger(__name__)
 
@@ -32,7 +31,7 @@ class WorkingMemory(Memory):
     def _make_key(self, key: str) -> str:
         return f"{self._key_prefix}:{key}"
 
-    async def store(self, key: str, value: Any, metadata: dict[str, Any] | None = None) -> None:
+    async def store(self, key: str, value: object, metadata: MetadataDict | None = None) -> None:
         redis_key = self._make_key(key)
         item = MemoryItem(
             key=key,
@@ -57,10 +56,14 @@ class WorkingMemory(Memory):
             value=item_dict["value"],
             metadata=item_dict.get("metadata", {}),
             score=item_dict.get("score", 1.0),
-            created_at=datetime.fromisoformat(item_dict["created_at"]) if item_dict.get("created_at") else datetime.now(timezone.utc),
+            created_at=datetime.fromisoformat(item_dict["created_at"])
+            if item_dict.get("created_at")
+            else datetime.now(timezone.utc),
         )
 
-    async def search(self, query: str, top_k: int = 5, filters: dict[str, Any] | None = None) -> list[MemoryItem]:
+    async def search(
+        self, query: str, top_k: int = 5, filters: MetadataDict | None = None
+    ) -> list[MemoryItem]:
         """Working Memory 不支持语义检索，按 key 前缀匹配"""
         pattern = self._make_key(f"{query}*")
         keys = []
@@ -74,13 +77,15 @@ class WorkingMemory(Memory):
             data = await self._redis.get(key)
             if data:
                 item_dict = json.loads(data)
-                items.append(MemoryItem(
+                items.append(
+                    MemoryItem(
                         key=item_dict["key"],
                         value=item_dict["value"],
                         metadata=item_dict.get("metadata", {}),
                         score=1.0,
                         created_at=datetime.now(timezone.utc),
-                ))
+                    )
+                )
         return items
 
     async def delete(self, key: str) -> bool:

src/agentkit/orchestrator/checkpoint.py

@@ -17,7 +17,7 @@ import logging
 import time
 from dataclasses import asdict, dataclass, field
 from datetime import datetime, timezone
-from typing import Any
+from typing import TYPE_CHECKING
 
 logger = logging.getLogger(__name__)
 
@@ -25,6 +25,28 @@ _TTL_SECONDS = 7 * 24 * 3600  # 7 days
 _KEY_PREFIX = "agentkit:pipeline:checkpoint"
 
 
+if TYPE_CHECKING:
+    from typing import Protocol
+
+    class _RedisPipelineLike(Protocol):
+        def set(self, key: str, value: str, ex: int | None = None) -> object: ...
+        def zadd(self, name: str, mapping: dict[str, float]) -> object: ...
+        def get(self, key: str) -> object: ...
+        def delete(self, *keys: str) -> object: ...
+        async def execute(self) -> list[object]: ...
+
+    class _RedisLike(Protocol):
+        async def set(self, key: str, value: str, ex: int | None = None) -> object: ...
+        async def get(self, key: str) -> object: ...
+        def pipeline(self) -> _RedisPipelineLike: ...
+        async def zrange(self, name: str, start: int, stop: int) -> list[object]: ...
+
+    class _PlanLike(Protocol):
+        @property
+        def id(self) -> str: ...
+        def to_dict(self) -> dict[str, object]: ...
+
+
 @dataclass
 class CheckpointData:
     """单个阶段的 checkpoint 数据。"""
@@ -33,15 +55,15 @@ class CheckpointData:
     phase_id: str
     phase_name: str
     phase_status: str
-    phase_result: dict[str, Any] | None = None
+    phase_result: dict[str, object] | None = None
     plan_status: str = ""
     saved_at: str = field(default_factory=lambda: datetime.now(timezone.utc).isoformat())
 
-    def to_dict(self) -> dict[str, Any]:
+    def to_dict(self) -> dict[str, object]:
         return asdict(self)
 
     @classmethod
-    def from_dict(cls, data: dict[str, Any]) -> CheckpointData:
+    def from_dict(cls, data: dict[str, object]) -> CheckpointData:
         return cls(
             plan_id=data.get("plan_id", ""),
             phase_id=data.get("phase_id", ""),
@@ -67,7 +89,7 @@ class PipelineCheckpoint:
 
     def __init__(
         self,
-        redis_client: Any = None,
+        redis_client: _RedisLike | None = None,
         prefix: str = _KEY_PREFIX,
         ttl_seconds: int = _TTL_SECONDS,
     ) -> None:
@@ -78,7 +100,7 @@ class PipelineCheckpoint:
         # P1 #6: 改用 dict keyed by phase_id，避免重复 append
         self._memory: dict[str, dict[str, CheckpointData]] = {}
         # 内存降级存储：plan_id → (plan_dict, saved_timestamp)
-        self._memory_plans: dict[str, tuple[dict[str, Any], float]] = {}
+        self._memory_plans: dict[str, tuple[dict[str, object], float]] = {}
 
     def _is_expired(self, saved_at: str) -> bool:
         """检查 checkpoint 是否已过期（内存模式 TTL）。"""
@@ -102,7 +124,7 @@ class PipelineCheckpoint:
         """完整 plan JSON 的存储键。"""
         return f"{self._prefix}:plan:{plan_id}"
 
-    async def save_plan(self, plan: Any) -> None:
+    async def save_plan(self, plan: _PlanLike) -> None:
         """保存完整 TeamPlan（用于 resume 重建）。
 
         Args:
@@ -121,7 +143,7 @@ class PipelineCheckpoint:
             except Exception as e:
                 logger.warning(f"PipelineCheckpoint.save_plan Redis failed for plan {plan_id}: {e}")
 
-    async def load_plan(self, plan_id: str) -> dict[str, Any] | None:
+    async def load_plan(self, plan_id: str) -> dict[str, object] | None:
         """加载完整 plan JSON。"""
         # 优先 Redis
         if self._redis is not None:
@@ -142,7 +164,7 @@ class PipelineCheckpoint:
             return None
         return plan_dict
 
-    async def save(self, plan_id: str, phase: Any, plan_status: str) -> None:
+    async def save(self, plan_id: str, phase: object, plan_status: str) -> None:
         """保存阶段 checkpoint。
 
         Args:
@@ -212,7 +234,8 @@ class PipelineCheckpoint:
                 if not phase_ids:
                     # Redis 无数据，检查内存（过滤过期）
                     return [
-                        c for c in self._memory.get(plan_id, {}).values()
+                        c
+                        for c in self._memory.get(plan_id, {}).values()
                         if not self._is_expired(c.saved_at)
                     ]
 
@@ -236,8 +259,7 @@ class PipelineCheckpoint:
 
         # 内存降级（过滤过期 checkpoint）
         return [
-            c for c in self._memory.get(plan_id, {}).values()
-            if not self._is_expired(c.saved_at)
+            c for c in self._memory.get(plan_id, {}).values() if not self._is_expired(c.saved_at)
         ]
 
     async def clear(self, plan_id: str) -> None:

src/agentkit/orchestrator/compensation.py

@@ -1,8 +1,8 @@
 """Saga compensation pattern for Pipeline execution"""
 
 import logging
-from dataclasses import dataclass, field
-from typing import Any, Awaitable, Callable
+from dataclasses import dataclass
+from typing import Awaitable, Callable
 
 logger = logging.getLogger(__name__)
 
@@ -12,7 +12,7 @@ class CompletedStep:
     """Record of a completed step with its compensation"""
 
     step_name: str
-    result: Any
+    result: object
     compensate_action: str | None = None
 
 
@@ -28,9 +28,7 @@ class CompensationResult:
 class SagaOrchestrator:
     """Orchestrates LIFO compensation for failed pipelines"""
 
-    def __init__(
-        self, execute_skill_func: Callable[..., Awaitable[Any]] | None = None
-    ):
+    def __init__(self, execute_skill_func: Callable[..., Awaitable[object]] | None = None):
         """
         Args:
             execute_skill_func: Async function to execute a skill by name
@@ -42,7 +40,7 @@ class SagaOrchestrator:
     def record_completed(
         self,
         step_name: str,
-        result: Any,
+        result: object,
         compensate_action: str | None = None,
     ):
         """Record a completed step for potential compensation"""
@@ -59,9 +57,7 @@ class SagaOrchestrator:
         results: list[CompensationResult] = []
         for step in reversed(self._completed_steps):
             if step.compensate_action is None:
-                logger.info(
-                    f"No compensation for step '{step.step_name}', skipping"
-                )
+                logger.info(f"No compensation for step '{step.step_name}', skipping")
                 results.append(
                     CompensationResult(
                         step_name=step.step_name,
@@ -82,9 +78,7 @@ class SagaOrchestrator:
                     )
                 )
             except Exception as e:
-                logger.error(
-                    f"Compensation for step '{step.step_name}' failed: {e}"
-                )
+                logger.error(f"Compensation for step '{step.step_name}' failed: {e}")
                 results.append(
                     CompensationResult(
                         step_name=step.step_name,

src/agentkit/orchestrator/dynamic_pipeline.py

@@ -4,7 +4,6 @@
 """
 
 import logging
-from typing import Any
 
 from agentkit.orchestrator.pipeline_engine import PipelineEngine
 from agentkit.orchestrator.pipeline_schema import Pipeline, PipelineResult, StageStatus
@@ -15,7 +14,7 @@ logger = logging.getLogger(__name__)
 class DynamicPipeline:
     """动态 Pipeline 组合器"""
 
-    def __init__(self, engine: PipelineEngine, loader: Any = None):
+    def __init__(self, engine: PipelineEngine, loader: object | None = None):
         self._engine = engine
         self._loader = loader
 
@@ -23,7 +22,7 @@ class DynamicPipeline:
         self,
         pipelines: dict[str, Pipeline],
         condition_key: str,
-        context: dict[str, Any] | None = None,
+        context: dict[str, object] | None = None,
     ) -> PipelineResult:
         """根据条件选择子 Pipeline 执行"""
         context = context or {}
@@ -37,14 +36,16 @@ class DynamicPipeline:
             )
 
         selected = pipelines[condition_value]
-        logger.info(f"DynamicPipeline selected '{selected.name}' for {condition_key}={condition_value}")
+        logger.info(
+            f"DynamicPipeline selected '{selected.name}' for {condition_key}={condition_value}"
+        )
         return await self._engine.execute(selected, context)
 
     async def execute_nested(
         self,
         parent: Pipeline,
         sub_pipeline_map: dict[str, Pipeline],
-        context: dict[str, Any] | None = None,
+        context: dict[str, object] | None = None,
     ) -> PipelineResult:
         """执行嵌套 Pipeline"""
         # 先执行父 Pipeline
@@ -52,7 +53,7 @@ class DynamicPipeline:
 
         # 根据父 Pipeline 结果选择子 Pipeline
         for stage_name, stage_result in parent_result.stage_results.items():
-            if hasattr(stage_result, 'output_data') and stage_result.output_data:
+            if hasattr(stage_result, "output_data") and stage_result.output_data:
                 sub_pipeline_name = stage_result.output_data.get("sub_pipeline")
                 if sub_pipeline_name and sub_pipeline_name in sub_pipeline_map:
                     sub = sub_pipeline_map[sub_pipeline_name]
@@ -66,7 +67,7 @@ class DynamicPipeline:
         pipeline: Pipeline,
         max_iterations: int = 5,
         exit_condition: str = "done",
-        context: dict[str, Any] | None = None,
+        context: dict[str, object] | None = None,
     ) -> PipelineResult:
         """循环执行 Pipeline 直到条件满足"""
         current_context = context or {}

src/agentkit/orchestrator/handoff.py

@@ -3,25 +3,42 @@
 import asyncio
 import json
 import logging
-from typing import Any
+from typing import Awaitable, Callable, Protocol
 
 from agentkit.core.protocol import HandoffMessage
 
 logger = logging.getLogger(__name__)
 
 
+class _RedisPubSubLike(Protocol):
+    """Structural type for Redis pubsub object."""
+
+    async def subscribe(self, channel: str) -> None: ...
+    async def unsubscribe(self, channel: str) -> None: ...
+    def listen(self) -> object: ...
+
+
+class _RedisLike(Protocol):
+    """Structural type for async Redis client."""
+
+    async def publish(self, channel: str, message: str) -> int: ...
+    def pubsub(self) -> _RedisPubSubLike: ...
+
+
 class HandoffManager:
     """Handoff 管理器
 
     通过 Redis Pub/Sub 管理 Agent 间的任务转交。
     """
 
-    def __init__(self, redis: Any = None, dispatcher: Any = None):
+    def __init__(self, redis: _RedisLike | None = None, dispatcher: object | None = None):
         self._redis = redis
         self._dispatcher = dispatcher
-        self._handlers: dict[str, list[Any]] = {}
+        self._handlers: dict[str, list[Callable[[HandoffMessage], Awaitable[None]]]] = {}
 
-    def register_handler(self, agent_name: str, handler: Any) -> None:
+    def register_handler(
+        self, agent_name: str, handler: Callable[[HandoffMessage], Awaitable[None]]
+    ) -> None:
         """注册 Handoff 处理器"""
         if agent_name not in self._handlers:
             self._handlers[agent_name] = []

src/agentkit/orchestrator/pipeline_engine.py

@@ -1,10 +1,12 @@
 """Pipeline Engine - DAG + 并行执行 + 步骤重试 + Saga 补偿"""
 
+from __future__ import annotations
+
 import asyncio
 import logging
 from collections import defaultdict
 from datetime import datetime, timezone
-from typing import Any
+from typing import TYPE_CHECKING
 
 from agentkit.orchestrator.compensation import SagaOrchestrator
 from agentkit.orchestrator.pipeline_schema import (
@@ -25,6 +27,23 @@ from agentkit.orchestrator.retry import execute_with_retry
 logger = logging.getLogger(__name__)
 
 
+if TYPE_CHECKING:
+    from typing import Protocol
+
+    class _DispatcherLike(Protocol):
+        async def dispatch(self, task: object) -> None: ...
+        async def get_task_status(self, task_id: str) -> dict[str, object]: ...
+
+    class _StateManagerLike(Protocol):
+        async def create_execution(self, **kwargs: object) -> str: ...
+        async def update_step(self, **kwargs: object) -> None: ...
+        async def complete_execution(self, **kwargs: object) -> None: ...
+        async def fail_execution(self, **kwargs: object) -> None: ...
+
+    class _LLMGatewayLike(Protocol):
+        async def chat(self, **kwargs: object) -> object: ...
+
+
 class PipelineEngine:
     """Pipeline 执行引擎
 
@@ -38,7 +57,12 @@ class PipelineEngine:
     - 状态持久化（可选）
     """
 
-    def __init__(self, dispatcher: Any = None, state_manager: Any = None, llm_gateway: Any = None):
+    def __init__(
+        self,
+        dispatcher: _DispatcherLike | None = None,
+        state_manager: _StateManagerLike | None = None,
+        llm_gateway: _LLMGatewayLike | None = None,
+    ):
         self._dispatcher = dispatcher
         self._state_manager = state_manager
         self._llm_gateway = llm_gateway
@@ -46,7 +70,7 @@ class PipelineEngine:
     async def execute(
         self,
         pipeline: Pipeline,
-        context: dict[str, Any] | None = None,
+        context: dict[str, object] | None = None,
         adaptive_config: AdaptiveConfig | None = None,
     ) -> PipelineResult:
         """执行 Pipeline
@@ -68,7 +92,7 @@ class PipelineEngine:
     async def _adaptive_loop(
         self,
         pipeline: Pipeline,
-        context: dict[str, Any] | None,
+        context: dict[str, object] | None,
         failed_result: PipelineResult,
         adaptive_config: AdaptiveConfig,
     ) -> PipelineResult:
@@ -92,34 +116,30 @@ class PipelineEngine:
 
             # Replan
             new_pipeline = await replanner.replan(current_pipeline, current_result, report)
-            logger.info(f"Pipeline replanned: {new_pipeline.name} ({len(new_pipeline.stages)} stages)")
+            logger.info(
+                f"Pipeline replanned: {new_pipeline.name} ({len(new_pipeline.stages)} stages)"
+            )
 
             # Re-execute
             current_result = await self._execute_pipeline(new_pipeline, context)
             current_pipeline = new_pipeline
 
             # Record reflection in metadata
-            current_result.metadata["reflections"] = [
-                r.model_dump() for r in reflections
-            ]
+            current_result.metadata["reflections"] = [r.model_dump() for r in reflections]
 
             if current_result.status == StageStatus.COMPLETED:
                 logger.info(f"Pipeline succeeded after {reflection_num} reflection(s)")
                 return current_result
 
         # Exhausted reflections
-        logger.warning(
-            f"Pipeline failed after {adaptive_config.max_reflections} reflection(s)"
-        )
-        current_result.metadata["reflections"] = [
-            r.model_dump() for r in reflections
-        ]
+        logger.warning(f"Pipeline failed after {adaptive_config.max_reflections} reflection(s)")
+        current_result.metadata["reflections"] = [r.model_dump() for r in reflections]
         return current_result
 
     async def _execute_pipeline(
         self,
         pipeline: Pipeline,
-        context: dict[str, Any] | None = None,
+        context: dict[str, object] | None = None,
     ) -> PipelineResult:
         """执行 Pipeline 的核心逻辑（不含反思-重规划）。"""
         result = PipelineResult(pipeline_name=pipeline.name)
@@ -151,7 +171,9 @@ class PipelineEngine:
 
         # 逐层执行
         for level, stages in enumerate(level_groups):
-            logger.info(f"Pipeline '{pipeline.name}' executing level {level} with {len(stages)} stage(s)")
+            logger.info(
+                f"Pipeline '{pipeline.name}' executing level {level} with {len(stages)} stage(s)"
+            )
 
             # 并行执行同层 stages
             tasks = []
@@ -173,9 +195,11 @@ class PipelineEngine:
                 # Update step state
                 if self._state_manager is not None and execution_id is not None:
                     try:
-                        step_status = "completed" if sr.status == StageStatus.COMPLETED else sr.status.value
-                        step_output = sr.output_data if hasattr(sr, 'output_data') else None
-                        step_error = sr.error_message if hasattr(sr, 'error_message') else None
+                        step_status = (
+                            "completed" if sr.status == StageStatus.COMPLETED else sr.status.value
+                        )
+                        step_output = sr.output_data if hasattr(sr, "output_data") else None
+                        step_error = sr.error_message if hasattr(sr, "error_message") else None
                         await self._state_manager.update_step(
                             execution_id=execution_id,
                             step_name=stage.name,
@@ -189,19 +213,21 @@ class PipelineEngine:
                 # 收集输出变量
                 if sr.output_data and isinstance(sr, dict):
                     pass
-                elif hasattr(sr, 'output_data') and sr.output_data:
+                elif hasattr(sr, "output_data") and sr.output_data:
                     for output_key in stage.outputs:
                         if output_key in sr.output_data:
                             result.variables[output_key] = sr.output_data[output_key]
 
                 # 检查是否需要中止
-                if hasattr(sr, 'status') and sr.status == StageStatus.FAILED:
+                if hasattr(sr, "status") and sr.status == StageStatus.FAILED:
                     if not stage.continue_on_failure:
                         # Execute Saga compensation for completed steps
                         compensation_results = await saga.compensate()
                         if compensation_results:
                             failed_compensations = [
-                                cr for cr in compensation_results if not cr.success and cr.error != "no_compensation_needed"
+                                cr
+                                for cr in compensation_results
+                                if not cr.success and cr.error != "no_compensation_needed"
                             ]
                             if failed_compensations:
                                 logger.warning(
@@ -219,7 +245,12 @@ class PipelineEngine:
                                     step_name=stage.name,
                                     error=result.error_message,
                                 )
-                            except (asyncio.TimeoutError, ConnectionError, RuntimeError, ValueError) as exc:
+                            except (
+                                asyncio.TimeoutError,
+                                ConnectionError,
+                                RuntimeError,
+                                ValueError,
+                            ) as exc:
                                 logger.warning(f"Failed to persist failure state: {exc}")
                         return result
 
@@ -252,7 +283,9 @@ class PipelineEngine:
         started_at = datetime.now(timezone.utc).isoformat()
 
         # 条件检查
-        if stage.condition and not self._evaluate_condition(stage.condition, pipeline_result.variables):
+        if stage.condition and not self._evaluate_condition(
+            stage.condition, pipeline_result.variables
+        ):
             return StageResult(
                 stage_name=stage.name,
                 status=StageStatus.SKIPPED,
@@ -312,7 +345,9 @@ class PipelineEngine:
                 if status["status"] in ("completed", "failed", "cancelled"):
                     return StageResult(
                         stage_name=stage.name,
-                        status=StageStatus.COMPLETED if status["status"] == "completed" else StageStatus.FAILED,
+                        status=StageStatus.COMPLETED
+                        if status["status"] == "completed"
+                        else StageStatus.FAILED,
                         output_data=status.get("output_data"),
                         error_message=status.get("error_message"),
                         started_at=started_at,
@@ -406,7 +441,7 @@ class PipelineEngine:
         return resolved
 
     @staticmethod
-    def _get_nested(data: dict, path: str) -> Any:
+    def _get_nested(data: dict, path: str) -> object:
         keys = path.split(".")
         current = data
         for key in keys:
@@ -497,9 +532,7 @@ class PipelineEngine:
 
             if verifier_feedback.passed:
                 # 审查通过，返回成功结果
-                logger.info(
-                    f"Stage '{stage.name}' passed review in round {round_num}"
-                )
+                logger.info(f"Stage '{stage.name}' passed review in round {round_num}")
                 worker_result.output_data = worker_result.output_data or {}
                 worker_result.output_data["adversarial_metadata"] = {
                     "passed_round": round_num,
@@ -553,7 +586,7 @@ class PipelineEngine:
         self,
         agent_name: str,
         action: str,
-        input_data: dict[str, Any],
+        input_data: dict[str, object],
         stage: PipelineStage,
         started_at: str,
         timeout_seconds: int | None = None,
@@ -568,7 +601,9 @@ class PipelineEngine:
             started_at: 开始时间
             timeout_seconds: 独立超时时间，不传则使用 stage.timeout_seconds
         """
-        effective_timeout = timeout_seconds if timeout_seconds is not None else stage.timeout_seconds
+        effective_timeout = (
+            timeout_seconds if timeout_seconds is not None else stage.timeout_seconds
+        )
         if self._dispatcher is None:
             # Dry-run 模式
             return StageResult(
@@ -602,7 +637,9 @@ class PipelineEngine:
                 if status["status"] in ("completed", "failed", "cancelled"):
                     return StageResult(
                         stage_name=stage.name,
-                        status=StageStatus.COMPLETED if status["status"] == "completed" else StageStatus.FAILED,
+                        status=StageStatus.COMPLETED
+                        if status["status"] == "completed"
+                        else StageStatus.FAILED,
                         output_data=status.get("output_data"),
                         error_message=status.get("error_message"),
                         started_at=started_at,
@@ -639,7 +676,7 @@ class PipelineEngine:
     async def _execute_verifier(
         self,
         verifier_name: str,
-        worker_output: dict[str, Any],
+        worker_output: dict[str, object],
         stage: PipelineStage,
         started_at: str,
     ) -> ReviewFeedback:
@@ -679,10 +716,7 @@ class PipelineEngine:
         try:
             feedback = ReviewFeedback(
                 passed=output_data.get("passed", False),
-                issues=[
-                    ReviewIssue(**issue)
-                    for issue in output_data.get("issues", [])
-                ],
+                issues=[ReviewIssue(**issue) for issue in output_data.get("issues", [])],
                 summary=output_data.get("summary", "No summary provided"),
                 score=output_data.get("score", 0.0),
             )
@@ -699,7 +733,7 @@ class PipelineEngine:
         self,
         feedback: ReviewFeedback,
         feedback_mode: str = "structured+natural",
-    ) -> dict[str, Any]:
+    ) -> dict[str, object]:
         """构建反馈上下文，让 Worker Agent 理解审查反馈并定向修复
 
         Args:
@@ -720,7 +754,7 @@ class PipelineEngine:
             for issue in feedback.issues
         ]
 
-        feedback_context: dict[str, Any] = {
+        feedback_context: dict[str, object] = {
             "previous_attempt_failed": True,
         }
 
@@ -756,7 +790,9 @@ class PipelineEngine:
             )
         else:
             # 未知模式，fallback 到 structured+natural
-            logger.warning(f"Unknown feedback_mode '{feedback_mode}', falling back to structured+natural")
+            logger.warning(
+                f"Unknown feedback_mode '{feedback_mode}', falling back to structured+natural"
+            )
             feedback_context["review_feedback"] = {
                 "summary": feedback.summary,
                 "issues": issues_list,

src/agentkit/orchestrator/pipeline_loader.py

@@ -2,7 +2,6 @@
 
 import logging
 from pathlib import Path
-from typing import Any
 
 import yaml
 
@@ -23,7 +22,9 @@ class PipelineLoader:
         if not yaml_path.exists():
             yaml_path = self._pipelines_dir / f"{pipeline_name}.yml"
         if not yaml_path.exists():
-            raise FileNotFoundError(f"Pipeline '{pipeline_name}' not found in {self._pipelines_dir}")
+            raise FileNotFoundError(
+                f"Pipeline '{pipeline_name}' not found in {self._pipelines_dir}"
+            )
 
         content = yaml_path.read_text(encoding="utf-8")
         return self.load_from_yaml(content, pipeline_name)

src/agentkit/orchestrator/pipeline_models.py

@@ -35,9 +35,7 @@ class PipelineExecutionModel(Base):
     )
     completed_at = Column(DateTime)
 
-    __table_args__ = (
-        Index("ix_pipeline_status_created", "status", "created_at"),
-    )
+    __table_args__ = (Index("ix_pipeline_status_created", "status", "created_at"),)
 
 
 class PipelineStepHistoryModel(Base):

src/agentkit/orchestrator/pipeline_schema.py

@@ -1,7 +1,7 @@
 """Pipeline 数据模型"""
 
 from enum import Enum
-from typing import Any, Literal
+from typing import Literal
 
 from pydantic import BaseModel, Field
 
@@ -18,8 +18,11 @@ class StageStatus(str, Enum):
 
 class ReviewIssue(BaseModel):
     """单条审查问题"""
+
     severity: Literal["critical", "major", "minor"] = Field(description="问题严重程度")
-    category: Literal["logic_error", "security", "style", "test_failure", "architecture"] = Field(description="问题类别")
+    category: Literal["logic_error", "security", "style", "test_failure", "architecture"] = Field(
+        description="问题类别"
+    )
     description: str = Field(min_length=1, description="问题描述")
     location: str | None = Field(default=None, description="文件路径/行号")
     suggestion: str | None = Field(default=None, description="修复建议")
@@ -27,6 +30,7 @@ class ReviewIssue(BaseModel):
 
 class ReviewFeedback(BaseModel):
     """Verifier 返回的结构化审查反馈"""
+
     passed: bool = Field(description="是否通过审查")
     issues: list[ReviewIssue] = Field(default_factory=list, description="问题列表")
     summary: str = Field(min_length=1, description="自然语言审查报告")
@@ -35,6 +39,7 @@ class ReviewFeedback(BaseModel):
 
 class AdversarialState(BaseModel):
     """对抗轮次状态追踪"""
+
     current_round: int = Field(default=0, description="当前对抗轮次")
     max_rounds: int = Field(default=3, description="最大对抗轮次")
     feedback_history: list[ReviewFeedback] = Field(default_factory=list, description="反馈历史")
@@ -46,7 +51,7 @@ class PipelineStage(BaseModel):
     agent: str
     action: str
     depends_on: list[str] = []
-    inputs: dict[str, Any] = {}
+    inputs: dict[str, object] = {}
     outputs: list[str] = []
     timeout_seconds: int = 300
     retry_count: int = 0
@@ -56,10 +61,17 @@ class PipelineStage(BaseModel):
     compensate: str | None = None
 
     # 对抗闭环相关字段
-    verifier: str | None = Field(default=None, description="Verifier Agent 名称，配置后启用对抗模式")
+    verifier: str | None = Field(
+        default=None, description="Verifier Agent 名称，配置后启用对抗模式"
+    )
     max_adversarial_rounds: int = Field(default=3, description="最大对抗轮次")
-    verifier_timeout_seconds: int = Field(default=120, description="Verifier Agent 独立超时时间（秒），避免与 Worker 共享 timeout_seconds")
-    feedback_mode: Literal["structured+natural", "structured", "natural"] = Field(default="structured+natural", description="反馈模式")
+    verifier_timeout_seconds: int = Field(
+        default=120,
+        description="Verifier Agent 独立超时时间（秒），避免与 Worker 共享 timeout_seconds",
+    )
+    feedback_mode: Literal["structured+natural", "structured", "natural"] = Field(
+        default="structured+natural", description="反馈模式"
+    )
     escalate_on_exhaust: str | None = Field(default=None, description="对抗轮次耗尽后的升级目标")
 
     model_config = {"arbitrary_types_allowed": True}
@@ -70,13 +82,13 @@ class Pipeline(BaseModel):
     version: str
     description: str
     stages: list[PipelineStage]
-    variables: dict[str, Any] = {}
+    variables: dict[str, object] = {}
 
 
 class StageResult(BaseModel):
     stage_name: str
     status: StageStatus = StageStatus.PENDING
-    output_data: dict[str, Any] | None = None
+    output_data: dict[str, object] | None = None
     error_message: str | None = None
     started_at: str | None = None
     completed_at: str | None = None
@@ -86,9 +98,9 @@ class PipelineResult(BaseModel):
     pipeline_name: str
     status: StageStatus = StageStatus.PENDING
     stage_results: dict[str, StageResult] = {}
-    variables: dict[str, Any] = {}
+    variables: dict[str, object] = {}
     error_message: str | None = None
-    metadata: dict[str, Any] = {}
+    metadata: dict[str, object] = {}
 
 
 class AdaptiveConfig(BaseModel):

src/agentkit/orchestrator/pipeline_state.py

@@ -13,7 +13,7 @@ import json
 import logging
 import uuid
 from datetime import datetime, timezone
-from typing import Any, Callable, Coroutine
+from typing import Callable, Coroutine
 
 from agentkit.orchestrator.pipeline_models import (
     PipelineExecutionModel,
@@ -183,7 +183,7 @@ class PipelineStateRedis:
         return self._redis
 
     async def _safe_redis_call(
-        self, fn: Callable[..., Coroutine[Any, Any, Any]], *args: object, **kwargs: object
+        self, fn: Callable[..., Coroutine[object, object, object]], *args: object, **kwargs: object
     ) -> object | None:
         """Execute a Redis call, falling back to memory on failure.
 

src/agentkit/orchestrator/reflection.py

@@ -4,22 +4,30 @@
 生成修正后的 Pipeline 重新执行。
 """
 
+from __future__ import annotations
+
 import json
 import logging
-from typing import Any
+from typing import TYPE_CHECKING
 
 from agentkit.orchestrator.pipeline_schema import (
     Pipeline,
     PipelineResult,
     PipelineStage,
     ReflectionReport,
-    StageResult,
     StageStatus,
 )
 
 logger = logging.getLogger(__name__)
 
 
+if TYPE_CHECKING:
+    from typing import Protocol
+
+    class _LLMGatewayLike(Protocol):
+        async def chat(self, **kwargs: object) -> object: ...
+
+
 class PipelineReflector:
     """分析 Pipeline 执行失败原因，生成结构化反思报告。
 
@@ -27,7 +35,7 @@ class PipelineReflector:
     输出 ReflectionReport 包含 failure_type、root_cause 和 suggested_fix。
     """
 
-    def __init__(self, llm_gateway: Any = None):
+    def __init__(self, llm_gateway: _LLMGatewayLike | None = None):
         self._llm_gateway = llm_gateway
 
     async def reflect(
@@ -54,19 +62,25 @@ class PipelineReflector:
         if self._llm_gateway is not None:
             try:
                 return await self._llm_reflect(
-                    pipeline, failed_stage, error_message,
-                    completed_outputs, reflection_number,
+                    pipeline,
+                    failed_stage,
+                    error_message,
+                    completed_outputs,
+                    reflection_number,
                 )
             except Exception as e:
                 logger.warning(f"LLM reflection failed, falling back to rule-based: {e}")
 
         # 规则兜底：基于错误信息分类
         return self._rule_based_reflect(
-            failed_stage, error_message, reflection_number,
+            failed_stage,
+            error_message,
+            reflection_number,
         )
 
     def _find_failure(
-        self, result: PipelineResult,
+        self,
+        result: PipelineResult,
     ) -> tuple[str, str]:
         """找到第一个失败的 stage 及其错误信息。"""
         for name, sr in result.stage_results.items():
@@ -75,8 +89,9 @@ class PipelineReflector:
         return "", "no failed stage found"
 
     def _collect_completed_outputs(
-        self, result: PipelineResult,
-    ) -> dict[str, Any]:
+        self,
+        result: PipelineResult,
+    ) -> dict[str, object]:
         """收集已完成步骤的输出。"""
         outputs = {}
         for name, sr in result.stage_results.items():
@@ -89,13 +104,16 @@ class PipelineReflector:
         pipeline: Pipeline,
         failed_stage: str,
         error_message: str,
-        completed_outputs: dict[str, Any],
+        completed_outputs: dict[str, object],
         reflection_number: int,
     ) -> ReflectionReport:
         """使用 LLM 分析失败原因。"""
         prompt = self._build_reflection_prompt(
-            pipeline, failed_stage, error_message,
-            completed_outputs, reflection_number,
+            pipeline,
+            failed_stage,
+            error_message,
+            completed_outputs,
+            reflection_number,
         )
 
         response = await self._llm_gateway.chat(
@@ -106,7 +124,9 @@ class PipelineReflector:
         # 解析 LLM 返回的 JSON
         content = response.content if hasattr(response, "content") else str(response)
         return self._parse_reflection_response(
-            content, failed_stage, reflection_number,
+            content,
+            failed_stage,
+            reflection_number,
         )
 
     def _build_reflection_prompt(
@@ -114,15 +134,14 @@ class PipelineReflector:
         pipeline: Pipeline,
         failed_stage: str,
         error_message: str,
-        completed_outputs: dict[str, Any],
+        completed_outputs: dict[str, object],
         reflection_number: int,
     ) -> str:
         """构建反思提示词。"""
         stage_descriptions = []
         for s in pipeline.stages:
             stage_descriptions.append(
-                f"  - {s.name}: agent={s.agent}, action={s.action}, "
-                f"depends_on={s.depends_on}"
+                f"  - {s.name}: agent={s.agent}, action={s.action}, depends_on={s.depends_on}"
             )
 
         completed_summary = json.dumps(
@@ -174,7 +193,9 @@ JSON response:"""
         except (json.JSONDecodeError, KeyError) as e:
             logger.warning(f"Failed to parse LLM reflection response: {e}")
             return self._rule_based_reflect(
-                failed_stage, content, reflection_number,
+                failed_stage,
+                content,
+                reflection_number,
             )
 
     def _rule_based_reflect(
@@ -218,7 +239,7 @@ class PipelineReplanner:
     保留已完成步骤的结果，仅重新规划失败及后续步骤。
     """
 
-    def __init__(self, llm_gateway: Any = None):
+    def __init__(self, llm_gateway: _LLMGatewayLike | None = None):
         self._llm_gateway = llm_gateway
 
     async def replan(
@@ -255,8 +276,7 @@ class PipelineReplanner:
     ) -> Pipeline:
         """使用 LLM 生成修正后的 Pipeline。"""
         completed_stages = [
-            name for name, sr in result.stage_results.items()
-            if sr.status == StageStatus.COMPLETED
+            name for name, sr in result.stage_results.items() if sr.status == StageStatus.COMPLETED
         ]
 
         prompt = f"""Based on the reflection report, generate a corrected pipeline.
@@ -284,7 +304,9 @@ JSON pipeline:"""
         return self._parse_pipeline_response(content, pipeline)
 
     def _parse_pipeline_response(
-        self, content: str, original: Pipeline,
+        self,
+        content: str,
+        original: Pipeline,
     ) -> Pipeline:
         """解析 LLM 返回的 Pipeline JSON。"""
         try:
@@ -294,9 +316,7 @@ JSON pipeline:"""
                 text = "\n".join(lines[1:-1])
 
             data = json.loads(text)
-            stages = [
-                PipelineStage(**s) for s in data.get("stages", [])
-            ]
+            stages = [PipelineStage(**s) for s in data.get("stages", [])]
             return Pipeline(
                 name=data.get("name", original.name),
                 version=data.get("version", original.version),
@@ -316,8 +336,7 @@ JSON pipeline:"""
     ) -> Pipeline:
         """基于规则的兜底重规划。"""
         completed_stages = {
-            name for name, sr in result.stage_results.items()
-            if sr.status == StageStatus.COMPLETED
+            name for name, sr in result.stage_results.items() if sr.status == StageStatus.COMPLETED
         }
 
         # 构建修正后的 stages 列表
@@ -345,17 +364,21 @@ JSON pipeline:"""
         )
 
     def _adjust_failed_stage(
-        self, stage: PipelineStage, report: ReflectionReport,
+        self,
+        stage: PipelineStage,
+        report: ReflectionReport,
     ) -> PipelineStage:
         """根据反思报告调整失败的步骤。"""
-        adjustments: dict[str, Any] = {}
+        adjustments: dict[str, object] = {}
 
         if report.failure_type == "timeout":
             adjustments["timeout_seconds"] = min(
-                stage.timeout_seconds * 2, 3600,
+                stage.timeout_seconds * 2,
+                3600,
             )
             if stage.retry_policy is None:
                 from agentkit.orchestrator.retry import StepRetryPolicy
+
                 adjustments["retry_policy"] = StepRetryPolicy(max_attempts=2)
 
         elif report.failure_type == "resource_error":
@@ -365,6 +388,7 @@ JSON pipeline:"""
             # 添加重试策略，可能输入在后续可用
             if stage.retry_policy is None:
                 from agentkit.orchestrator.retry import StepRetryPolicy
+
                 adjustments["retry_policy"] = StepRetryPolicy(max_attempts=2)
 
         return stage.model_copy(update=adjustments)

src/agentkit/orchestrator/retry.py

@@ -4,7 +4,7 @@ import asyncio
 import logging
 import random
 from dataclasses import dataclass
-from typing import Any, Awaitable, Callable
+from typing import Awaitable, Callable
 
 logger = logging.getLogger(__name__)
 
@@ -27,7 +27,7 @@ class StepRetryPolicy:
     def calculate_delay(self, attempt: int) -> float:
         """Calculate delay for given attempt number (0-based)"""
         delay = min(
-            self.base_delay * (self.exponential_base ** attempt),
+            self.base_delay * (self.exponential_base**attempt),
             self.max_delay,
         )
         if self.jitter:
@@ -36,10 +36,10 @@ class StepRetryPolicy:
 
 
 async def execute_with_retry(
-    func: Callable[..., Awaitable[Any]],
+    func: Callable[..., Awaitable[object]],
     retry_policy: StepRetryPolicy | None = None,
     step_name: str = "",
-) -> Any:
+) -> object:
     """Execute a function with retry policy"""
     if retry_policy is None:
         return await func()

src/agentkit/orchestrator/workflow_schema.py

@@ -3,18 +3,17 @@
 from __future__ import annotations
 
 from datetime import datetime, timezone
-from typing import Any
 
 from pydantic import BaseModel, Field
 
-from agentkit.orchestrator.pipeline_schema import Pipeline, PipelineStage
+from agentkit.orchestrator.pipeline_schema import PipelineStage
 
 
 class WorkflowStage(PipelineStage):
     """A workflow stage extending PipelineStage with type and config."""
 
     type: str = "skill"  # "skill" | "condition" | "approval" | "parallel"
-    config: dict[str, Any] = Field(default_factory=dict)
+    config: dict[str, object] = Field(default_factory=dict)
 
 
 class WorkflowDefinition(BaseModel):
@@ -24,9 +23,9 @@ class WorkflowDefinition(BaseModel):
     name: str
     version: int = 1
     stages: list[WorkflowStage] = Field(default_factory=list)
-    triggers: list[dict[str, Any]] = Field(default_factory=list)
-    variables_schema: dict[str, Any] = Field(default_factory=dict)
-    output_schema: dict[str, Any] = Field(default_factory=dict)
+    triggers: list[dict[str, object]] = Field(default_factory=list)
+    variables_schema: dict[str, object] = Field(default_factory=dict)
+    output_schema: dict[str, object] = Field(default_factory=dict)
     created_at: str = Field(default_factory=lambda: datetime.now(timezone.utc).isoformat())
     updated_at: str = Field(default_factory=lambda: datetime.now(timezone.utc).isoformat())
 
@@ -38,11 +37,11 @@ class WorkflowExecution(BaseModel):
     workflow_id: str = ""
     status: str = "pending"  # pending|running|paused|completed|failed|cancelled
     current_stage: str | None = None
-    stage_results: dict[str, Any] = Field(default_factory=dict)
+    stage_results: dict[str, object] = Field(default_factory=dict)
     started_at: str | None = None
     completed_at: str | None = None
     error: str | None = None
-    variables: dict[str, Any] = Field(default_factory=dict)
+    variables: dict[str, object] = Field(default_factory=dict)
 
 
 class WorkflowSummary(BaseModel):
@@ -62,15 +61,15 @@ class CreateWorkflowRequest(BaseModel):
 
     name: str
     stages: list[WorkflowStage] = Field(default_factory=list)
-    triggers: list[dict[str, Any]] = Field(default_factory=list)
-    variables_schema: dict[str, Any] = Field(default_factory=dict)
-    output_schema: dict[str, Any] = Field(default_factory=dict)
+    triggers: list[dict[str, object]] = Field(default_factory=list)
+    variables_schema: dict[str, object] = Field(default_factory=dict)
+    output_schema: dict[str, object] = Field(default_factory=dict)
 
 
 class ExecuteWorkflowRequest(BaseModel):
     """Request body for executing a workflow."""
 
-    variables: dict[str, Any] = Field(default_factory=dict)
+    variables: dict[str, object] = Field(default_factory=dict)
 
 
 class ApproveRequest(BaseModel):

src/agentkit/server/routes/agents.py

@@ -37,7 +37,7 @@ async def create_agent(request: CreateAgentRequest, req: Request):
         config_dict = request.config
         try:
             config = SkillConfig.from_dict(config_dict)
-        except Exception:
+        except (ValueError, KeyError, TypeError):
             config = AgentConfig.from_dict(config_dict)
         agent = await pool.create_agent(config)
     else:

src/agentkit/server/routes/auth.py

@@ -53,6 +53,8 @@ from agentkit.server.auth.session_service import (
     REVOKE_REASON_PASSWORD_CHANGED,
     REVOKE_REASON_USER_TERMINATED,
     SessionCreate,
+    SessionNotFound,
+    SessionReuseDetected,
     SessionService,
     get_session_service,
 )
@@ -253,7 +255,7 @@ def _is_legacy_client(request: Request) -> bool:
         if client_v is None or cutoff_v is None:
             return False
         return client_v < cutoff_v
-    except Exception:  # noqa: BLE001
+    except (ValueError, TypeError, AttributeError):  # noqa: BLE001
         logger.debug("Failed to parse X-Client-Version %r", raw)
         return False
 
@@ -492,7 +494,7 @@ async def refresh(payload: RefreshRequest, request: Request) -> TokenResponse:
     # 1. Verify signature + type
     try:
         refresh_payload = verify_token(payload.refresh_token, secret, expected_type="refresh")
-    except Exception as exc:  # noqa: BLE001
+    except (jwt.PyJWTError, ValueError, KeyError) as exc:  # noqa: BLE001
         raise HTTPException(status_code=401, detail="Invalid refresh token") from exc
 
     # 2-3. Validate the session (also handles reuse detection)
@@ -510,7 +512,7 @@ async def refresh(payload: RefreshRequest, request: Request) -> TokenResponse:
             new_refresh_token=new_pair.refresh_token,
             new_ttl_seconds=int(REFRESH_TOKEN_TTL.total_seconds()),
         )
-    except Exception as exc:  # noqa: BLE001 — SessionReuseDetected / SessionNotFound
+    except (SessionReuseDetected, SessionNotFound, ValueError, KeyError, RuntimeError) as exc:  # noqa: BLE001 — SessionReuseDetected / SessionNotFound
         logger.info("Refresh rejected: %s", exc)
         raise HTTPException(status_code=401, detail="Invalid refresh token") from exc
 

src/agentkit/server/routes/bitable.py

@@ -483,7 +483,7 @@ async def validate_formula(
         parse_formula(body.formula)
     except (FormulaParseError, FormulaSecurityError, UnknownFunctionError) as e:
         return {"valid": False, "error": str(e)}
-    except Exception as e:  # pragma: no cover — defensive
+    except (ValueError, TypeError, KeyError, AttributeError) as e:  # pragma: no cover — defensive
         return {"valid": False, "error": f"Unexpected error: {e}"}
     return {"valid": True}
 
@@ -750,7 +750,7 @@ async def upload_file(
                 f.write(chunk)
     except HTTPException:
         raise
-    except Exception as exc:
+    except (OSError, RuntimeError) as exc:
         file_path.unlink(missing_ok=True)
         logger.error(f"Failed to save uploaded bitable file: {exc}")
         raise HTTPException(status_code=500, detail="Failed to save file") from exc

src/agentkit/server/routes/channels.py

@@ -553,7 +553,7 @@ async def _invalidate_adapter_cache(channel_id: str) -> None:
     if old is not None:
         try:
             await old.close()
-        except Exception:  # noqa: BLE001 — 关闭异常不应阻塞配置变更
+        except (ConnectionError, RuntimeError, OSError, asyncio.TimeoutError):  # noqa: BLE001 — 关闭异常不应阻塞配置变更
             logger.debug("关闭旧适配器异常已忽略: channel_id=%s", channel_id)
 
 
@@ -562,7 +562,7 @@ async def close_all_adapters() -> None:
     for channel_id, adapter in list(_adapter_cache.items()):
         try:
             await adapter.close()
-        except Exception:  # noqa: BLE001
+        except (ConnectionError, RuntimeError, OSError, asyncio.TimeoutError):  # noqa: BLE001
             logger.debug("关闭适配器异常已忽略: channel_id=%s", channel_id)
     _adapter_cache.clear()
 
@@ -614,6 +614,8 @@ async def _process_inbound_message(app_state: Any, adapter: MessageAdapter, mess
                         model=routing.model or "default",
                     )
                     final_content = getattr(result, "content", "") or ""
+                except asyncio.CancelledError:
+                    raise
                 except Exception as exc:  # noqa: BLE001 — 回退路径需捕获全部异常
                     logger.warning("ReActEngine 执行失败，回退到 DIRECT_CHAT: %s", exc)
                     final_content = await _direct_chat(llm_gateway, routing)
@@ -628,6 +630,8 @@ async def _process_inbound_message(app_state: Any, adapter: MessageAdapter, mess
                 content=final_content,
             )
             await adapter.send_message(outgoing)
+        except asyncio.CancelledError:
+            raise
         except Exception as exc:  # noqa: BLE001 — webhook 必须保持响应能力
             logger.exception("处理入站消息失败: %s", exc)
 
@@ -703,7 +707,7 @@ async def channel_webhook(channel_id: str, request: Request) -> Any:
     except WeComURLVerification as e:
         # 企微 URL 验证 — 返回 XML 响应
         return Response(content=e.response_xml, media_type="application/xml")
-    except Exception as exc:  # noqa: BLE001 — 防止 receive_message 异常导致 500 触发平台重试风暴
+    except (ValueError, KeyError, RuntimeError, AttributeError, OSError) as exc:  # noqa: BLE001 — 防止 receive_message 异常导致 500 触发平台重试风暴
         logger.warning("receive_message 解析失败 channel=%s: %s", channel_id, exc)
         return {"code": 0, "msg": "invalid_payload"}
 

src/agentkit/server/routes/chat.py

@@ -108,7 +108,7 @@ class ChatConnectionManager:
         for ws, _ in conns:
             try:
                 await ws.send_json(message)
-            except Exception:
+            except (ConnectionError, RuntimeError, asyncio.TimeoutError):
                 stale.append(ws)
         for ws in stale:
             self.remove(session_id, ws)
@@ -295,6 +295,8 @@ async def _execute_board_meeting(
         await team.create_board(topic=routing_result.topic, expert_configs=expert_configs)
         orchestrator = BoardOrchestrator(team=team)
         result = await orchestrator.execute(routing_result.topic)
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Board meeting failed for session {session_id}: {e}", exc_info=True)
         await websocket.send_json(
@@ -302,7 +304,7 @@ async def _execute_board_meeting(
         )
         try:
             await team.dissolve()
-        except Exception:
+        except (RuntimeError, asyncio.TimeoutError, ConnectionError):
             pass
         return True
     finally:
@@ -348,7 +350,7 @@ async def _execute_board_meeting(
     # Dissolve the team to release expert agents
     try:
         await team.dissolve()
-    except Exception as e:
+    except (RuntimeError, asyncio.TimeoutError, ConnectionError) as e:
         logger.warning(f"Board team dissolve failed: {e}")
 
     return True
@@ -467,7 +469,7 @@ async def _execute_team_collab(
         # Always dissolve the team and remove handler to avoid leaks
         try:
             await team.dissolve()
-        except Exception as e:
+        except (RuntimeError, asyncio.TimeoutError, ConnectionError) as e:
             logger.warning(f"Team dissolve failed: {e}")
         # dissolve() already clears handlers via handoff_transport.close()
 
@@ -585,7 +587,7 @@ def _build_phase_engine(
         if phase_policy is None:
             # Empty config (no `plan_exec:` section) → use KTD5 defaults.
             phase_policy = default_policy()
-    except Exception as e:
+    except (ValueError, TypeError, KeyError) as e:
         logger.error(
             "PLAN_EXEC phase policy construction failed for session %s: %s",
             session_id,
@@ -695,6 +697,8 @@ async def send_message(session_id: str, request: SendMessageRequest, req: Reques
                     agent_name=agent.name,
                     system_prompt=system_prompt,
                 )
+            except asyncio.CancelledError:
+                raise
             except Exception as e:
                 logger.error(f"PLAN_EXEC execution error for session {session_id}: {e}")
                 raise HTTPException(status_code=500, detail=str(e))
@@ -773,6 +777,8 @@ async def send_message(session_id: str, request: SendMessageRequest, req: Reques
             return response_dict
         return response
 
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Chat execution error for session {session_id}: {e}")
         raise HTTPException(status_code=500, detail=str(e))
@@ -829,7 +835,7 @@ async def _resolve_ws_dept_context(
         db_path_resolved = Path(db_path)
         try:
             department_ids = await _fetch_user_department_ids(db_path_resolved, user_id)
-        except Exception:
+        except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
             logger.exception(
                 "Failed to fetch department ids for WebSocket user %s — fail-closed",
                 user_id,
@@ -946,7 +952,7 @@ async def chat_websocket(websocket: WebSocket, session_id: str) -> None:
                                 "data": {"content": content},
                             }
                         )
-                    except Exception as e:
+                    except (asyncio.QueueFull, RuntimeError, ConnectionError) as e:
                         logger.warning(f"Failed to enqueue intervention: {e}")
                         await websocket.send_json(
                             {
@@ -1022,11 +1028,13 @@ async def chat_websocket(websocket: WebSocket, session_id: str) -> None:
 
     except WebSocketDisconnect:
         logger.debug(f"Chat WebSocket disconnected for session {session_id}")
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Chat WebSocket error for session {session_id}: {e}")
         try:
             await websocket.send_json({"type": "error", "data": {"message": str(e)}})
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             pass
     finally:
         # Clean up pending futures
@@ -1174,6 +1182,8 @@ async def _handle_chat_message(
                 content=final_content,
                 agent_name=agent.name,
             )
+        except asyncio.CancelledError:
+            raise
         except Exception as e:
             # Check if this is a QuotaExceededError (U4: WebSocket quota).
             from agentkit.llm.gateway import QuotaExceededError
@@ -1422,6 +1432,8 @@ async def _handle_chat_message(
                 agent_name=agent.name,
             )
 
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Chat execution error for session {session_id}: {e}")
         # Show meaningful error to user, but avoid leaking full stack traces
@@ -1473,6 +1485,8 @@ async def upload_chat_file(file: UploadFile = File(...)) -> dict[str, Any]:
         file_path.write_bytes(contents)
     except HTTPException:
         raise
+    except asyncio.CancelledError:
+        raise
     except Exception as exc:
         logger.error(f"Failed to save uploaded file: {exc}")
         raise HTTPException(status_code=500, detail="Failed to save file") from exc

src/agentkit/server/routes/config_sync.py

@@ -64,7 +64,7 @@ def _collect_skill_configs(request: Request) -> list[dict[str, Any]]:
                 "version": getattr(skill.config, "version", "1.0.0"),
                 "config": config_dict,
             })
-    except Exception as e:
+    except (AttributeError, ValueError, KeyError, RuntimeError) as e:
         logger.warning(f"Failed to collect skill configs: {e}")
     return configs
 
@@ -81,7 +81,7 @@ def _collect_workflow_configs(request: Request) -> list[dict[str, Any]]:
             from agentkit.server.routes.workflows import _workflow_store
 
             workflow_store = _workflow_store
-        except Exception:
+        except (ImportError, AttributeError):
             return []
 
     configs: list[dict[str, Any]] = []
@@ -94,7 +94,7 @@ def _collect_workflow_configs(request: Request) -> list[dict[str, Any]]:
                 configs.append(wf.to_dict())
             else:
                 configs.append(dict(wf))
-    except Exception as e:
+    except (RuntimeError, AttributeError, ValueError) as e:
         logger.warning(f"Failed to collect workflow configs: {e}")
     return configs
 

src/agentkit/server/routes/documents.py

@@ -13,6 +13,7 @@ Endpoints:
 
 from __future__ import annotations
 
+import asyncio
 import hmac
 import logging
 import uuid
@@ -155,6 +156,8 @@ async def create_document(
         raise HTTPException(status_code=400, detail=str(e)) from e
     except FileNotFoundError as e:
         raise HTTPException(status_code=404, detail=str(e)) from e
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Document creation failed: {e}")
         raise HTTPException(status_code=500, detail="Document creation failed") from e
@@ -187,7 +190,7 @@ async def upload_template(
         file_path.write_bytes(contents)
     except HTTPException:
         raise
-    except Exception as exc:
+    except (OSError, RuntimeError) as exc:
         logger.error(f"Failed to save template: {exc}")
         raise HTTPException(status_code=500, detail="Failed to save template") from exc
     finally:

src/agentkit/server/routes/evolution.py

@@ -1,5 +1,6 @@
 """Evolution API routes"""
 
+import asyncio
 import logging
 
 from fastapi import APIRouter, HTTPException, Request
@@ -42,6 +43,8 @@ async def list_evolution_events(
             agent_name=agent_name,
             change_type=event_type,
         )
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Failed to list evolution events: {e}")
         raise HTTPException(status_code=500, detail="Failed to list evolution events")
@@ -63,6 +66,8 @@ async def get_skill_versions(skill_name: str, req: Request = None):
     store = _get_evolution_store(req)
     try:
         versions = await store.list_skill_versions(skill_name)
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Failed to get skill versions for '{skill_name}': {e}")
         raise HTTPException(status_code=500, detail="Failed to get skill versions")
@@ -103,6 +108,8 @@ async def trigger_evolution(request: TriggerEvolutionRequest, req: Request = Non
     )
     try:
         event_id = await store.record(event)
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Failed to record trigger event: {e}")
         raise HTTPException(status_code=500, detail="Failed to trigger evolution")
@@ -153,6 +160,8 @@ async def list_ab_tests(
                     for e in entries
                 ]
                 return {"items": results[:limit], "total": len(results)}
+        except asyncio.CancelledError:
+            raise
         except Exception as e:
             logger.error(f"Failed to list A/B tests from persistent store: {e}")
             raise HTTPException(status_code=500, detail="Failed to list A/B tests")

src/agentkit/server/routes/evolution_dashboard.py

@@ -194,7 +194,7 @@ async def list_experiences(
                     }
                 )
             return {"experiences": experiences, "total": len(experiences)}
-        except Exception as e:
+        except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError) as e:
             logger.error(f"Failed to list experiences from store: {e}")
 
     # Fallback to in-memory store
@@ -324,7 +324,7 @@ async def get_metrics(
 
                 # Generate daily trends from the metrics
                 trends = _generate_trends(metrics_list, period)
-        except Exception as e:
+        except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError) as e:
             logger.error(f"Failed to get metrics from store: {e}")
     else:
         # Generate from in-memory experiences
@@ -501,7 +501,7 @@ async def get_usage(
                     "errors": 0,
                     "avg_latency_ms": round(d["latency"] / max(d["requests"], 1), 1),
                 })
-        except Exception as e:
+        except (ConnectionError, OSError, ValueError, KeyError, RuntimeError, AttributeError) as e:
             logger.error(f"Failed to get usage from LLMGateway: {e}")
 
     # Fill in missing dates with zero
@@ -587,7 +587,7 @@ async def check_pitfalls(
             })
 
         return {"warnings": warnings_data}
-    except Exception as e:
+    except (RuntimeError, ValueError, KeyError, AttributeError, asyncio.TimeoutError, ConnectionError) as e:
         logger.error(f"Failed to check pitfalls: {e}")
         return {"warnings": []}
 
@@ -642,7 +642,7 @@ async def list_path_optimizations(
                             else None,
                         }
                     )
-        except Exception as e:
+        except (ValueError, KeyError, RuntimeError, AttributeError) as e:
             logger.error(f"Failed to get path optimizations: {e}")
 
     # Also include in-memory optimizations
@@ -767,7 +767,7 @@ async def evolution_dashboard_ws(websocket: WebSocket):
                 )
     except WebSocketDisconnect:
         logger.debug("Evolution dashboard WebSocket disconnected")
-    except Exception as e:
+    except (RuntimeError, asyncio.TimeoutError, ConnectionError) as e:
         logger.error(f"Evolution dashboard WebSocket error: {e}")
     finally:
         if websocket in _ws_connections:
@@ -781,7 +781,7 @@ async def _broadcast_event(event_type: str, data: dict):
     for ws in _ws_connections:
         try:
             await ws.send_json(message)
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             disconnected.append(ws)
     for ws in disconnected:
         if ws in _ws_connections:

src/agentkit/server/routes/health.py

@@ -1,5 +1,7 @@
 """Health check route"""
 
+import asyncio
+
 from fastapi import APIRouter, Request
 
 router = APIRouter(tags=["health"])
@@ -23,14 +25,14 @@ async def health_check(request: Request):
                     redis_client = await task_store._get_redis()
                     await redis_client.ping()
                     redis_status = "available"
-                except Exception as ping_exc:
+                except (ConnectionError, OSError, asyncio.TimeoutError, RuntimeError) as ping_exc:
                     redis_status = f"error: {str(ping_exc)[:100]}"
                     overall_status = "degraded"
             else:
                 redis_status = "not_configured"
         else:
             redis_status = "not_configured"
-    except Exception as exc:
+    except (ConnectionError, OSError, asyncio.TimeoutError, RuntimeError, AttributeError) as exc:
         redis_status = f"error: {str(exc)[:100]}"
         overall_status = "degraded"
     checks["redis"] = redis_status
@@ -42,7 +44,7 @@ async def health_check(request: Request):
         try:
             agents = agent_pool.list_agents()
             pool_size = len(agents)
-        except Exception:
+        except (RuntimeError, AttributeError):
             pass
     checks["agent_pool"] = {"status": "available", "size": pool_size}
 
@@ -57,7 +59,7 @@ async def health_check(request: Request):
             else:
                 llm_status = "no_providers"
                 overall_status = "degraded"
-        except Exception:
+        except (RuntimeError, AttributeError, ValueError):
             llm_status = "error"
             overall_status = "degraded"
     checks["llm_gateway"] = llm_status
@@ -68,7 +70,7 @@ async def health_check(request: Request):
     if skill_registry:
         try:
             skill_count = len(skill_registry.list_skills())
-        except Exception:
+        except (RuntimeError, AttributeError):
             pass
     checks["skill_registry"] = {
         "status": "available" if skill_registry else "not_configured",

src/agentkit/server/routes/kb_management.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import asyncio
 import hmac
 import logging
 import os
@@ -252,7 +253,7 @@ async def list_sources(
         visible_ids = await filter_kb_sources_by_department(
             db_path, dept_ctx.department_ids, all_ids
         )
-    except Exception:  # noqa: BLE001 — never block listing on DB errors
+    except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):  # noqa: BLE001 — never block listing on DB errors
         logger.exception("Department KB filtering failed — returning empty list")
         return {"sources": []}
     visible_set = set(visible_ids)
@@ -398,7 +399,7 @@ async def list_documents(
         visible_ids = await filter_kb_sources_by_department(
             db_path, dept_ctx.department_ids, all_source_ids
         )
-    except Exception:  # noqa: BLE001 — never block listing on DB errors
+    except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):  # noqa: BLE001 — never block listing on DB errors
         logger.exception("Department KB filtering failed — returning empty list")
         return {"documents": []}
     visible_set = set(visible_ids)
@@ -484,7 +485,7 @@ async def upload_document(
         try:
             text = processor.parse(tmp_path, file_type)
             chunks = processor.segment(text)
-        except Exception as e:
+        except (ValueError, OSError, RuntimeError, UnicodeDecodeError) as e:
             logger.warning("Document parsing failed: %s", e)
             raise HTTPException(status_code=422, detail=f"Document parsing failed: {e}") from e
 
@@ -567,7 +568,7 @@ async def preview_document(
                 chunk_size=chunk_size,
                 chunk_overlap=chunk_overlap,
             )
-        except Exception as e:
+        except (ValueError, OSError, RuntimeError, UnicodeDecodeError) as e:
             logger.warning("Document preview failed: %s", e)
             raise HTTPException(status_code=422, detail=f"Document preview failed: {e}") from e
 
@@ -628,7 +629,7 @@ async def search_knowledge(
                     for r in results
                 ]
             }
-        except Exception as e:
+        except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError) as e:
             logger.warning(f"Semantic search failed: {e}")
 
     # Fallback: return empty results with a hint

src/agentkit/server/routes/mcp_publish.py

@@ -115,7 +115,7 @@ async def publish_skill(
 
     try:
         skill = skill_registry.get(skill_name)
-    except Exception:
+    except (KeyError, ValueError, AttributeError):
         raise HTTPException(status_code=404, detail=f"Skill '{skill_name}' not found")
 
     try:

src/agentkit/server/routes/memory.py

@@ -1,5 +1,6 @@
 """Memory API routes"""
 
+import asyncio
 import logging
 
 from fastapi import APIRouter, HTTPException, Request
@@ -40,6 +41,8 @@ async def search_episodic_memory(
         if agent_name:
             filters["agent_name"] = agent_name
         items = await retriever._episodic.search(query, top_k=top_k, filters=filters or None)
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Failed to search episodic memory: {e}")
         raise HTTPException(status_code=500, detail="Failed to search episodic memory")
@@ -76,6 +79,8 @@ async def search_semantic_memory(
         if knowledge_base_ids:
             filters["knowledge_base_ids"] = [kid.strip() for kid in knowledge_base_ids.split(",")]
         items = await retriever._semantic.search(query, top_k=top_k, filters=filters or None)
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Failed to search semantic memory: {e}")
         raise HTTPException(status_code=500, detail="Failed to search semantic memory")
@@ -104,6 +109,8 @@ async def delete_episodic_memory(key: str, req: Request = None):
 
     try:
         deleted = await retriever._episodic.delete(key)
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Failed to delete episodic memory '{key}': {e}")
         raise HTTPException(status_code=500, detail="Failed to delete episodic memory")

src/agentkit/server/routes/metrics.py

@@ -1,5 +1,6 @@
 """Metrics route — /api/v1/metrics"""
 
+import asyncio
 import logging
 
 from fastapi import APIRouter, Request
@@ -29,7 +30,7 @@ async def get_metrics(request: Request):
             task_metrics["completed_tasks"] = counts.get("completed", 0)
             task_metrics["failed_tasks"] = counts.get("failed", 0)
             task_metrics["pending_tasks"] = counts.get("pending", 0)
-        except Exception as e:
+        except (RuntimeError, AttributeError, ConnectionError, asyncio.TimeoutError) as e:
             logger.warning(f"Failed to collect task metrics: {e}")
 
     # Agent pool metrics
@@ -41,7 +42,7 @@ async def get_metrics(request: Request):
         try:
             agents = agent_pool.list_agents()
             agent_metrics["total_agents"] = len(agents)
-        except Exception as e:
+        except (RuntimeError, AttributeError) as e:
             logger.warning(f"Failed to collect agent metrics: {e}")
 
     # Skill registry metrics
@@ -53,7 +54,7 @@ async def get_metrics(request: Request):
         try:
             skills = skill_registry.list_skills()
             skill_metrics["total_skills"] = len(skills)
-        except Exception as e:
+        except (RuntimeError, AttributeError) as e:
             logger.warning(f"Failed to collect skill metrics: {e}")
 
     return {

src/agentkit/server/routes/portal.py

@@ -94,7 +94,7 @@ async def _emit_event_safe(
             data=data or {},
         )
         await event_queue.emit(event)
-    except Exception as e:
+    except (asyncio.QueueFull, RuntimeError, ConnectionError) as e:
         logger.warning(f"EventQueue emit failed (type={event_type}): {e}", exc_info=True)
 
 
@@ -211,7 +211,7 @@ class PortalConnectionManager:
                 import asyncio
 
                 asyncio.create_task(oldest.close(code=1008, reason="Connection limit exceeded"))
-            except Exception:
+            except (ConnectionError, RuntimeError):
                 pass
         conns.append(ws)
 
@@ -235,7 +235,7 @@ class PortalConnectionManager:
         for ws in conns:
             try:
                 await ws.send_json(message)
-            except Exception as e:
+            except (ConnectionError, RuntimeError, asyncio.TimeoutError) as e:
                 logger.debug(
                     "Portal WS send failed for user %s (marking stale): %s", user_id, e
                 )
@@ -285,7 +285,7 @@ async def _build_history_messages(
     """
     try:
         history = await _conversation_store.get_history(conv_id, limit=limit)
-    except Exception:
+    except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
         return []
 
     # The last message in history is the current user message (just added),
@@ -553,6 +553,8 @@ async def chat(request: ChatRequest, req: Request, _auth: None = Depends(_verify
             ):
                 if event.event_type == "final_answer":
                     collected_output.append(event.data.get("output", ""))
+        except asyncio.CancelledError:
+            raise
         except Exception as e:
             response_text = f"执行出错: {e}"
         else:
@@ -682,6 +684,8 @@ async def chat_stream(request: ChatRequest, req: Request, _auth: None = Depends(
                             }
                         ),
                     }
+            except asyncio.CancelledError:
+                raise
             except Exception as e:
                 yield {
                     "event": "error",
@@ -862,7 +866,7 @@ async def _execute_react_background(
                 await _task_store_update_status(
                     task_store, task_id, TaskStatus.RUNNING, started_at=datetime.now(timezone.utc)
                 )
-            except Exception:
+            except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                 logger.warning("Failed to update TaskStore RUNNING", exc_info=True)
 
         async for event in react_engine.execute_stream(
@@ -909,7 +913,7 @@ async def _execute_react_background(
                     progress=1.0,
                     progress_message="Completed",
                 )
-            except Exception:
+            except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                 logger.warning("Failed to update TaskStore COMPLETED", exc_info=True)
 
         # Emit task.completed so subscribers know the task is done
@@ -932,7 +936,15 @@ async def _execute_react_background(
             partial = _ensure_non_empty("".join(collected_output))
             try:
                 await asyncio.shield(conversation_store.add_message(conv_id, "assistant", partial))
-            except (Exception, asyncio.CancelledError):
+            except (
+                asyncio.CancelledError,
+                ConnectionError,
+                OSError,
+                asyncio.TimeoutError,
+                ValueError,
+                KeyError,
+                RuntimeError,
+            ):
                 logger.warning("Failed to persist partial output on cancel")
         if task_store is not None:
             try:
@@ -945,7 +957,15 @@ async def _execute_react_background(
                         completed_at=datetime.now(timezone.utc),
                     )
                 )
-            except (Exception, asyncio.CancelledError):
+            except (
+                asyncio.CancelledError,
+                ConnectionError,
+                OSError,
+                asyncio.TimeoutError,
+                ValueError,
+                KeyError,
+                RuntimeError,
+            ):
                 logger.warning("Failed to update TaskStore on cancel", exc_info=True)
         # P0 #2 fix: _emit_event_safe is async (it awaits event_queue.emit).
         # Shield it so a re-entrant CancelledError doesn't kill the emit
@@ -963,7 +983,7 @@ async def _execute_react_background(
                     },
                 )
             )
-        except (Exception, asyncio.CancelledError):
+        except (asyncio.CancelledError, asyncio.QueueFull, RuntimeError, ConnectionError):
             logger.warning("Failed to emit TASK_FAILED on cancel")
         raise  # Propagate cancellation
 
@@ -973,7 +993,7 @@ async def _execute_react_background(
             partial = _ensure_non_empty("".join(collected_output))
             try:
                 await conversation_store.add_message(conv_id, "assistant", partial)
-            except Exception:
+            except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                 logger.warning("Failed to persist partial output in background task")
 
         if task_store is not None:
@@ -985,7 +1005,7 @@ async def _execute_react_background(
                     error_message=str(e),
                     completed_at=datetime.now(timezone.utc),
                 )
-            except Exception:
+            except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                 logger.warning("Failed to update TaskStore FAILED", exc_info=True)
 
         # Emit task.failed so subscribers know the task failed
@@ -1120,7 +1140,7 @@ async def portal_websocket(websocket: WebSocket):
 
                 try:
                     record = await _task_store_get(resume_task_store, resume_task_id)
-                except Exception:
+                except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                     logger.warning("TaskStore.get failed during resume", exc_info=True)
                     record = None
                 if record is not None:
@@ -1333,7 +1353,7 @@ async def portal_websocket(websocket: WebSocket):
                         },
                     )
                     await _broadcast_dashboard_event("metrics_updated", {"period": "7d"})
-                except Exception as e:
+                except (asyncio.QueueFull, RuntimeError, ConnectionError, ValueError, KeyError) as e:
                     logger.warning(f"Failed to record experience: {e}")
 
             # Unified preprocessing via RequestPreprocessor (minimal: @skill prefix + greeting regex + REACT)
@@ -1414,7 +1434,7 @@ async def portal_websocket(websocket: WebSocket):
                         TaskStatus.PENDING,
                         metadata={"conversation_id": conv.id},
                     )
-                except Exception:
+                except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                     logger.warning("Failed to register task in TaskStore", exc_info=True)
 
             # Execute based on routing result's execution_mode
@@ -1455,7 +1475,7 @@ async def portal_websocket(websocket: WebSocket):
                             progress=1.0,
                             progress_message="Completed",
                         )
-                    except Exception:
+                    except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                         logger.warning("Failed to update TaskStore for DIRECT_CHAT", exc_info=True)
 
                 # Emit turn.final_answer and task.completed to EQ
@@ -1526,7 +1546,7 @@ async def portal_websocket(websocket: WebSocket):
                             chat_messages.insert(
                                 -1, {"role": hist_msg.role, "content": hist_msg.content}
                             )
-                except Exception:
+                except (ConnectionError, OSError, asyncio.TimeoutError, ValueError, KeyError, RuntimeError):
                     pass
                 response = await llm_gateway.chat(
                     messages=chat_messages,
@@ -1627,7 +1647,7 @@ async def portal_websocket(websocket: WebSocket):
                 logger.warning("EventQueue not configured; awaiting background task directly")
                 try:
                     await bg_task
-                except Exception:
+                except (RuntimeError, ConnectionError, asyncio.TimeoutError):
                     pass  # errors handled inside _execute_react_background
                 active_bg_task = None
                 continue
@@ -1734,6 +1754,8 @@ async def portal_websocket(websocket: WebSocket):
         # kill the task, lose the full output, and mark it FAILED —
         # defeating layers 2 and 3. The task is only cancelled on explicit
         # user cancel (msg_type == 'cancel') or application shutdown.
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Portal WebSocket error: {e}")
         # P1 #6 fix: Do NOT cancel the background task on connection-level
@@ -1758,7 +1780,7 @@ async def portal_websocket(websocket: WebSocket):
             )
         try:
             await websocket.send_json({"type": "error", "data": {"message": str(e)}})
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             pass
     finally:
         # Remove from user-scoped push tracking on any disconnect/error/return.

src/agentkit/server/routes/skill_management.py

@@ -119,7 +119,7 @@ def _skill_to_detail(skill: Any) -> dict[str, Any]:
     if hasattr(skill, "config"):
         try:
             config = skill.config.to_dict() if hasattr(skill.config, "to_dict") else {}
-        except Exception:
+        except (AttributeError, ValueError, TypeError):
             config = {}
 
     return {
@@ -174,7 +174,7 @@ async def get_skill_detail(skill_name: str, req: Request):
     skill_registry = req.app.state.skill_registry
     try:
         skill = skill_registry.get(skill_name)
-    except Exception:
+    except (KeyError, ValueError, AttributeError):
         raise HTTPException(status_code=404, detail=f"Skill '{skill_name}' not found")
 
     return _skill_to_detail(skill)
@@ -186,7 +186,7 @@ async def check_skill_health(skill_name: str, req: Request):
     skill_registry = req.app.state.skill_registry
     try:
         skill_registry.get(skill_name)
-    except Exception:
+    except (KeyError, ValueError, AttributeError):
         raise HTTPException(status_code=404, detail=f"Skill '{skill_name}' not found")
 
     # Basic health check - skill exists and is registered
@@ -243,7 +243,7 @@ async def reload_skill(skill_name: str, req: Request):
     # Verify the skill is currently registered (404 if not).
     try:
         skill_registry.get(skill_name)
-    except Exception:
+    except (KeyError, ValueError, AttributeError):
         raise HTTPException(status_code=404, detail=f"Skill '{skill_name}' not found")
 
     # Resolve the skills directory (mirrors routes.skills._get_skills_dir).

src/agentkit/server/routes/skills.py

@@ -136,7 +136,7 @@ async def register_skill(request: RegisterSkillRequest, req: Request):
 
     try:
         config = SkillConfig.from_dict(request.config)
-    except Exception as e:
+    except (ValueError, TypeError, KeyError) as e:
         raise HTTPException(status_code=422, detail=f"Invalid skill config: {e}")
 
     skill = Skill(config=config)
@@ -279,7 +279,7 @@ async def install_skill(request: InstallSkillRequest, req: Request):
                 resp = await client.get(source)
                 resp.raise_for_status()
                 yaml_content = resp.text
-        except Exception as e:
+        except (httpx.HTTPError, OSError) as e:
             raise HTTPException(status_code=400, detail=f"Failed to download from source: {e}")
     elif source and source.startswith("file://"):
         # Read from local file path
@@ -295,7 +295,7 @@ async def install_skill(request: InstallSkillRequest, req: Request):
         try:
             with open(local_path, encoding="utf-8") as f:
                 yaml_content = f.read()
-        except Exception as e:
+        except OSError as e:
             raise HTTPException(status_code=400, detail=f"Failed to read local file: {e}")
     else:
         # Search GitHub for skills (YAML config files)
@@ -313,7 +313,7 @@ async def install_skill(request: InstallSkillRequest, req: Request):
                     },
                 )
                 gh_data = gh_resp.json()
-        except Exception as e:
+        except (httpx.HTTPError, OSError, ValueError) as e:
             raise HTTPException(status_code=502, detail=f"GitHub search failed: {e}")
 
         items = gh_data.get("items", [])
@@ -334,7 +334,7 @@ async def install_skill(request: InstallSkillRequest, req: Request):
                         },
                     )
                     items = gh_resp2.json().get("items", [])
-            except Exception:
+            except (httpx.HTTPError, OSError, ValueError, KeyError):
                 items = []
 
         if not items:
@@ -362,7 +362,7 @@ async def install_skill(request: InstallSkillRequest, req: Request):
                 resp = await client.get(raw_url)
                 resp.raise_for_status()
                 yaml_content = resp.text
-        except Exception as e:
+        except (httpx.HTTPError, OSError) as e:
             raise HTTPException(status_code=400, detail=f"Failed to download skill: {e}")
 
     # Validate YAML content before writing to disk
@@ -391,14 +391,14 @@ async def install_skill(request: InstallSkillRequest, req: Request):
         )
         loader.load_from_file(file_path)
         registration_ok = True
-    except Exception as e:
+    except (ValueError, TypeError, KeyError, OSError, RuntimeError) as e:
         logger.warning(f"Failed to register installed skill: {e}")
 
     if not registration_ok:
         # Remove the invalid YAML file and report error
         try:
             os.remove(file_path)
-        except Exception:
+        except OSError:
             pass
         raise HTTPException(status_code=500, detail="Skill downloaded but registration failed")
 
@@ -419,7 +419,7 @@ async def uninstall_skill(name: str, req: Request):
 
     try:
         skill_registry.get(validated_name)
-    except Exception:
+    except (KeyError, ValueError, RuntimeError):
         raise HTTPException(status_code=404, detail=f"Skill '{name}' not found")
 
     # Remove from registry
@@ -487,7 +487,7 @@ async def execute_pipeline(name: str, request: ExecutePipelineRequest, req: Requ
 
     try:
         result = await pipeline.execute(input_data=request.input_data)
-    except Exception as e:
+    except (ValueError, TypeError, KeyError, RuntimeError, OSError, ConnectionError) as e:
         logger.error(f"Pipeline execution failed for '{name}': {e}", exc_info=True)
         raise HTTPException(status_code=500, detail="Pipeline execution failed")
 

src/agentkit/server/routes/system.py

@@ -24,7 +24,7 @@ def _read_meminfo() -> dict[str, int]:
                     key = parts[0].strip()
                     value = parts[1].strip().split()[0]
                     values[key] = int(value) * 1024  # kB -> bytes
-    except Exception as exc:
+    except (OSError, ValueError, FileNotFoundError) as exc:
         logger.debug(f"Failed to read /proc/meminfo: {exc}")
     return values
 
@@ -52,7 +52,7 @@ async def get_system_resources() -> dict[str, Any]:
     if hasattr(os, "getloadavg"):
         try:
             loadavg = list(os.getloadavg())
-        except Exception as exc:
+        except (OSError, AttributeError) as exc:
             logger.debug(f"Failed to get loadavg: {exc}")
 
     meminfo = _read_meminfo()
@@ -68,7 +68,7 @@ async def get_system_resources() -> dict[str, Any]:
         disk_total = du.total
         disk_used = du.used
         disk_free = du.free
-    except Exception as exc:
+    except (OSError, FileNotFoundError) as exc:
         logger.debug(f"Failed to get disk usage: {exc}")
 
     return {

src/agentkit/server/routes/tasks.py

@@ -83,7 +83,7 @@ async def submit_task(request: SubmitTaskRequest, req: Request):
     elif request.skill_name:
         try:
             skill = skill_registry.get(request.skill_name)
-        except Exception:
+        except (KeyError, ValueError, AttributeError):
             raise HTTPException(
                 status_code=404,
                 detail=f"Skill '{request.skill_name}' not found",
@@ -145,7 +145,7 @@ async def submit_task(request: SubmitTaskRequest, req: Request):
             quality_result = await quality_gate.validate(
                 task_result.output_data or {}, skill, skill_context=skill_context
             )
-        except Exception:
+        except (RuntimeError, ValueError, KeyError, AttributeError, asyncio.TimeoutError):
             pass  # Quality gate failure shouldn't block the response
 
     # 7. Standardize output if skill available
@@ -167,7 +167,7 @@ async def submit_task(request: SubmitTaskRequest, req: Request):
                 "task_id": task.task_id,
                 "status": task_result.status,
             }
-        except Exception:
+        except (ValueError, KeyError, AttributeError, RuntimeError):
             pass  # Fall through to raw output
 
     # 8. Return raw result if no skill or standardization failed
@@ -307,7 +307,7 @@ async def resume_task(task_id: str, req: Request, plan_id: str | None = None):
         finally:
             try:
                 await team.dissolve()
-            except Exception:
+            except (RuntimeError, asyncio.TimeoutError, ConnectionError):
                 pass
 
     return {
@@ -343,7 +343,7 @@ async def stream_task(request: SubmitTaskRequest, req: Request):
     elif request.skill_name:
         try:
             skill_registry.get(request.skill_name)
-        except Exception:
+        except (KeyError, ValueError, AttributeError):
             raise HTTPException(
                 status_code=404,
                 detail=f"Skill '{request.skill_name}' not found",

src/agentkit/server/routes/terminal.py

@@ -359,14 +359,14 @@ async def execute_command(
                 await check_pty.close()
                 if cwd_result.exit_code == 0:
                     state.cwd = cwd_result.output.strip()
-            except Exception:
+            except (OSError, RuntimeError, asyncio.TimeoutError):
                 pass
 
     except asyncio.TimeoutError:
         output = f"命令执行超时（{request.timeout}s）"
         exit_code = -1
         duration_ms = int((time.monotonic() - start_time) * 1000)
-    except Exception as e:
+    except (OSError, RuntimeError, ValueError) as e:
         output = str(e)
         exit_code = -1
         duration_ms = int((time.monotonic() - start_time) * 1000)
@@ -515,7 +515,7 @@ async def terminal_websocket(websocket: WebSocket) -> None:
                     })
                     await websocket.close(code=4003, reason="Permission denied")
                     return
-        except Exception:
+        except (ValueError, KeyError, RuntimeError, OSError):
             pass  # Fall through to API key / dev mode
 
     # 2. API key via ?api_key=
@@ -721,7 +721,7 @@ async def terminal_websocket(websocket: WebSocket) -> None:
                                 })
                         except asyncio.TimeoutError:
                             continue
-                        except Exception:
+                        except (OSError, RuntimeError):
                             break
 
                     # Get final result
@@ -741,7 +741,7 @@ async def terminal_websocket(websocket: WebSocket) -> None:
                             await check_pty.close()
                             if cwd_result.exit_code == 0:
                                 state.cwd = cwd_result.output.strip()
-                        except Exception:
+                        except (OSError, RuntimeError, asyncio.TimeoutError):
                             pass
 
                     # Record history
@@ -790,6 +790,8 @@ async def terminal_websocket(websocket: WebSocket) -> None:
                         "cwd": state.cwd,
                         "duration_ms": duration_ms,
                     })
+                except asyncio.CancelledError:
+                    raise
                 except Exception as e:
                     duration_ms = int((time.monotonic() - start_time) * 1000)
                     await websocket.send_json({
@@ -800,7 +802,7 @@ async def terminal_websocket(websocket: WebSocket) -> None:
                     if active_pty and active_pty.is_running:
                         try:
                             await active_pty.close()
-                        except Exception:
+                        except (OSError, RuntimeError):
                             pass
                         active_pty = None
 
@@ -836,11 +838,13 @@ async def terminal_websocket(websocket: WebSocket) -> None:
 
     except WebSocketDisconnect:
         logger.debug(f"Terminal WebSocket disconnected for session {state.session_id}")
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Terminal WebSocket error for session {state.session_id}: {e}")
         try:
             await websocket.send_json({"type": "error", "message": str(e)[:200]})
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             pass
     finally:
         # Clean up pending confirmations
@@ -851,5 +855,5 @@ async def terminal_websocket(websocket: WebSocket) -> None:
         if active_pty and active_pty.is_running:
             try:
                 await active_pty.close()
-            except Exception:
+            except (OSError, RuntimeError):
                 pass

src/agentkit/server/routes/terminal_server.py

@@ -549,7 +549,7 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
                     })
                     await websocket.close(code=4003, reason="Permission denied")
                     return
-        except Exception:
+        except (ValueError, KeyError, RuntimeError, OSError):
             pass
 
     if not auth_ok:
@@ -585,7 +585,7 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
                         })
                         await websocket.close(code=4003, reason="Not authorized")
                         return
-            except Exception as e:
+            except (aiosqlite.Error, OSError, ValueError, KeyError, RuntimeError) as e:
                 logger.warning(f"Failed to check server terminal authorization: {e}")
                 # If DB check fails, deny access
                 await websocket.send_json({
@@ -822,7 +822,7 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
                                 })
                         except asyncio.TimeoutError:
                             continue
-                        except Exception:
+                        except (OSError, RuntimeError):
                             break
 
                     result = await run_task
@@ -846,7 +846,7 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
                             else:
                                 base = state.cwd or os.path.expanduser("~")
                                 state.cwd = os.path.normpath(os.path.join(base, cd_arg))
-                        except Exception:
+                        except (ValueError, OSError, RuntimeError):
                             pass  # cd parsing failed — keep old cwd
 
                     # Record history
@@ -893,6 +893,8 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
                         "cwd": state.cwd,
                         "duration_ms": duration_ms,
                     })
+                except asyncio.CancelledError:
+                    raise
                 except Exception as e:
                     duration_ms = int((time.monotonic() - start_time) * 1000)
                     await websocket.send_json({
@@ -903,7 +905,7 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
                     if active_pty and active_pty.is_running:
                         try:
                             await active_pty.close()
-                        except Exception:
+                        except (OSError, RuntimeError):
                             pass
                         active_pty = None
 
@@ -942,17 +944,19 @@ async def server_terminal_websocket(websocket: WebSocket) -> None:
 
     except WebSocketDisconnect:
         logger.debug(f"Server terminal WebSocket disconnected for session {state.session_id}")
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"Server terminal WebSocket error for session {state.session_id}: {e}")
         try:
             await websocket.send_json({"type": "error", "message": str(e)[:200]})
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             pass
     finally:
         if active_pty and active_pty.is_running:
             try:
                 await active_pty.close()
-            except Exception:
+            except (OSError, RuntimeError):
                 pass
         # Clean up session from in-memory stores to prevent leaks
         _server_sessions.pop(state.session_id, None)

src/agentkit/server/routes/workflows.py

@@ -433,6 +433,8 @@ async def _execute_workflow(
                         )
                         result = await agent.handle_task(task)
                         stage_result = {"output": result, "skill": stage.action}
+                    except asyncio.CancelledError:
+                        raise
                     except Exception as e:
                         stage_result = {"error": str(e), "skill": stage.action}
                 else:
@@ -474,6 +476,8 @@ async def _execute_workflow(
                             )
                             result = await agent.handle_task(task)
                             return {"output": result, "skill": action}
+                        except asyncio.CancelledError:
+                            raise
                         except Exception as e:
                             return {"error": str(e), "skill": action}
                     return {"dry_run": True, "action": action}
@@ -515,6 +519,8 @@ async def _execute_workflow(
                 execution_id=execution.execution_id,
             )
 
+        except asyncio.CancelledError:
+            raise
         except Exception as e:
             execution.stage_results[stage_name] = {
                 "status": "failed",
@@ -633,7 +639,7 @@ async def _broadcast_ws(message: dict[str, Any], execution_id: str | None = None
     for ws in targets:
         try:
             await ws.send_json(message)
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             disconnected.append(ws)
     if disconnected:
         async with _ws_subscribers_lock:
@@ -952,7 +958,7 @@ async def workflow_websocket(websocket: WebSocket):
             # Keep connection alive - messages are primarily server-push
     except WebSocketDisconnect:
         logger.debug("Workflow WebSocket disconnected")
-    except Exception as e:
+    except (RuntimeError, asyncio.TimeoutError, ConnectionError) as e:
         logger.error(f"Workflow WebSocket error: {e}")
     finally:
         if subscribed_execution_id:

src/agentkit/server/routes/ws.py

@@ -47,7 +47,7 @@ class ConnectionManager:
         for ws, _ in conns:
             try:
                 await ws.send_json(message)
-            except Exception:
+            except (ConnectionError, RuntimeError, asyncio.TimeoutError):
                 stale.append(ws)
         for ws in stale:
             self.remove(task_id, ws)
@@ -153,6 +153,8 @@ async def task_websocket(websocket: WebSocket, task_id: str) -> None:
 
     except WebSocketDisconnect:
         logger.debug(f"WebSocket disconnected for task {task_id}")
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         logger.error(f"WebSocket error for task {task_id}: {e}")
         try:
@@ -162,7 +164,7 @@ async def task_websocket(websocket: WebSocket, task_id: str) -> None:
                     "data": {"message": str(e)},
                 }
             )
-        except Exception:
+        except (ConnectionError, RuntimeError, asyncio.TimeoutError):
             pass
     finally:
         manager.remove(task_id, websocket)
@@ -243,6 +245,8 @@ async def _run_react_and_stream(
                 },
             )
 
+    except asyncio.CancelledError:
+        raise
     except Exception as e:
         await websocket.send_json(
             {