Merge remote-tracking branch 'upstream/main' into main

Author: xiaowei

.env.example

@@ -229,6 +229,11 @@ AGENT_SKILLS=
 # 在飞书群 -> 设置 -> 群机器人 -> 添加机器人 -> 自定义机器人 -> 复制 Webhook 地址
 #
 # FEISHU_WEBHOOK_URL=https://open.feishu.cn/open-apis/bot/v2/hook/your_key_here
+# 如果机器人安全设置开启了“签名校验”，需同步填写 secret
+# FEISHU_WEBHOOK_SECRET=your_feishu_webhook_secret
+# 如果机器人安全设置开启了“关键词”，需填写同一个关键词；系统会自动在每条消息前补上
+# FEISHU_WEBHOOK_KEYWORD=股票日报
+# 注意：FEISHU_APP_ID / FEISHU_APP_SECRET 用于飞书应用、云文档或 Stream 模式，不会直接启用群 Webhook 推送
 #
 # 【方式三】Telegram 机器人（需同时配置两项）
 # 1. 在 Telegram 找 @BotFather -> /newbot -> 获取 Bot Token
@@ -369,14 +374,25 @@ DINGTALK_APP_SECRET=xxxx
 # 启用 Stream 模式
 DINGTALK_STREAM_ENABLED=false
 
-# 飞书应用机器人配置
+# 飞书应用机器人配置（用于 Stream Bot / 云文档模式，不会直接开启群 Webhook 推送）
 FEISHU_APP_ID=xxxx
-FEISHU_APP_SECRET=xxxx
+FEISHU_APP_SECRET=xxxx  # 飞书应用 Secret，仅应用/Stream Bot/云文档模式使用；群通知推送请使用 FEISHU_WEBHOOK_URL
 # 启用长连接模式
 FEISHU_STREAM_ENABLED=false
+# 飞书群机器人 Webhook 安全配置（仅 Webhook 推送模式使用）
+# FEISHU_WEBHOOK_SECRET=your_feishu_webhook_secret
+# FEISHU_WEBHOOK_KEYWORD=股票日报
 
 # 数据库路径
 DATABASE_PATH=./data/stock_analysis.db
+# SQLite 写入优化：文件库默认启用 WAL，降低批量写入和并发回写时的锁竞争
+SQLITE_WAL_ENABLED=true
+# SQLite 等锁超时（毫秒）
+SQLITE_BUSY_TIMEOUT_MS=5000
+# SQLite 遇到 database is locked / database table is locked 时的最大重试次数
+SQLITE_WRITE_RETRY_MAX=3
+# SQLite 写入重试基础退避时间（秒，按指数退避递增）
+SQLITE_WRITE_RETRY_BASE_DELAY=0.1
 
 # ===================================
 # 回测配置（可选）

api/v1/endpoints/agent.py

@@ -447,8 +447,13 @@ async def event_generator():
         finally:
             try:
                 await asyncio.wait_for(fut, timeout=5.0)
-            except Exception:
+            except asyncio.CancelledError:
                 pass
+            except asyncio.TimeoutError:
+                # Cleanup taking longer than 5s is treated as an expected timeout; no warning.
+                logger.debug("agent executor cleanup timed out after 5s for session %s", session_id)
+            except Exception as exc:
+                logger.warning("agent executor cleanup error (ignored): %s", exc, exc_info=True)
 
     return StreamingResponse(
         event_generator(),

api/v1/endpoints/analysis.py

@@ -347,11 +347,12 @@ def _handle_sync_analysis(
         )
 
         if result is None:
+            error_message = service.last_error or f"分析股票 {stock_code} 失败"
             raise HTTPException(
                 status_code=500,
                 detail={
                     "error": "analysis_failed",
-                    "message": f"分析股票 {stock_code} 失败"
+                    "message": error_message,
                 }
             )
 
@@ -482,6 +483,7 @@ async def task_stream():
     - connected: 连接成功
     - task_created: 新任务创建
     - task_started: 任务开始执行
+    - task_progress: 任务阶段进度更新
     - task_completed: 任务完成
     - task_failed: 任务失败
     - heartbeat: 心跳（每 30 秒）
@@ -516,8 +518,8 @@ async def event_generator():
                         "timestamp": datetime.now().isoformat()
                     })
         except asyncio.CancelledError:
-            # 客户端断开连接
-            pass
+            logger.debug("SSE client disconnected, cancelling event generator")
+            raise
         finally:
             task_queue.unsubscribe(event_queue)
 

api/v1/endpoints/system_config.py

@@ -11,6 +11,8 @@
 from api.deps import get_system_config_service
 from api.v1.schemas.common import ErrorResponse
 from api.v1.schemas.system_config import (
+    DiscoverLLMChannelModelsRequest,
+    DiscoverLLMChannelModelsResponse,
     ExportSystemConfigResponse,
     ImportSystemConfigRequest,
     SystemConfigConflictResponse,
@@ -313,6 +315,50 @@ def test_llm_channel(
         )
 
 
+@router.post(
+    "/config/llm/discover-models",
+    response_model=DiscoverLLMChannelModelsResponse,
+    responses={
+        200: {"description": "Model discovery completed"},
+        500: {"description": "Internal server error", "model": ErrorResponse},
+    },
+    summary="Discover models for one LLM channel",
+    description="Call one unsaved or saved channel's `/models` endpoint and return discovered model IDs.",
+)
+def discover_llm_channel_models(
+    request: DiscoverLLMChannelModelsRequest,
+    service: SystemConfigService = Depends(get_system_config_service),
+) -> DiscoverLLMChannelModelsResponse:
+    """Discover models for one channel definition without writing `.env`."""
+    try:
+        payload = service.discover_llm_channel_models(
+            name=request.name,
+            protocol=request.protocol,
+            base_url=request.base_url,
+            api_key=request.api_key,
+            models=request.models,
+            timeout_seconds=request.timeout_seconds,
+        )
+        return DiscoverLLMChannelModelsResponse.model_validate(payload)
+    except (ValueError, TypeError) as exc:
+        raise HTTPException(
+            status_code=422,
+            detail={
+                "error": "validation_error",
+                "message": str(exc),
+            },
+        )
+    except Exception as exc:
+        logger.error("Failed to discover LLM channel models: %s", exc, exc_info=True)
+        raise HTTPException(
+            status_code=500,
+            detail={
+                "error": "internal_error",
+                "message": "Failed to discover LLM channel models",
+            },
+        )
+
+
 @router.get(
     "/config/schema",
     response_model=SystemConfigSchemaResponse,

api/v1/schemas/system_config.py

@@ -162,6 +162,28 @@ class TestLLMChannelResponse(BaseModel):
     latency_ms: Optional[int] = None
 
 
+class DiscoverLLMChannelModelsRequest(BaseModel):
+    """Request payload for discovering models from one LLM channel."""
+
+    name: str = "channel"
+    protocol: str = "openai"
+    base_url: str = ""
+    api_key: str = ""
+    models: List[str] = Field(default_factory=list)
+    timeout_seconds: float = 20.0
+
+
+class DiscoverLLMChannelModelsResponse(BaseModel):
+    """Response payload for one LLM channel model discovery request."""
+
+    success: bool
+    message: str
+    error: Optional[str] = None
+    resolved_protocol: Optional[str] = None
+    models: List[str] = Field(default_factory=list)
+    latency_ms: Optional[int] = None
+
+
 class SystemConfigValidationErrorResponse(BaseModel):
     """Error payload for failed update validation."""