feat: 完成项目助手与监控演示并修复聊天及云部署配置

backend/api/chat.py

@@ -1,6 +1,7 @@
 """POST /api/chat — SSE streaming chat with Agent."""
 
 import json
+import asyncio
 import os
 import traceback
 from pathlib import Path
@@ -12,6 +13,7 @@ from sse_starlette.sse import EventSourceResponse
 
 from graph.agent import agent_manager
 from graph.session_manager import session_manager
+from graph.chat_model_options import chat_model_options
 
 router = APIRouter()
 
@@ -47,6 +49,7 @@ async def _generate_title(session_id: str) -> str | None:
             api_key=os.getenv("DEEPSEEK_API_KEY"),
             api_base=os.getenv("DEEPSEEK_BASE_URL", "https://api.deepseek.com"),
             temperature=0.3,
+            **chat_model_options(),
         )
 
         prompt = (
@@ -64,6 +67,27 @@ async def _generate_title(session_id: str) -> str | None:
         return None
 
 
+class EmptyChatResponse(RuntimeError):
+    pass
+
+
+async def bounded_events(iterator, idle_seconds=75, total_seconds=180):
+    """Bound stalls and total execution; never replay possible tool side effects."""
+    deadline = asyncio.get_running_loop().time() + total_seconds
+    try:
+        while True:
+            remaining = deadline - asyncio.get_running_loop().time()
+            if remaining <= 0:
+                raise TimeoutError()
+            try:
+                item = await asyncio.wait_for(anext(iterator), timeout=min(idle_seconds, remaining))
+            except StopAsyncIteration:
+                break
+            yield item
+    finally:
+        await iterator.aclose()
+
+
 async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict, None]:
     """Generate SSE events from agent stream.
 
@@ -72,6 +96,10 @@ async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict,
     emitted and a new segment begins. Each segment is saved as a
     separate assistant message in the session history.
     """
+    segments: list[dict] = []
+    current_segment: dict = {"content": "", "tool_calls": []}
+    saved = False
+    completed = False
     try:
         # Use merged history for agent context (combines consecutive assistant msgs)
         history = session_manager.load_session_for_agent(session_id)
@@ -81,7 +109,7 @@ async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict,
         segments: list[dict] = []  # finalized segments
         current_segment: dict = {"content": "", "tool_calls": []}
 
-        async for event in agent_manager.astream(message, history, session_id):
+        async for event in bounded_events(agent_manager.astream(message, history, session_id)):
             event_type = event.get("type", "unknown")
 
             if event_type == "retrieval":
@@ -137,6 +165,12 @@ async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict,
                 }
 
             elif event_type == "done":
+                if not str(event.get("content") or "").strip():
+                    raise EmptyChatResponse()
+                # Some providers deliver final content without token events.
+                if not current_segment["content"] and not any(s["content"] for s in segments):
+                    current_segment["content"] = event["content"]
+                    yield {"event":"token", "data":json.dumps({"content":event["content"]}, ensure_ascii=False)}
                 # Finalize the last segment
                 segments.append(current_segment)
 
@@ -145,10 +179,14 @@ async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict,
 
                 # Save each segment as a separate assistant message
                 for seg in segments:
+                    if not seg["content"].strip() and not seg["tool_calls"]:
+                        continue
                     tc = seg["tool_calls"] if seg["tool_calls"] else None
                     session_manager.save_message(
                         session_id, "assistant", seg["content"], tool_calls=tc
                     )
+                saved = True
+                completed = True
 
                 yield {
                     "event": "done",
@@ -160,7 +198,10 @@ async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict,
 
                 # Auto-generate title on first conversation
                 if is_first_message:
-                    title = await _generate_title(session_id)
+                    try:
+                        title = await asyncio.wait_for(_generate_title(session_id), timeout=8)
+                    except TimeoutError:
+                        title = None
                     if title:
                         yield {
                             "event": "title",
@@ -170,11 +211,26 @@ async def event_generator(message: str, session_id: str) -> AsyncGenerator[dict,
                             ),
                         }
 
+        if not completed:
+            raise ConnectionError("stream ended before done")
     except Exception as e:
-        traceback.print_exc()
+        is_timeout = isinstance(e, TimeoutError) or type(e).__name__ in ("APITimeoutError", "ReadTimeout", "ConnectTimeout")
+        code = "empty_response" if isinstance(e, EmptyChatResponse) else "timeout" if is_timeout else "chat_failed"
+        message_text = {
+            "empty_response":"模型未返回回答正文，可能只返回了推理字段。请检查聊天模型兼容配置后重试。",
+            "timeout":"模型响应超时，已停止本次等待。请检查服务后重试；不会自动重复执行工具。",
+            "chat_failed":"聊天未完成，请检查当前后端、模型连接或访问凭据后重试。已接收的内容会保留。",
+        }[code]
+        print(f"Chat failed: {type(e).__name__}")
+        if not saved:
+            session_manager.save_message(session_id, "user", message)
+            for seg in segments:
+                if seg["content"] or seg["tool_calls"]:
+                    session_manager.save_message(session_id, "assistant", seg["content"], tool_calls=seg["tool_calls"] or None)
+            session_manager.save_message(session_id, "assistant", current_segment["content"] + "\n\n【未完成】" + message_text, tool_calls=current_segment["tool_calls"] or None)
         yield {
             "event": "error",
-            "data": json.dumps({"error": str(e)}, ensure_ascii=False),
+            "data": json.dumps({"error": message_text, "code": code}, ensure_ascii=False),
         }