Add system memory monitoring, Bedrock support, and Discord integrations

- Add CPU/GPU memory monitoring endpoint and frontend widget
- Add Amazon Bedrock model support (API key / inference profiles) plus model selector entries
- Add Discord weather/laundry notification job (JMA forecast + LLM-generated message)
- Add Discord bot that replies to mentions via the configured LLM, with weather-aware context
- Add scheduled "useless tips" job with separate webhook and quiet-hours skip
- Switch FastAPI startup to lifespan handler, log viewer, markdown rendering, and toast UI additions

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/api/chat.py

@@ -30,11 +30,29 @@ class SessionResponse(BaseModel):
         from_attributes = True
 
 
+class MessageResponse(BaseModel):
+    id: int
+    role: str
+    content: str
+    created_at: datetime
+
+    class Config:
+        from_attributes = True
+
+
 @router.get("/sessions", response_model=list[SessionResponse])
 def list_sessions(db: DBSession = Depends(get_db)):
     return db.query(Session).order_by(Session.updated_at.desc()).all()
 
 
+@router.get("/sessions/{session_id}/messages", response_model=list[MessageResponse])
+def get_messages(session_id: str, db: DBSession = Depends(get_db)):
+    session = db.get(Session, session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="Session not found")
+    return session.messages
+
+
 @router.delete("/sessions/{session_id}", status_code=204)
 def delete_session(session_id: str, db: DBSession = Depends(get_db)):
     session = db.get(Session, session_id)
@@ -68,24 +86,40 @@ async def stream_chat(req: ChatRequest, db: DBSession = Depends(get_db)):
     context = mem.get_context_messages(db, session.id)
 
     async def generate():
+        import json as _json
+
+        def sse(payload: dict) -> str:
+            return f"data: {_json.dumps(payload, ensure_ascii=False)}\n\n"
+
         # セッションIDをまず返す（新規セッション時にフロントへ通知）
-        yield f"data: {{\"type\":\"session_id\",\"value\":\"{session.id}\"}}\n\n"
+        yield sse({"type": "session_id", "value": session.id})
 
         full_response = []
-        async for chunk in llm.stream_chat(context, model=req.model or session.model):
-            full_response.append(chunk)
-            escaped = chunk.replace('"', '\\"').replace('\n', '\\n')
-            yield f"data: {{\"type\":\"chunk\",\"value\":\"{escaped}\"}}\n\n"
+        try:
+            async for chunk in llm.stream_chat(context, model=req.model or session.model, session_id=session.id):
+                full_response.append(chunk)
+                yield sse({"type": "chunk", "value": chunk})
 
-        # アシスタントメッセージを保存
-        assistant_content = "".join(full_response)
-        db.add(Message(session_id=session.id, role="assistant", content=assistant_content))
-        session.updated_at = datetime.now(timezone.utc)
-        db.commit()
+            # アシスタントメッセージを保存
+            assistant_content = "".join(full_response)
+            db.add(Message(session_id=session.id, role="assistant", content=assistant_content))
+            session.updated_at = datetime.now(timezone.utc)
+            db.commit()
 
-        # メモリ要約チェック
-        await mem.maybe_summarize(db, session)
+            # メモリ要約チェック
+            await mem.maybe_summarize(db, session)
 
-        yield "data: {\"type\":\"done\"}\n\n"
+            yield sse({"type": "done"})
 
-    return StreamingResponse(generate(), media_type="text/event-stream")
+        except Exception as e:
+            yield sse({"type": "error", "value": str(e)})
+
+    return StreamingResponse(
+        generate(),
+        media_type="text/event-stream",
+        headers={
+            "Cache-Control": "no-cache",
+            "X-Accel-Buffering": "no",
+            "Connection": "keep-alive",
+        },
+    )