style: 简化文档卡片样式，优化布局间距

backend/main.py

@@ -1,6 +1,5 @@
 import asyncio
 import base64
-import json
 import logging
 import os
 import re
@@ -12,7 +11,7 @@ from typing import Optional
 
 from fastapi import FastAPI, HTTPException, Request, Security, File, UploadFile
 from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import JSONResponse, StreamingResponse, Response
+from fastapi.responses import JSONResponse, Response
 from fastapi.security import APIKeyHeader
 from pydantic import BaseModel
 
@@ -125,10 +124,6 @@ def _sanitize_converted_markdown(text: str) -> str:
     return value.strip()
 
 
-def _sse_payload(payload: dict) -> str:
-    return f"data: {json.dumps(payload)}\n\n"
-
-
 def get_client_ip(request: Request) -> str:
     if request.client:
         return request.headers.get("X-Client-IP") or request.client.host
@@ -186,11 +181,10 @@ async def create_completion(request: Request, req: CompletionRequest, api_key: s
             )
         )
 
-        async with ACTIVE_COMPLETIONS_LOCK:
-            existing = ACTIVE_COMPLETIONS.get(request_id)
-            if existing and not existing.done():
-                existing.cancel()
-            ACTIVE_COMPLETIONS[request_id] = inference_task
+        existing = ACTIVE_COMPLETIONS.get(request_id)
+        if existing and not existing.done():
+            existing.cancel()
+        ACTIVE_COMPLETIONS[request_id] = inference_task
 
         result = await inference_task
         content = result["content"] or ""
@@ -204,26 +198,17 @@ async def create_completion(request: Request, req: CompletionRequest, api_key: s
             _preview(content, 120),
         )
 
-        async def generate():
-            yield _sse_payload({"content": content})
-            yield _sse_payload({"done": True})
-
-        return StreamingResponse(generate(), media_type="text/event-stream")
+        return JSONResponse(content={"content": content, "request_id": request_id})
     except asyncio.CancelledError:
         logger.info("[%s] /v1/completions cancelled request_id=%s", request_tag, request_id)
-
-        async def cancelled():
-            yield _sse_payload({"cancelled": True, "request_id": request_id, "done": True})
-
-        return StreamingResponse(cancelled(), media_type="text/event-stream")
+        return JSONResponse(content={"cancelled": True, "request_id": request_id}, status_code=499)
     except Exception as e:
         logger.exception("[%s] /v1/completions failed request_id=%s: %s", request_tag, request_id, e)
         return JSONResponse(content={"error": str(e)}, status_code=500)
     finally:
-        async with ACTIVE_COMPLETIONS_LOCK:
-            active = ACTIVE_COMPLETIONS.get(request_id)
-            if active is not None and active is inference_task:
-                ACTIVE_COMPLETIONS.pop(request_id, None)
+        active = ACTIVE_COMPLETIONS.get(request_id)
+        if active is not None and active is inference_task:
+            ACTIVE_COMPLETIONS.pop(request_id, None)
 
 
 @app.post("/v1/completions/cancel")
@@ -399,7 +384,10 @@ if __name__ == "__main__":
     uvicorn.run(app, host="0.0.0.0", port=8001)
 
 
-# TTS and ASR routes
-from tts_asr import register_tts_asr_routes
-register_tts_asr_routes(app)
+# TTS and ASR routes (lazy loaded to avoid heavy import on startup)
+def _register_tts_asr_routes():
+    from tts_asr import register_tts_asr_routes
+    register_tts_asr_routes(app)
+
+_register_tts_asr_routes()