Update app.py

app.py

@@ -80,7 +80,7 @@ def predict(vector):
     print(f"输入向量类型: {type(vector)}, 长度: {len(vector) if hasattr(vector, '__len__') else 'N/A'}")
 
     try:
-        # API调用会收到二维数组
+        # 处理不同输入格式
         if isinstance(vector, list) and len(vector) > 0 and isinstance(vector[0], list):
             # 已经是正确的二维数组格式
             pass
@@ -90,12 +90,12 @@ def predict(vector):
         else:
             error_msg = "错误：输入格式无效"
             print(error_msg)
-            return error_msg
+            return {"error": error_msg}
 
         if len(vector) != 1 or len(vector[0]) != 768:
             error_msg = f"错误：需要1x768的二维数组，收到{len(vector)}x{len(vector[0]) if vector else 0}"
             print(error_msg)
-            return error_msg
+            return {"error": error_msg}
 
         # 添加实际处理逻辑
         vector_array = np.array(vector, dtype=np.float32)
@@ -106,31 +106,74 @@ def predict(vector):
             try:
                 result = metadata.iloc[I[0][i]]
                 confidence = 1/(1+D[0][i])
-                results.append(f"匹配结果 {i+1}: {result['source']} | 置信度: {confidence:.2f}")
+                results.append({
+                    "source": result['source'],
+                    "confidence": round(float(confidence), 2)
+                })
             except Exception as e:
                 print(f"结果处理错误: {str(e)}")
-                results.append(f"结果 {i+1}: 数据获取失败")
+                results.append({"error": f"结果 {i+1}: 数据获取失败"})
 
         print(f"处理完成 | 耗时: {time.time()-start_time:.2f}秒")
         
-        # 返回Gradio期望的格式
-        return json.dumps({
-            "data": [
-                json.dumps({
-                    "results": results  
-                })
-            ]
-        })
+        # 返回标准化的JSON格式
+        return {
+            "status": "success",
+            "results": results
+        }
         
     except Exception as e:
         import traceback
         error_msg = f"处理错误: {str(e)}\n{traceback.format_exc()}"
         print(error_msg)
-        return json.dumps({
-            "error": "处理错误，请重试或联系管理员"
-        })
+        return {
+            "status": "error",
+            "message": "处理错误，请重试或联系管理员"
+        }
+
+# 创建FastAPI应用 (替代Gradio Blocks)
+import fastapi
+from fastapi import FastAPI, Request
+from fastapi.responses import JSONResponse
+from fastapi.middleware.cors import CORSMiddleware
+
+app = FastAPI()
+
+# 添加CORS支持
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.post("/predict")
+async def api_predict(request: Request):
+    """API预测端点"""
+    try:
+        data = await request.json()
+        vector = data.get("vector")
+        
+        if not vector or not isinstance(vector, list):
+            return JSONResponse(
+                status_code=400,
+                content={"status": "error", "message": "无效输入: 需要向量数据"}
+            )
+            
+        result = predict(vector)
+        return JSONResponse(content=result)
+        
+    except Exception as e:
+        return JSONResponse(
+            status_code=500,
+            content={
+                "status": "error",
+                "message": f"服务器内部错误: {str(e)}"
+            }
+        )
 
-# 创建Blocks应用
+# 保留Gradio界面
 with gr.Blocks() as demo:
     gr.Markdown("## 🛍 电商智能客服系统 (768维专用)")
     gr.Markdown("**使用CLIP-vit-large-patch14模型 | 向量维度: 768**")
@@ -142,7 +185,7 @@ with gr.Blocks() as demo:
             label="输入向量 (768维)",
             value=[[0.1]*768]
         )
-        output = gr.Textbox(label="智能回答", lines=5)
+        output = gr.JSON(label="智能回答")
     
     submit_btn = gr.Button("生成回答")
     submit_btn.click(
@@ -153,6 +196,21 @@ with gr.Blocks() as demo:
 
 # 启动应用
 if __name__ == "__main__":
+    import uvicorn
+    import threading
+    
+    # 启动FastAPI服务
+    def run_fastapi():
+        uvicorn.run(app, host="0.0.0.0", port=7860)
+    
+    # 启动Gradio界面
+    def run_gradio():
+        demo.launch(
+            server_name="0.0.0.0",
+            server_port=7861,
+            share=False
+        )
+    
     # 验证资源
     print("="*50)
     print("Space启动完成 | 准备接收请求")
@@ -160,8 +218,6 @@ if __name__ == "__main__":
     print(f"元数据记录数: {len(metadata)}")
     print("="*50)
     
-    # API端点自动创建在 /run/predict
-    demo.launch(
-        server_name="0.0.0.0",
-        server_port=7860
-    )
+    # 启动两个服务
+    threading.Thread(target=run_fastapi, daemon=True).start()
+    run_gradio()