代码初始化

inference/inference.py

@@ -1,22 +1,41 @@
-import requests
-from fastapi import APIRouter, Header
+import httpx
+from fastapi import APIRouter, Header, HTTPException
 from .apiModel import Payload
+import time
 
 router = APIRouter()
 
 API_URL = "https://api-inference.huggingface.co/models/cardiffnlp/twitter-roberta-base-sentiment-latest"
 
 
+# 使用httpx异步请求
+async def fetch_model_response(payload: dict, headers: dict):
+    async with httpx.AsyncClient() as client:
+        try:
+            response = await client.post(API_URL, headers=headers, json=payload)
+            if response.status_code == 503:
+                # 如果模型正在加载，等待并重试
+                print("模型加载中，等待中...")
+                await asyncio.sleep(20)  # 等待20秒
+                return await fetch_model_response(payload, headers)  # 重试请求
+            response.raise_for_status()  # 如果返回错误状态码，会抛出异常
+            return response.json()
+        except httpx.RequestError as e:
+            raise HTTPException(status_code=500, detail=f"请求错误: {e}")
+        except httpx.HTTPStatusError as e:
+            raise HTTPException(status_code=response.status_code, detail=f"HTTP 错误: {e}")
+
+
 @router.post("/api-inference/")
 async def api_inference(
         authorization: str = Header(...),
         item: Payload = None):
     print("请求：", item)
+
+    # 设置请求头
     headers = {"Authorization": authorization}
-    # Forward request to Hugging Face API
-    response = requests.post(
-        API_URL,
-        headers=headers,
-        json=item.dict()
-    )
-    return response.json()
+
+    # 使用异步请求
+    response_data = await fetch_model_response(item.dict(), headers)
+
+    return response_data

requirements.txt

@@ -1,3 +1,4 @@
 fastapi
+httpx
 requests
 uvicorn[standard]