use threads

main.py

@@ -1,10 +1,12 @@
+import uvicorn
+import re
+import asyncio
 from fastapi import FastAPI
 from pydantic import BaseModel
 from huggingface_hub import InferenceClient
-import uvicorn
-import re
 from typing import List
 
+
 app = FastAPI()
 
 client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
@@ -69,4 +71,7 @@ async def generate_text(
         top_p=top_p,
         repetition_penalty=repetition_penalty
     )
-    return {"response": generate(item)}
+    
+    response = await asyncio.to_thread(generate, item)
+
+    return {"response": response}