Update main.py

main.py

@@ -4,7 +4,6 @@ from fastapi import FastAPI, HTTPException, Request
 from fastapi.responses import StreamingResponse, HTMLResponse
 from pydantic import BaseModel
 import httpx
-from functools import lru_cache
 
 # Load environment variables from .env file
 load_dotenv()
@@ -51,9 +50,9 @@ async def root():
     """
     return HTMLResponse(content=html_content)
 
-# Cache function with lru_cache
-@lru_cache(maxsize=1)
-async def get_cached_models():
+# Remove cache from get_models
+@app.get("/v1/models")
+async def get_models():
     async with httpx.AsyncClient() as client:
         try:
             response = await client.get(f"{secret_api_endpoint}/v1/models", timeout=3)
@@ -62,10 +61,6 @@ async def get_cached_models():
         except httpx.RequestError as e:
             raise HTTPException(status_code=500, detail=f"Request failed: {e}")
 
-@app.get("/v1/models")
-async def get_models():
-    return await get_cached_models()
-
 @app.post("/v1/chat/completions")
 async def get_completion(payload: Payload, request: Request):
     api_key = request.headers.get("Authorization")
@@ -103,4 +98,4 @@ async def startup_event():
 # Run the server with Uvicorn using the 'main' module
 if __name__ == "__main__":
     import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=8000)
+    uvicorn.run(app, host="0.0.0.0", port=8000)