Update main.py

main.py

@@ -30,10 +30,9 @@ async def lifespan(app: FastAPI):
     """
     Manages the lifecycle of the HTTPX client.
     The client is created on startup and gracefully closed on shutdown.
-    WARNING: This client has no timeout and no explicit connection pool limits.
+    NOTE: The client has no timeout, which is a deliberate choice for a proxy
+    that should not impose its own timeout limits on long-running requests.
     """
-    # timeout=None disables all client-side timeouts.
-    # The absence of a `limits` parameter means we rely on system defaults.
     async with httpx.AsyncClient(base_url=TARGET_URL, timeout=None) as client:
         app.state.http_client = client
         yield
@@ -53,25 +52,15 @@ async def _reverse_proxy(request: Request):
     url = httpx.URL(path=request.url.path, query=request.url.query.encode("utf-8"))
 
     # --- Header Processing ---
-    # Start with headers from the incoming request.
     request_headers = dict(request.headers)
+    request_headers.pop("host", None) # The 'host' header is managed by httpx.
 
-    # 1. CRITICAL: Remove host header.
-    #    The 'host' header is managed by httpx.
-    request_headers.pop("host", None)
-
-    # 2. Get the user's authorization key from the incoming request.
     authorization_header = request.headers.get("authorization")
-
-    # 3. Generate a random IP for spoofing headers.
     random_ip = generate_random_ip()
     
-    # --- ADDED LOGGING ---
-    # Log the original client IP and the spoofed IP being used for the request.
     logging.info(f"Client '{request.client.host}' is being proxied with spoofed IP: {random_ip}")
 
-    # 4. Set the specific, required headers for the target API.
-    #    This will overwrite any conflicting headers from the original request.
+    # Set the specific, required headers for the target API.
     specific_headers = {
         "accept": "application/json, text/plain, */*",
         "accept-language": "en-US,en;q=0.9,ru;q=0.8",
@@ -91,29 +80,24 @@ async def _reverse_proxy(request: Request):
     }
     request_headers.update(specific_headers)
 
-    # 5. Add the user's authorization key to the headers if it exists.
     if authorization_header:
         request_headers["authorization"] = authorization_header
 
-    # Read the request body once, as it will be reused in case of retries.
     body = await request.body()
     
     # --- Retry Logic ---
     last_exception = None
     for attempt in range(MAX_RETRIES):
         try:
-            # Build the request for each attempt to ensure the content stream is fresh.
             rp_req = client.build_request(
                 method=request.method,
                 url=url,
                 headers=request_headers,
                 content=body,
             )
-            # Send the request and get a streaming response.
             rp_resp = await client.send(rp_req, stream=True)
 
-            # If the status code is not in our retry list, we can return the response.
-            if rp_resp.status_code not in RETRY_STATUS_CODES:
+            if rp_resp.status_code not in RETRY_STATUS_CODES or attempt == MAX_RETRIES - 1:
                 return StreamingResponse(
                     rp_resp.aiter_raw(),
                     status_code=rp_resp.status_code,
@@ -121,26 +105,16 @@ async def _reverse_proxy(request: Request):
                     background=BackgroundTask(rp_resp.aclose),
                 )
             
-            # If it's a retryable error but this is the last attempt, return the error response.
-            if attempt == MAX_RETRIES - 1:
-                 return StreamingResponse(
-                    rp_resp.aiter_raw(),
-                    status_code=rp_resp.status_code,
-                    headers=rp_resp.headers,
-                    background=BackgroundTask(rp_resp.aclose),
-                )
-
-            # Otherwise, close the unsuccessful response before trying again.
             await rp_resp.aclose()
 
         except httpx.ConnectError as e:
             last_exception = e
-            # Continue to the next attempt if a connection error occurs.
+            logging.warning(f"Attempt {attempt + 1}/{MAX_RETRIES} failed with connection error: {e}")
 
-    # If all retry attempts failed with a connection error, raise a final exception.
+    # If all retry attempts failed, raise a final exception.
     raise HTTPException(
         status_code=502,
-        detail=f"Bad Gateway: Cannot connect to target service after {MAX_RETRIES} attempts. {last_exception}"
+        detail=f"Bad Gateway: Cannot connect to target service after {MAX_RETRIES} attempts. Last error: {last_exception}"
     )
 
 
@@ -156,10 +130,7 @@ async def chat_proxy_handler(request: Request):
     """
     return await _reverse_proxy(request)
 
-# A simple root endpoint for health checks.
 @app.get("/")
 async def health_check():
     """Provides a basic health check endpoint."""
-    return {"status": "ok", "proxying_endpoint": "/completions", "target": "TypeGPT"}
-
-# Any request to a path other than "/completions" or "/" will result in a 404 Not Found.
+    return {"status": "ok", "proxying_endpoint": "/completions", "target": "TypeGPT"}