Update main.py

main.py

@@ -3,7 +3,7 @@ import httpx
 import json
 import time
 import asyncio
-import secrets  # <-- Added for new ID generation
+import secrets
 from fastapi import FastAPI, HTTPException, Security, Depends, status
 from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
 from fastapi.responses import StreamingResponse
@@ -22,7 +22,7 @@ if not SERVER_API_KEY:
     raise ValueError("SERVER_API_KEY environment variable not set. This is required to protect your server.")
 
 # FastAPI Init
-app = FastAPI(title="Replicate to OpenAI Compatibility Layer", version="9.2.6 (Dynamic Tokens & ID)")
+app = FastAPI(title="Replicate to OpenAI Compatibility Layer", version="9.2.7 (Non-Stream Fix)")
 
 # --- Authentication ---
 security = HTTPBearer()
@@ -288,8 +288,6 @@ async def create_chat_completion(request: ChatCompletionRequest):
     replicate_model_id = SUPPORTED_MODELS[request.model]
     formatted = format_messages_for_replicate(request.messages, request.functions)
     
-    # ### MAJOR FIX HERE (Max Tokens) ###
-    # Build the payload dynamically.
     replicate_input = {
         "prompt": formatted["prompt"],
         "temperature": request.temperature or 0.7,
@@ -297,14 +295,12 @@ async def create_chat_completion(request: ChatCompletionRequest):
     }
     
     # Only add max_new_tokens if the user *actually* provided it.
-    # If not provided, Replicate will use the model's own default.
     if request.max_tokens is not None:
         replicate_input["max_new_tokens"] = request.max_tokens
     
     if formatted["system_prompt"]: replicate_input["system_prompt"] = formatted["system_prompt"]
     if formatted["image"]: replicate_input["image"] = formatted["image"]
     
-    # ### MAJOR FIX HERE (Request ID) ###
     request_id = generate_request_id()
     
     if request.stream:
@@ -323,8 +319,19 @@ async def create_chat_completion(request: ChatCompletionRequest):
             resp = await client.post(url, headers=headers, json={"input": replicate_input}, timeout=300.0)
             resp.raise_for_status()
             pred = resp.json()
-            output = "".join(pred.get("output", []))
-            
+
+            # ### MAJOR FIX HERE (Non-Streaming Join Error) ###
+            # Robustly handle the 'output' field which could be a list, string, or null
+            raw_output = pred.get("output")
+
+            if isinstance(raw_output, list):
+                output = "".join(raw_output)  # Expected case: list of strings
+            elif isinstance(raw_output, str):
+                output = raw_output          # Handle if it's just a single string
+            else:
+                # Handle None, null, int, bool, or other unexpected types
+                output = ""                  
+
             output = output.strip() # Clean up any leading/trailing whitespace
             
             end_time = time.time()
@@ -360,6 +367,7 @@ async def create_chat_completion(request: ChatCompletionRequest):
         except httpx.HTTPStatusError as e:
             raise HTTPException(status_code=e.response.status_code, detail=f"Error from Replicate API: {e.response.text}")
         except Exception as e:
+            # Catch the join error and any others
             raise HTTPException(status_code=500, detail=f"Internal error: {str(e)}")
 
 @app.get("/")
@@ -367,7 +375,7 @@ async def root():
     """
     Root endpoint for health checks. Does not require authentication.
     """
-    return {"message": "Replicate to OpenAI Compatibility Layer API", "version": "9.2.6"}
+    return {"message": "Replicate to OpenAI Compatibility Layer API", "version": "9.2.7"}
 
 @app.middleware("http")
 async def add_performance_headers(request, call_next):
@@ -375,5 +383,5 @@ async def add_performance_headers(request, call_next):
     response = await call_next(request)
     process_time = time.time() - start_time
     response.headers["X-Process-Time"] = str(round(process_time, 3))
-    response.headers["X-API-Version"] = "9.2.6"
+    response.headers["X-API-Version"] = "9.2.7"
     return response