Fix: Use direct FastAPI decorators with JSONResponse to avoid Content-Length conflicts

app.py

@@ -5,7 +5,7 @@ from functools import lru_cache
 from typing import List, Optional, Tuple
 
 import torch
-from fastapi import APIRouter, HTTPException
+from fastapi import HTTPException
 from pydantic import BaseModel
 
 try:
@@ -483,24 +483,26 @@ with gr.Blocks(
         outputs=[prompt_input, output, status_display],
     )
 
-# Attach API routes directly onto Gradio's FastAPI instance using APIRouter
-api_router = APIRouter()
+# Add API routes directly to Gradio's FastAPI app
+# These routes are added after Gradio Blocks context but before queue/launch
+from fastapi.responses import JSONResponse
+from fastapi import Request
 
-
-@api_router.get("/health")
-def api_health() -> dict[str, str]:
+@gradio_app.app.get("/health", response_class=JSONResponse)
+def api_health():
     """API health check endpoint."""
     return healthcheck()
 
-
-@api_router.post("/v1/generate", response_model=GenerateResponse)
-def api_generate(payload: GeneratePayload) -> GenerateResponse:
+@gradio_app.app.post("/v1/generate", response_class=JSONResponse)
+async def api_generate(request: Request):
     """API generate endpoint."""
-    return generate_endpoint(payload)
-
-
-# Include the router in Gradio's FastAPI app
-gradio_app.app.include_router(api_router)
+    try:
+        data = await request.json()
+        payload = GeneratePayload(**data)
+        result = generate_endpoint(payload)
+        return {"text": result.text}
+    except Exception as exc:
+        raise HTTPException(status_code=500, detail=str(exc))
 
 # Call warm start
 warm_start()