Fix: Use demo.launch() for ZeroGPU, insert route at position 0 for priority

- Revert to demo.launch() which is required for ZeroGPU detection
- Insert custom route at position 0 in demo.app.routes to take priority over Gradio's catch-all
- Remove unused FastAPI app creation

app.py

@@ -11,7 +11,7 @@ import re
 import sys
 import warnings
 
-from fastapi import FastAPI, Request
+from fastapi import Request
 from fastapi.responses import StreamingResponse
 
 from diffusers import AutoencoderKL, FlowMatchEulerDiscreteScheduler
@@ -586,12 +586,7 @@ with gr.Blocks(title="Z-Image Generation MCP") as demo:
         api_name="generate_image",  # nice, stable name for tool clients
     )
 
-# ==================== FastAPI App with Custom Routes ====================
-# Create our own FastAPI app so custom routes take precedence over Gradio's catch-all
-app = FastAPI()
-
-
-@app.get("/generate/{prompt_path:path}")
+# ==================== /generate/<prompt>.png Endpoint ====================
 def http_generate_png(
     prompt_path: str,
     request: Request,
@@ -691,9 +686,16 @@ def http_generate_png(
     return StreamingResponse(io.BytesIO(png_bytes), media_type="image/png")
 
 
-# Mount Gradio app onto our FastAPI app
-app = gr.mount_gradio_app(app, demo, path="/")
+# Add custom route to Gradio's FastAPI app (insert at position 0 for priority)
+from fastapi.routing import APIRoute
+demo.app.routes.insert(
+    0,
+    APIRoute(
+        "/generate/{prompt_path:path}",
+        http_generate_png,
+        methods=["GET"],
+    ),
+)
 
 if __name__ == "__main__":
-    import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=7860)
+    demo.launch(mcp_server=True)