Update app.py

app.py

@@ -1,60 +1,23 @@
-# ---------- DO THIS FIRST ----------
+# ---------- Gradio Space entrypoint (no FastAPI/Uvicorn) ----------
 import os
 os.environ.setdefault("GRADIO_USE_CDN", "true")
 
-# Safe import of 'spaces' (works on CPU or ZeroGPU)
-try:
-    import spaces  # HF Spaces SDK
-except Exception:
-    class _DummySpaces:
-        def GPU(self, *_, **__):
-            def _decorator(fn): return fn
-            return _decorator
-    spaces = _DummySpaces()
-
 import gradio as gr
-from fastapi import FastAPI
 
-# Log a few things so we can see them in the container logs
+# (Optional) ZeroGPU probe — harmless on CPU. Will work once you switch HW to ZeroGPU.
 try:
-    import sys
-    print("== Sanity ==")
-    print("gradio.__version__:", getattr(gr, "__version__", "unknown"))
-    print("SPACE_RUNTIME:", os.getenv("SPACE_RUNTIME"))
-    print("PY:", sys.version)
-except Exception as e:
-    print("sanity log error:", e)
-
-# ---------- PUBLIC GPU PROBES (harmless on CPU) ----------
-@spaces.GPU(duration=10)
-def gpu_probe(a: int = 1, b: int = 1):
-    return a + b
-
-@spaces.GPU(duration=10)
-def gpu_echo(x: str = "ok"):
-    return x
+    import spaces
+    @spaces.GPU(duration=10)
+    def gpu_probe(a: int = 1, b: int = 1):
+        return a + b
+except Exception:
+    pass
 
-# ---------- Tiny Gradio UI ----------
 with gr.Blocks(title="Hello Space") as demo:
     gr.Markdown("### ✅ App is alive\nType to echo.")
     inp = gr.Textbox(label="Input", value="hello")
     out = gr.Textbox(label="Output")
     inp.submit(lambda s: f"echo: {s}", inp, out)
 
-# Queue BEFORE mounting
-demo = demo.queue(max_size=8)
-
-# ---------- ASGI app and fast health ----------
-app = FastAPI()
-
-@app.get("/health")
-def _health():
-    return {"ok": True}
-
-# Mount Gradio at "/" so the supervisor sees it
-app = gr.mount_gradio_app(app, demo, path="/")
-
-# (Do NOT call demo.launch() on Spaces)
-if __name__ == "__main__":
-    import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=7860)
+# IMPORTANT: expose a top-level `demo`, and DO NOT call .launch()
+demo = demo.queue(max_size=8)