init commit

app.py

@@ -1,31 +1,14 @@
-"""
-app.py — HF Space frontend server
-===================================
-Lives in the Hugging Face Space. Does two things only:
-
-  1. Serves index.html (and static assets) to the browser
-  2. Proxies POST /generate to the remote Gradio server running on your machine,
-     translating between the frontend's JSON schema and Gradio's /run/predict API
-
-Set GRADIO_SERVER_URL to the public *.gradio.live URL that server.py prints
-when it starts (or your own domain if you're running behind a reverse proxy).
-
-Architecture
-------------
-Browser  →  POST /generate (HF Space)  →  POST /run/predict (your machine)
-                    app.py                         server.py
-"""
-
 from flask import Flask, request, jsonify, send_from_directory
 import requests
 import os
 
 app = Flask(__name__, static_folder="static")
 
-# Paste the *.gradio.live URL printed by server.py here (no trailing slash)
-GRADIO_SERVER_URL = "https://c9ceb67adbae7238f6.gradio.live"
+# Put your *.gradio.live or *.hf.space URL (no trailing slash)
+GRADIO_SERVER_URL = os.getenv("GRADIO_SERVER_URL", "https://c9ceb67adbae7238f6.gradio.live").rstrip("/")
 
-GRADIO_PREDICT_URL = f"{GRADIO_SERVER_URL}/run/predict"
+# Default endpoint — some Spaces expose /predict by default
+GRADIO_PREDICT_URL = f"{GRADIO_SERVER_URL}/predict"
 
 
 @app.route("/")
@@ -35,19 +18,27 @@ def index():
 
 @app.route("/static/<path:filename>")
 def static_files(filename):
-    return send_from_directory("static", filename)
+    return send_from_directory(app.static_folder, filename)
 
 
 @app.route("/generate", methods=["POST"])
 def generate():
-    body = request.get_json(force=True)
+    body = request.get_json(silent=True) or {}
 
-    prompt = body.get("prompt", "").strip()
+    prompt = str(body.get("prompt", "")).strip()
     if not prompt:
         return jsonify({"error": "prompt is required"}), 400
 
-    # Gradio expects inputs as a positional list matching the server.py fn signature:
-    # [prompt, max_new_tokens, temperature, top_p, repetition_penalty]
+    # You can optionally override if the Space uses a custom API name.
+    # Example: body["api_name"] = "/generate" or "/my_api"
+    api_name = body.get("api_name", "/predict")
+    if not api_name.startswith("/"):
+        api_name = "/" + api_name
+
+    api_url = f"{GRADIO_SERVER_URL}{api_name}"
+
+    # Gradio expects positional arguments if the endpoint is /predict
+    # Usually it's a list matching the input components
     gradio_payload = {
         "data": [
             prompt,
@@ -59,18 +50,25 @@ def generate():
     }
 
     try:
-        resp = requests.post(GRADIO_PREDICT_URL, json=gradio_payload, timeout=120)
+        resp = requests.post(api_url, json=gradio_payload, timeout=120)
         resp.raise_for_status()
     except requests.exceptions.ConnectionError:
-        return jsonify({"error": "Cannot reach inference server. Is server.py running?"}), 502
+        return jsonify({"error": "Cannot reach inference server."}), 502
     except requests.exceptions.Timeout:
         return jsonify({"error": "Inference server timed out."}), 504
     except requests.exceptions.HTTPError as e:
-        return jsonify({"error": f"Inference server error: {e}"}), 502
+        return jsonify({"error": f"Inference server error: {str(e)}", "response": resp.text}), 502
+
+    try:
+        result = resp.json()
+        # for many Spaces, the first item in "data" is the output
+        generated_text = result.get("data", [None])[0] if isinstance(result, dict) else None
+    except Exception:
+        return jsonify({"error": "Unexpected inference server response", "raw": resp.text}), 502
 
-    generated_text = resp.json()["data"][0]
     return jsonify({"generated_text": generated_text, "prompt": prompt})
 
 
 if __name__ == "__main__":
-    app.run(debug=True, port=7860)
+    port = int(os.getenv("PORT", "7860"))
+    app.run(port=port, debug=False)