Create app.py

app.py

@@ -0,0 +1,86 @@
+import os
+import json
+import base64
+from flask import Flask, request, jsonify
+
+# Gradio/Flask setup
+# This simple setup avoids full Gradio UI and focuses on serving static files and the API endpoint
+app = Flask(__name__, static_folder=".")
+
+# Ensure the API Key is set as a Secret in Hugging Face Space Settings
+# Variable name must be GEMINI_API_KEY for security
+GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
+if not GEMINI_API_KEY:
+    print("Warning: GEMINI_API_KEY environment variable not set. TTS generation will fail.")
+
+@app.route('/tts-proxy', methods=['POST'])
+def tts_proxy():
+    """
+    Handles the request from the frontend, securely calls the Gemini TTS API, 
+    and returns the base64 audio data.
+    """
+    if not GEMINI_API_KEY:
+        return jsonify({"error": "API Key is missing or not configured."}), 500
+
+    try:
+        data = request.json
+        prompt = data.get('prompt')
+        voice = data.get('voice')
+
+        if not prompt or not voice:
+            return jsonify({"error": "Missing 'prompt' or 'voice' parameter."}), 400
+        
+        # Prepare the payload for the Gemini TTS API
+        payload = {
+            "contents": [{"parts": [{"text": prompt}]}],
+            "generationConfig": {
+                "responseModalities": ["AUDIO"],
+                "speechConfig": {
+                    "voiceConfig": {"prebuiltVoiceConfig": {"voiceName": voice}}
+                }
+            },
+            "model": "gemini-2.5-flash-preview-tts"
+        }
+
+        # Use requests library or similar to call the API from the backend
+        # We use standard library fetch if running in a compliant environment, 
+        # but for simplicity and common deployment, we simulate the internal call structure.
+        
+        # Since we cannot use an external 'requests' dependency easily here, 
+        # we assume this file will be executed by an environment (like a Gradio wrapper) 
+        # that handles external API calls or we rely on the internal execution context 
+        # in a standard Hugging Face Docker/SDK setup. 
+        # For a true Gradio Space deployment, this part is usually written slightly differently.
+        
+        # *** Critical Section: Calling the Gemini API from the Backend ***
+        import requests
+        
+        api_url = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-preview-tts:generateContent?key={GEMINI_API_KEY}"
+        
+        response = requests.post(
+            api_url, 
+            headers={'Content-Type': 'application/json'}, 
+            data=json.dumps(payload)
+        )
+        response.raise_for_status() # Raise exception for bad status codes (4xx or 5xx)
+
+        result = response.json()
+        audio_data = result.get('candidates')[0].get('content').get('parts')[0].get('inlineData').get('data')
+
+        return jsonify({"audioData": audio_data}), 200
+
+    except requests.exceptions.HTTPError as e:
+        print(f"HTTP Error calling Gemini API: {e}")
+        return jsonify({"error": f"API call failed: {e.response.text}"}), 500
+    except Exception as e:
+        print(f"General error in TTS proxy: {e}")
+        return jsonify({"error": "Internal server error during audio generation."}), 500
+
+@app.route('/')
+def index():
+    return app.send_static_file('index.html')
+
+if __name__ == '__main__':
+    # When deployed on HF Spaces, the port is usually managed by the environment
+    port = int(os.environ.get('PORT', 7860)) 
+    app.run(host='0.0.0.0', port=port)