Update app.py

app.py

@@ -76,7 +76,7 @@ def structured_retriever(question: str) -> str:
 
 # Function to generate audio with Eleven Labs TTS
 def generate_audio_elevenlabs(text):
-    XI_API_KEY = os.environ['ELEVENLABS_API']
+    XI_API_KEY = os.environ.get('ELEVENLABS_API')
     VOICE_ID = 'ehbJzYLQFpwbJmGkqbnW'
     tts_url = f"https://api.elevenlabs.io/v1/text-to-speech/{VOICE_ID}/stream"
     headers = {
@@ -93,20 +93,27 @@ def generate_audio_elevenlabs(text):
             "use_speaker_boost": False
         }
     }
-    response = requests.post(tts_url, headers=headers, json=data, stream=True)
-    if response.ok:
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as f:
-            for chunk in response.iter_content(chunk_size=1024):
-                if chunk:
-                    f.write(chunk)
-            audio_path = f.name
-        logging.debug(f"Audio saved to {audio_path}")
-        return audio_path  # Return audio path for automatic playback
-    else:
-        logging.error(f"Error generating audio: {response.text}")
+    
+    try:
+        logging.debug(f"Sending request to Eleven Labs with text: {text[:100]}...")
+        response = requests.post(tts_url, headers=headers, json=data, stream=True)
+        
+        if response.ok:
+            logging.debug("Received successful response from Eleven Labs API.")
+            with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as f:
+                for chunk in response.iter_content(chunk_size=1024):
+                    if chunk:
+                        f.write(chunk)
+                audio_path = f.name
+            logging.debug(f"Audio successfully saved to {audio_path}")
+            return audio_path
+        else:
+            logging.error(f"Error generating audio: {response.status_code} - {response.text}")
+            return None
+    except Exception as e:
+        logging.error(f"Exception during audio generation: {str(e)}")
         return None
 
-
 # Define the ASR model with Whisper
 model_id = 'openai/whisper-large-v3'
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -165,12 +172,10 @@ def transcribe_and_respond(audio):
     
     return audio_path, response_text
 
-
 # Function to clear the transcription state
 def clear_transcription_state():
     return None, None
 
-
 # Define the Gradio interface with only audio input and output
 with gr.Blocks(theme="rawrsor1/Everforest") as demo:
     with gr.Row():