Update app.py

app.py

@@ -140,6 +140,118 @@
 # if __name__ == "__main__":
 #     demo.launch()
 
+# import gradio as gr
+# from huggingface_hub import InferenceClient
+# import tempfile
+
+# # Initialize clients
+# chat_client = InferenceClient("Qwen/Qwen2.5-Coder-32B-Instruct")
+# stt_client = InferenceClient("openai/whisper-large-v3")
+
+# def transcribe_audio(audio_file):
+#     """Convert audio to text using Whisper model"""
+#     with open(audio_file, "rb") as f:
+#         result = stt_client.automatic_speech_recognition(f.read())
+#     return result.text  # Extract only the text from the response
+
+# def respond(history, query):
+#     system_message = """You are a friendly Product Assistant. Follow these rules:
+#     1. If the query is product-related, provide structured recommendations
+#     2. Consider both voice and text inputs equally
+#     3. Format responses with bullet points and emojis
+#     4. Always acknowledge voice queries specifically"""
+    
+#     product_prompt = f"""Analyze this {'voice' if history[-1][0] == query else 'text'} query:
+#     "{query}"
+#     Recommend products considering:
+#     - User intent
+#     - Semantic meaning
+#     - Potential use cases
+#     - Price ranges
+#     Provide ranked list with brief explanations"""
+    
+#     messages = [
+#         {"role": "system", "content": system_message},
+#         {"role": "user", "content": product_prompt}
+#     ]
+
+#     # Build conversation history
+#     for entry in history[:-1]:  # Exclude current query
+#         messages.extend([
+#             {"role": "user", "content": entry[0]},
+#             {"role": "assistant", "content": entry[1]}
+#         ])
+
+#     # Generate streamed response
+#     response = ""
+#     for chunk in chat_client.chat_completion(
+#         messages,
+#         max_tokens=2048,
+#         stream=True,
+#         temperature=0.7,
+#         top_p=0.95,
+#     ):
+#         token = chunk.choices[0].delta.content
+#         response += token
+#         history[-1] = (history[-1][0], response)  # Update last entry
+#         yield history
+
+# # Custom styling
+# css = """
+# .gradio-container { background: #f5f7fa !important; }
+# .audio-input { background: white !important; border-radius: 10px; }
+# """
+
+# with gr.Blocks(css=css) as demo:
+#     gr.Markdown("# 🎤 Voice-Activated Product Advisor 🛍️")
+    
+#     with gr.Row():
+#         chatbot = gr.Chatbot(height=600, bubble_full_width=False)
+        
+#         with gr.Column():
+#             with gr.Tab("🎙️ Voice Input"):
+#                 audio_input = gr.Audio(
+#                     sources="microphone",
+#                     type="filepath",
+#                     label="Speak your product request",
+#                     elem_classes="audio-input"
+#                 )
+#             with gr.Tab("📝 Text Input"):
+#                 text_input = gr.Textbox(label="Type your request")
+#             submit_btn = gr.Button("🚀 Get Recommendations", variant="primary")
+
+#     def process_inputs(text, audio, history):
+#         """Handle both input types"""
+#         query = text.strip()
+#         if audio:
+#             query = transcribe_audio(audio)
+#             # Add voice-specific marker
+#             query = f"🎤 Voice Query: {query}"
+        
+#         if query:
+#             return history + [(query, None)]  # Proper tuple format
+#         return history
+
+#     submit_btn.click(
+#         process_inputs,
+#         [text_input, audio_input, chatbot],
+#         chatbot,
+#         queue=False
+#     ).then(
+#         respond,
+#         [chatbot, text_input],
+#         chatbot
+#     )
+
+#     # Clear inputs after submission
+#     submit_btn.click(
+#         lambda: [None, None],  # Clear audio and text inputs
+#         outputs=[text_input, audio_input]
+#     )
+
+# if __name__ == "__main__":
+#     demo.launch()
+
 import gradio as gr
 from huggingface_hub import InferenceClient
 import tempfile
@@ -150,38 +262,33 @@ stt_client = InferenceClient("openai/whisper-large-v3")
 
 def transcribe_audio(audio_file):
     """Convert audio to text using Whisper model"""
-    with open(audio_file, "rb") as f:
-        result = stt_client.automatic_speech_recognition(f.read())
-    return result.text  # Extract only the text from the response
+    try:
+        with open(audio_file, "rb") as f:
+            result = stt_client.automatic_speech_recognition(f.read())
+        return result.text
+    except Exception as e:
+        print(f"Transcription error: {e}")
+        return "Could not process audio. Please try again."
 
 def respond(history, query):
-    system_message = """You are a friendly Product Assistant. Follow these rules:
-    1. If the query is product-related, provide structured recommendations
-    2. Consider both voice and text inputs equally
-    3. Format responses with bullet points and emojis
-    4. Always acknowledge voice queries specifically"""
+    system_message = """You are a Voice-Aware Product Assistant. Rules:
+    1. Always acknowledge voice queries with 🎧 icon
+    2. Provide structured recommendations with emojis
+    3. Consider both voice and text inputs equally"""
     
-    product_prompt = f"""Analyze this {'voice' if history[-1][0] == query else 'text'} query:
-    "{query}"
+    product_prompt = f"""Analyze this {'🎧 VOICE' if '🎧' in query else '📝 TEXT'} query:
+    "{query.replace('🎧 VOICE: ', '')}"
     Recommend products considering:
-    - User intent
+    - Voice tone analysis (if audio)
     - Semantic meaning
-    - Potential use cases
-    - Price ranges
-    Provide ranked list with brief explanations"""
+    - User intent
+    - Price context"""
     
     messages = [
         {"role": "system", "content": system_message},
         {"role": "user", "content": product_prompt}
     ]
 
-    # Build conversation history
-    for entry in history[:-1]:  # Exclude current query
-        messages.extend([
-            {"role": "user", "content": entry[0]},
-            {"role": "assistant", "content": entry[1]}
-        ])
-
     # Generate streamed response
     response = ""
     for chunk in chat_client.chat_completion(
@@ -193,49 +300,65 @@ def respond(history, query):
     ):
         token = chunk.choices[0].delta.content
         response += token
-        history[-1] = (history[-1][0], response)  # Update last entry
+        history[-1] = (history[-1][0], response)
         yield history
 
-# Custom styling
 css = """
-.gradio-container { background: #f5f7fa !important; }
+.gradio-container { background: #f8f9fa !important; }
 .audio-input { background: white !important; border-radius: 10px; }
+.mic-status { color: #4a90e2; font-weight: bold; }
 """
 
-with gr.Blocks(css=css) as demo:
-    gr.Markdown("# 🎤 Voice-Activated Product Advisor 🛍️")
+with gr.Blocks(css=css, title="Voice Product Assistant") as demo:
+    gr.Markdown("# 🎧 Voice-Activated Product Advisor 🛒")
     
     with gr.Row():
         chatbot = gr.Chatbot(height=600, bubble_full_width=False)
         
         with gr.Column():
-            with gr.Tab("🎙️ Voice Input"):
+            # Audio input with status indicator
+            with gr.Group():
                 audio_input = gr.Audio(
                     sources="microphone",
                     type="filepath",
-                    label="Speak your product request",
-                    elem_classes="audio-input"
+                    label="Click mic & speak",
+                    elem_classes="audio-input",
+                    interactive=True
                 )
-            with gr.Tab("📝 Text Input"):
-                text_input = gr.Textbox(label="Type your request")
+                mic_status = gr.Markdown("🔴 Mic offline", elem_classes="mic-status")
+            
+            # Text input
+            text_input = gr.Textbox(label="Or type your request")
+            
+            # Unified submit button
             submit_btn = gr.Button("🚀 Get Recommendations", variant="primary")
 
+    # Handle audio permissions
+    def request_mic_access():
+        return gr.update(text="🟢 Mic ready") if audio_input.is_enabled else gr.update(text="🔴 Mic blocked")
+
+    # Process inputs
     def process_inputs(text, audio, history):
-        """Handle both input types"""
         query = text.strip()
         if audio:
-            query = transcribe_audio(audio)
-            # Add voice-specific marker
-            query = f"🎤 Voice Query: {query}"
+            transcript = transcribe_audio(audio)
+            query = f"🎧 VOICE: {transcript}"
         
         if query:
-            return history + [(query, None)]  # Proper tuple format
-        return history
+            return history + [(query, None)], ""
+        return history, ""
 
+    # Connect all components
+    audio_input.change(
+        request_mic_access,
+        outputs=mic_status,
+        queue=False
+    )
+    
     submit_btn.click(
         process_inputs,
         [text_input, audio_input, chatbot],
-        chatbot,
+        [chatbot, text_input],
         queue=False
     ).then(
         respond,
@@ -243,11 +366,5 @@ with gr.Blocks(css=css) as demo:
         chatbot
     )
 
-    # Clear inputs after submission
-    submit_btn.click(
-        lambda: [None, None],  # Clear audio and text inputs
-        outputs=[text_input, audio_input]
-    )
-
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(server_port=7860, share=False)