Update app.py

app.py

@@ -28,9 +28,7 @@ llama_cloud_api_key = os.environ.get("LLAMA_CLOUD_API_KEY")
 ACCOUNT_ID = os.environ.get("CLOUDFARE_ACCOUNT_ID")
 API_TOKEN = os.environ.get("CLOUDFLARE_AUTH_TOKEN")
 API_BASE_URL = "https://api.cloudflare.com/client/v4/accounts/a17f03e0f049ccae0c15cdcf3b9737ce/ai/run/"
-
-# Add this to your existing imports and configurations
-whisper_client = InferenceClient("openai/whisper-large-v3", token=huggingface_token)
+whisper_api = InferenceApi("openai/whisper-large-v2", token=huggingface_token)
 
 print(f"ACCOUNT_ID: {ACCOUNT_ID}")
 print(f"CLOUDFLARE_AUTH_TOKEN: {API_TOKEN[:5]}..." if API_TOKEN else "Not set")
@@ -280,46 +278,21 @@ def generate_chunked_response(prompt, model, max_tokens=10000, num_calls=3, temp
     print(f"Final clean response: {final_response[:100]}...")
     return final_response
 
-def chatbot_interface(message, history, model, temperature, num_calls, use_web_search, selected_docs, audio):
+def chatbot_interface(message, history, model, temperature, num_calls):
+    if not message.strip():
+        return "", history
+
+    history = history + [(message, "")]
+
     try:
-        if not isinstance(history, list):
-            history = []
-        
-        # Prioritize text input
-        if message and isinstance(message, str) and message.strip():
-            input_text = message.strip()
-        elif audio is not None:
-            # Only use audio if text input is empty and audio is provided
-            transcribed_text = transcribe_audio(audio)
-            if transcribed_text:
-                input_text = transcribed_text
-            else:
-                return history  # Return if both text and audio are empty
-        else:
-            return history  # Return if both text and audio are empty
-        
-        # Add user message to history
-        history.append((input_text, None))
-        
-        # Get response from the model
-        response_generator = respond(input_text, history, model, temperature, num_calls, use_web_search, selected_docs)
-        
-        full_response = ""
-        for partial_response in response_generator:
-            full_response += str(partial_response)
-            # Update the last message in history
-            history[-1] = (input_text, full_response)
-            yield history
-        
-        # If no response was generated, yield an error message
-        if not full_response:
-            history[-1] = (input_text, "I'm sorry, I couldn't generate a response. Please try again.")
+        for response in respond(message, history, model, temperature, num_calls):
+            history[-1] = (message, response)
             yield history
-
+    except gr.CancelledError:
+        yield history
     except Exception as e:
-        logging.error(f"Error in chatbot_interface: {str(e)}")
-        error_message = f"An error occurred: {str(e)}"
-        history.append((input_text if 'input_text' in locals() else "Error", error_message))
+        logging.error(f"Unexpected error in chatbot_interface: {str(e)}")
+        history[-1] = (message, f"An unexpected error occurred: {str(e)}")
         yield history
 
 def retry_last_response(history, model, temperature, num_calls):
@@ -430,12 +403,13 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
     logging.info(f"Model Used: {model}")
     logging.info(f"Selected Documents: {selected_docs}")
     logging.info(f"Use Web Search: {use_web_search}")
-    
+
     if use_web_search:
         original_query = message
         rephrased_query = rephrase_query(message, conversation_manager)
         logging.info(f"Original query: {original_query}")
         logging.info(f"Rephrased query: {rephrased_query}")
+
         final_summary = ""
         for _ in range(num_calls):
             search_results = get_web_search_results(rephrased_query)
@@ -445,7 +419,8 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
                 final_summary += search_results[0]["error"] + "\n\n"
             else:
                 summary = summarize_web_results(rephrased_query, search_results, conversation_manager)
-                final_summary += str(summary) + "\n\n"
+                final_summary += summary + "\n\n"
+
         if final_summary:
             conversation_manager.add_interaction(original_query, final_summary)
             yield final_summary
@@ -476,21 +451,21 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
             if model.startswith("duckduckgo/"):
                 # Use DuckDuckGo chat with context
                 for partial_response in get_response_from_duckduckgo(message, model, context_str, num_calls, temperature):
-                    yield str(partial_response)
+                    yield partial_response
             elif model == "@cf/meta/llama-3.1-8b-instruct":
                 # Use Cloudflare API
                 for partial_response in get_response_from_cloudflare(prompt="", context=context_str, query=message, num_calls=num_calls, temperature=temperature, search_type="pdf"):
-                    yield str(partial_response)
+                    yield partial_response
             else:
                 # Use Hugging Face API
                 for partial_response in get_response_from_pdf(message, model, selected_docs, num_calls=num_calls, temperature=temperature):
-                    yield str(partial_response)
+                    yield partial_response
         except Exception as e:
             logging.error(f"Error with {model}: {str(e)}")
             if "microsoft/Phi-3-mini-4k-instruct" in model:
                 logging.info("Falling back to Mistral model due to Phi-3 error")
                 fallback_model = "mistralai/Mistral-7B-Instruct-v0.3"
-                yield from (str(response) for response in respond(message, history, fallback_model, temperature, num_calls, selected_docs))
+                yield from respond(message, history, fallback_model, temperature, num_calls, selected_docs)
             else:
                 yield f"An error occurred with the {model} model: {str(e)}. Please try again or select a different model."
         
@@ -632,28 +607,15 @@ Write a detailed and complete response that answers the following user question:
         
         logging.info("Finished generating response")
 
-def transcribe_audio(audio_file):
+def transcribe(audio_file):
     if audio_file is None:
         return ""
     
-    if isinstance(audio_file, list):
-        if len(audio_file) == 0:
-            return ""
-        audio_file = audio_file[0]  # Take the first element if it's a list
-    
-    if not isinstance(audio_file, (str, bytes, os.PathLike)):
-        logging.error(f"Unexpected audio_file type: {type(audio_file)}")
-        return f"Error: Unexpected audio file type {type(audio_file)}"
+    with open(audio_file, "rb") as f:
+        audio_data = f.read()
     
-    try:
-        with open(audio_file, "rb") as f:
-            data = f.read()
-        
-        response = whisper_client.audio_to_text(data)
-        return response["text"]
-    except Exception as e:
-        logging.error(f"Error transcribing audio: {str(e)}")
-        return f"Error transcribing audio: {str(e)}"
+    response = whisper_api(audio_data)
+    return response["text"]
 
 def vote(data: gr.LikeData):
     if data.liked:
@@ -704,42 +666,74 @@ use_web_search = gr.Checkbox(label="Use Web Search", value=False)
 
 custom_placeholder = "Ask a question (Note: You can toggle between Web Search and PDF Chat in Additional Inputs below)"
 
-# Update the demo interface
-# Update the Gradio interface
-demo = gr.ChatInterface(
-    fn=chatbot_interface,
-    additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=True, render=False),
-    additional_inputs=[
-        gr.Dropdown(choices=MODELS, label="Select Model", value=MODELS[3]),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.2, step=0.1, label="Temperature"),
-        gr.Slider(minimum=1, maximum=5, value=1, step=1, label="Number of API Calls"),
-        gr.Checkbox(label="Use Web Search", value=True),
-        gr.CheckboxGroup(label="Select documents to query"),
-        gr.Audio(sources="microphone", type="filepath", label="Or speak your question (optional)")
+asr_interface = gr.Interface(
+    fn=transcribe,
+    inputs=[
+        gr.Audio(sources="microphone", type="filepath", optional=True)
     ],
-    title="AI-powered PDF Chat and Web Search Assistant",
-    description="Chat with your PDFs, use web search, or speak your questions.",
-    theme=gr.themes.Soft(
-        # ... (keep the existing theme configuration)
-    ),
-    css=css,
-    examples=[
-        # ... (keep the existing examples)
-    ],
-    cache_examples=False,
-    analytics_enabled=False,
-    textbox=gr.Textbox(placeholder="Type or speak your question", container=False, scale=7),
-    chatbot = gr.Chatbot(  
-        show_copy_button=True,
-        likeable=True,
-        layout="bubble",
-        height=400,
-        value=initial_conversation()
-    )
+    outputs="text",
+    title="Speech to Text",
+    description="Speak your query, and it will be transcribed."
 )
 
-# Add file upload functionality
-with demo:
+# Update the Gradio interface
+with gr.Blocks() as demo:
+    gr.Markdown("# AI-powered PDF Chat and Web Search Assistant with Speech Input")
+    
+    with gr.Row():
+        with gr.Column(scale=1):
+            asr_interface.render()
+        
+        with gr.Column(scale=2):
+            chatbot = gr.ChatInterface(
+                respond,
+                additional_inputs=[
+                    gr.Dropdown(choices=MODELS, label="Select Model", value=MODELS[3]),
+                    gr.Slider(minimum=0.1, maximum=1.0, value=0.2, step=0.1, label="Temperature"),
+                    gr.Slider(minimum=1, maximum=5, value=1, step=1, label="Number of API Calls"),
+                    gr.Checkbox(label="Use Web Search", value=True),
+                    gr.CheckboxGroup(label="Select documents to query")
+                ],
+                title="Chat with PDFs or Web Search",
+                description="Chat with your PDFs or use web search to answer questions.",
+                theme=gr.themes.Soft(
+                    primary_hue="orange",
+                    secondary_hue="amber",
+                    neutral_hue="gray",
+                    font=[gr.themes.GoogleFont("Exo"), "ui-sans-serif", "system-ui", "sans-serif"]
+                ).set(
+                    body_background_fill_dark="#0c0505",
+                    block_background_fill_dark="#0c0505",
+                    block_border_width="1px",
+                    block_title_background_fill_dark="#1b0f0f",
+                    input_background_fill_dark="#140b0b",
+                    button_secondary_background_fill_dark="#140b0b",
+                    border_color_accent_dark="#1b0f0f",
+                    border_color_primary_dark="#1b0f0f",
+                    background_fill_secondary_dark="#0c0505",
+                    color_accent_soft_dark="transparent",
+                    code_background_fill_dark="#140b0b"
+                ),
+                css=css,
+                examples=[
+                    ["Tell me about the contents of the uploaded PDFs."],
+                    ["What are the main topics discussed in the documents?"],
+                    ["Can you summarize the key points from the PDFs?"],
+                    ["What's the latest news about artificial intelligence?"]
+                ],
+                cache_examples=False,
+                analytics_enabled=False,
+                textbox=gr.Textbox(placeholder="Ask a question about the uploaded PDFs or any topic", container=False, scale=7),
+                chatbot=gr.Chatbot(
+                    show_copy_button=True,
+                    likeable=True,
+                    layout="bubble",
+                    height=400,
+                    value=initial_conversation()
+                )
+            )
+    
+    # Add file upload functionality
     gr.Markdown("## Upload and Manage PDF Documents")
     with gr.Row():
         file_input = gr.Files(label="Upload your PDF documents", file_types=[".pdf"])
@@ -754,35 +748,35 @@ with demo:
     update_button.click(
         update_vectors, 
         inputs=[file_input, parser_dropdown], 
-        outputs=[update_output, demo.additional_inputs[-1]]  # Use the CheckboxGroup from additional_inputs
+        outputs=[update_output, chatbot.additional_inputs[-1]]
     )
     
     # Add the refresh button functionality
     refresh_button.click(
         refresh_documents, 
         inputs=[], 
-        outputs=[demo.additional_inputs[-1]]  # Use the CheckboxGroup from additional_inputs
+        outputs=[chatbot.additional_inputs[-1]]
     )
     
     # Add the delete button functionality
     delete_button.click(
         delete_documents,
-        inputs=[demo.additional_inputs[-1]],  # Use the CheckboxGroup from additional_inputs
-        outputs=[update_output, demo.additional_inputs[-1]]
+        inputs=[chatbot.additional_inputs[-1]],
+        outputs=[update_output, chatbot.additional_inputs[-1]]
     )
 
     gr.Markdown(
     """
     ## How to use
-    1. Upload PDF documents using the file input at the top.
-    2. Select the PDF parser (pypdf or llamaparse) and click "Upload Document" to update the vector store.
-    3. Select the documents you want to query using the checkboxes.
-    4. Ask questions in the chat interface. 
-    5. Toggle "Use Web Search" to switch between PDF chat and web search.
-    6. Adjust Temperature and Number of API Calls to fine-tune the response generation.
-    7. Use the provided examples or ask your own questions.
+    1. Use the microphone to speak your query, or type it in the chat interface.
+    2. Upload PDF documents using the file input at the bottom.
+    3. Select the PDF parser (pypdf or llamaparse) and click "Upload Document" to update the vector store.
+    4. Select the documents you want to query using the checkboxes.
+    5. Ask questions in the chat interface. 
+    6. Toggle "Use Web Search" to switch between PDF chat and web search.
+    7. Adjust Temperature and Number of API Calls to fine-tune the response generation.
+    8. Use the provided examples or ask your own questions.
     """
     )
-
 if __name__ == "__main__":
     demo.launch(share=True)