Update app.py

app.py

@@ -17,8 +17,6 @@ from huggingface_hub import InferenceClient
 import inspect
 import logging
 import shutil
-import numpy as np
-import soundfile as sf
 
 
 # Set up basic configuration for logging
@@ -30,6 +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/"
+whisper_api = InferenceClient("openai/whisper-small", token=huggingface_token)
 
 print(f"ACCOUNT_ID: {ACCOUNT_ID}")
 print(f"CLOUDFLARE_AUTH_TOKEN: {API_TOKEN[:5]}..." if API_TOKEN else "Not set")
@@ -399,20 +398,13 @@ def summarize_web_results(query: str, search_results: List[Dict[str, str]], conv
         return f"An error occurred during summarization: {str(e)}"
 
 # Modify the existing respond function to handle both PDF and web search
-def respond(message, history, model, temperature, num_calls, use_web_search, selected_docs, audio_input):
-    if audio_input:
-        message = transcribe_audio(audio_input)
-        logging.info(f"Transcribed audio: {message}")
-    
-    if not message.strip():
-        return "Please provide a text or audio query.", history
-
+def respond(message, history, model, temperature, num_calls, use_web_search, selected_docs):
     logging.info(f"User Query: {message}")
     logging.info(f"Model Used: {model}")
     logging.info(f"Selected Documents: {selected_docs}")
     logging.info(f"Use Web Search: {use_web_search}")
 
-    history = history + [(message, "")]
+    response = ""
 
     if use_web_search:
         original_query = message
@@ -433,9 +425,10 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
 
         if final_summary:
             conversation_manager.add_interaction(original_query, final_summary)
-            yield final_summary
+            response = final_summary
         else:
-            yield "Unable to generate a response. Please try a different query."
+            response = "Unable to generate a response. Please try a different query."
+
     else:
         # Existing PDF search logic
         try:
@@ -448,45 +441,41 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
                 relevant_docs = [doc for doc in all_relevant_docs if doc.metadata["source"] in selected_docs]
                 
                 if not relevant_docs:
-                    yield "No relevant information found in the selected documents. Please try selecting different documents or rephrasing your query."
-                    return
-    
-                context_str = "\n".join([doc.page_content for doc in relevant_docs])
-                logging.info(f"Context length: {len(context_str)}")
-            else:
-                context_str = "No documents available."
-                yield "No documents available. Please upload PDF documents to answer questions."
-                return
-            
-            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 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 partial_response
+                    response = "No relevant information found in the selected documents. Please try selecting different documents or rephrasing your query."
+                else:
+                    context_str = "\n".join([doc.page_content for doc in relevant_docs])
+                    logging.info(f"Context length: {len(context_str)}")
+
+                    if model.startswith("duckduckgo/"):
+                        # Use DuckDuckGo chat with context
+                        for partial_response in get_response_from_duckduckgo(message, model, context_str, num_calls, temperature):
+                            response += 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"):
+                            response += 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):
+                            response += 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 partial_response
+                response = "No documents available. Please upload PDF documents to answer questions."
+
         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 respond(message, history, fallback_model, temperature, num_calls, selected_docs)
+                return respond(message, history, fallback_model, temperature, num_calls, use_web_search, selected_docs)
             else:
-                yield f"An error occurred with the {model} model: {str(e)}. Please try again or select a different model
+                response = f"An error occurred with the {model} model: {str(e)}. Please try again or select a different model."
+
+    # Update the conversation history
+    history.append((message, response))
+    
+    # Yield the updated history
+    yield history
 
-            history[-1] = (message, response)
-            return response, history
-        except Exception as e:
-            logging.error(f"Error in respond: {str(e)}")
-            error_message = f"An error occurred: {str(e)}"
-            history[-1] = (message, error_message)
-            return error_message, history
-        
 logging.basicConfig(level=logging.DEBUG)
 
 def get_response_from_cloudflare(prompt, context, query, num_calls=3, temperature=0.2, search_type="pdf"):
@@ -625,23 +614,15 @@ Write a detailed and complete response that answers the following user question:
         
         logging.info("Finished generating response")
 
-def transcribe_audio(audio_file):
-    client = InferenceClient("openai/whisper-large-v3", token=huggingface_token)
-    
-    # Load the audio file
-    audio, sample_rate = sf.read(audio_file)
+def transcribe(audio_file):
+    if audio_file is None:
+        return ""
     
-    # Ensure audio is mono
-    if len(audio.shape) > 1:
-        audio = audio.mean(axis=1)
+    with open(audio_file, "rb") as f:
+        audio_data = f.read()
     
-    # Normalize audio
-    audio = (audio / np.max(np.abs(audio))) * 32767
-    audio = audio.astype(np.int16)
-    
-    # Transcribe
-    result = client.automatic_speech_recognition(audio, sampling_rate=sample_rate)
-    return result["text"]
+    response = whisper_api(audio_data)
+    return response["text"]
 
 def vote(data: gr.LikeData):
     if data.liked:
@@ -692,96 +673,40 @@ 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.Interface(
-    fn=respond,
-    inputs=[
-        gr.Textbox(placeholder=custom_placeholder, container=False, scale=7),
-        gr.State([]),  # for history
-        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")
-    ],
-    outputs=[gr.Chatbot()],
-    title="AI-powered PDF Chat and Web Search Assistant",
-    description="Chat with your PDFs or use web search to answer questions. You can type or speak your query.",
-    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,
-)
+def update_textbox(transcription):
+    return gr.Textbox.update(value=transcription)
 
-# Add file upload functionality
-# Add file upload functionality
+# Update the Gradio interface
 with gr.Blocks() as demo:
-    chatbot = gr.Chatbot(
-        show_copy_button=True,
-        likeable=True,
-        layout="bubble",
-        height=400,
-        value=initial_conversation()
-    )
-    state = gr.State([])
-
+    gr.Markdown("# AI-powered PDF Chat and Web Search Assistant with Speech Input")
+    
     with gr.Row():
-        text_input = gr.Textbox(
-            placeholder=custom_placeholder,
-            container=False,
-            scale=7
-        )
-        audio_input = gr.Audio(source="microphone", type="filepath")
-
+        with gr.Column(scale=1):
+            audio_input = gr.Audio(sources="microphone", type="filepath", label="Speak your query")
+            transcribe_button = gr.Button("Transcribe")
+        
+        with gr.Column(scale=2):
+            chatbot = gr.Chatbot(
+                show_copy_button=True,
+                likeable=True,
+                layout="bubble",
+                height=400,
+                value=initial_conversation()
+            )
+            query_textbox = gr.Textbox(
+                placeholder="Ask a question about the uploaded PDFs or any topic", 
+                container=False, 
+                scale=7
+            )
+            submit_button = gr.Button("Submit")
+    
     with gr.Accordion("⚙️ Parameters", open=False):
         model = gr.Dropdown(choices=MODELS, label="Select Model", value=MODELS[3])
         temperature = gr.Slider(minimum=0.1, maximum=1.0, value=0.2, step=0.1, label="Temperature")
         num_calls = gr.Slider(minimum=1, maximum=5, value=1, step=1, label="Number of API Calls")
         use_web_search = gr.Checkbox(label="Use Web Search", value=True)
-        selected_docs = gr.CheckboxGroup(label="Select documents to query")
-
-    submit_button = gr.Button("Submit")
-
-    submit_button.click(
-        fn=respond,
-        inputs=[
-            text_input,
-            state,
-            model,
-            temperature,
-            num_calls,
-            use_web_search,
-            selected_docs,
-            audio_input
-        ],
-        outputs=[chatbot, state]
-    )
-
+        document_selector = gr.CheckboxGroup(label="Select documents to query")
+    
     # Add file upload functionality
     gr.Markdown("## Upload and Manage PDF Documents")
     with gr.Row():
@@ -793,35 +718,50 @@ with gr.Blocks() as demo:
     update_output = gr.Textbox(label="Update Status")
     delete_button = gr.Button("Delete Selected Documents")
     
+    # Connect components
+    transcribe_button.click(
+        transcribe, 
+        inputs=[audio_input], 
+        outputs=[query_textbox]
+    )
+    
+    submit_button.click(
+        respond,
+        inputs=[query_textbox, chatbot, model, temperature, num_calls, use_web_search, document_selector],
+        outputs=[chatbot]
+    )
+    
     update_button.click(
         update_vectors, 
         inputs=[file_input, parser_dropdown], 
-        outputs=[update_output, selected_docs]
+        outputs=[update_output, document_selector]
     )
     
     refresh_button.click(
         refresh_documents, 
         inputs=[], 
-        outputs=[selected_docs]
+        outputs=[document_selector]
     )
     
     delete_button.click(
         delete_documents,
-        inputs=[selected_docs],
-        outputs=[update_output, selected_docs]
+        inputs=[document_selector],
+        outputs=[update_output, document_selector]
     )
 
     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 by typing in the text box or using the microphone for speech input.
-    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, then click "Transcribe", or type directly in the text box.
+    2. Click "Submit" to get a response from the AI.
+    3. Upload PDF documents using the file input at the bottom.
+    4. Select the PDF parser (pypdf or llamaparse) and click "Upload Document" to update the vector store.
+    5. Select the documents you want to query using the checkboxes.
+    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.
     """
     )
+
 if __name__ == "__main__":
-    demo.launch(share=True)
+    demo.launch(share=True)
+Troubleshooting Python Audio Recording Issues - Claude