Update app.py

app.py

@@ -17,6 +17,11 @@ from huggingface_hub import InferenceClient
 import inspect
 import logging
 import shutil
+import soundfile as sf
+from transformers import WhisperProcessor, WhisperForConditionalGeneration
+from transformers import pipeline
+import numpy as np
+
 
 
 # Set up basic configuration for logging
@@ -28,6 +33,7 @@ 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_URL = "https://api-inference.huggingface.co/models/openai/whisper-large-v3"
 
 print(f"ACCOUNT_ID: {ACCOUNT_ID}")
 print(f"CLOUDFLARE_AUTH_TOKEN: {API_TOKEN[:5]}..." if API_TOKEN else "Not set")
@@ -397,7 +403,16 @@ 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):
+def respond(message, history, model, temperature, num_calls, use_web_search, selected_docs, audio_input):
+    if audio_input is not None:
+        # If audio input is provided, transcribe it
+        try:
+            message = transcribe(audio_input)
+            print(f"Transcribed audio: {message}")
+        except Exception as e:
+            print(f"Error transcribing audio: {str(e)}")
+            return f"Error transcribing audio: {str(e)}", history
+    
     logging.info(f"User Query: {message}")
     logging.info(f"Model Used: {model}")
     logging.info(f"Selected Documents: {selected_docs}")
@@ -606,6 +621,13 @@ Write a detailed and complete response that answers the following user question:
         
         logging.info("Finished generating response")
 
+def transcribe(audio_path):
+    headers = {"Authorization": f"Bearer {huggingface_token}"}
+    with open(audio_path, "rb") as f:
+        data = f.read()
+    response = requests.post(WHISPER_API_URL, headers=headers, data=data)
+    return response.json().get("text", "")
+
 def vote(data: gr.LikeData):
     if data.liked:
         print(f"You upvoted this response: {data.value}")
@@ -659,16 +681,16 @@ custom_placeholder = "Ask a question (Note: You can toggle between Web Search an
 # Update the Gradio interface
 demo = gr.ChatInterface(
     respond,
-    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.CheckboxGroup(label="Select documents to query"),
+        gr.Audio(source="microphone", type="filepath", label="Voice Input")
     ],
     title="AI-powered PDF Chat and Web Search Assistant",
-    description="Chat with your PDFs or use web search to answer questions.",
+    description="Chat with your PDFs, use web search to answer questions, or speak your query.",
     theme=gr.themes.Soft(
         primary_hue="orange",
         secondary_hue="amber",