Added audio input.

app.py

@@ -7,6 +7,9 @@ For more information on `huggingface_hub` Inference API support, please check th
 client = InferenceClient("meta-llama/Llama-3.2-3B-Instruct") 
 # client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
+# Create a separate client for audio transcription
+audio_client = InferenceClient("openai/whisper-large-v3")
+
 _sys_msg = """
 You are a reporter writing a biographical article about your interviewee and you only ask one question at a time and let the user answer. Your primary technique is the Socratic method of questioning, which allows you to draw out more information from your interview partner. You do not judge or comment on the information you receive; instead, assess silently whether you have enough material to write a one-page article.
 
@@ -28,13 +31,23 @@ Welcome to the interview. I want to write a short biography about you and need s
 Can you please start by stating your name and talking about your early childhood?
 """
 
+def transcribe(audio):
+    if audio is None:
+        return None
+    text = audio_client.audio_transcription(audio)
+    return text
+
 def respond(
     message,
     history: list[tuple[str, str]],
-    max_tokens,
-    temperature,
-    top_p,
+    audio,
 ):
+    # Handle voice input if provided
+    if audio is not None:
+        transcribed_text = transcribe(audio)
+        if transcribed_text:
+            message = transcribed_text
+
     messages = [{"role": "system", "content": _sys_msg},
                 {"role": "assistant", "content": _ass_msg_start}]
     
@@ -44,7 +57,6 @@ def respond(
         if val[1]:
             messages.append({"role": "assistant", "content": val[1]})
 
-
     messages.append({"role": "user", "content": message})
 
     response = ""
@@ -57,24 +69,25 @@ def respond(
         top_p=0.95,      # Fixed value
     ):
         token = message.choices[0].delta.content
-
         response += token
         yield response
 
-
-demo  = gr.ChatInterface(
+demo = gr.ChatInterface(
     respond,
-    title = "AI Biographical Interview Assistant",
-    description = """
+    additional_inputs=[
+        gr.Audio(source="microphone", type="filepath", label="Voice Input", interactive=True)
+    ],
+    title="AI Biographical Interview Assistant",
+    description="""
     Welcome to the AI Biographical Interview Assistant! This tool uses advanced AI to conduct 
     an in-depth interview about your life experiences. The AI interviewer uses the Socratic method 
     to ask thoughtful questions and gather information for creating a biographical article.
     
     Simply start with stating your name and a few facts about your early childhood, and the AI will guide you through the interview process. 
+    You can either type your responses or use the microphone button to speak them.
     When finished ask for the final article and the AI will give wrap it up and give you a summary.
     """
-    )
-
+)
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()