app.py

# imports
import os
import base64
import gradio as gr
from openai import OpenAI

# Initialization
openai_api_key = os.getenv('OPENAI_API_KEY')

if openai_api_key:
    print(f"OpenAI API Key exists and begins {openai_api_key[:8]}")
else:
    print("OpenAI API Key not set")

MODEL = "gpt-4o-mini"
openai = OpenAI()

system_message = "You are a language tutor, and as such provide only with \
    helpful tips and accurate translations. You are entertaining and polite. \
    If you don't know something, you say so."


def talker(message):
    response = openai.audio.speech.create(
        model="tts-1",
        voice="nova",  # Can I change the vibe parameter?
        input=message
    )

    # Convert audio content to base64 for embedding in HTML
    audio_base64 = base64.b64encode(response.content).decode()
    audio_html = f'<audio autoplay controls src="data:audio/mp3;base64,{audio_base64}"></audio>'

    return message, audio_html  # Returning text + HTML audio (no separate "Press Play" button)


# Transcription function
def transcribe_audio(audio_file):
    # Ensure the audio file is opened as a binary file
    with open(audio_file, "rb") as audio:
        translation = openai.audio.translations.create(
            model="whisper-1",
            file=audio  # Pass the opened file, not the filepath
        )

    print(translation.text)
    return translation.text


# Wrapper function to combine microphone input, transcription, and chat
def process_microphone_input(audio, history=[]):
    if audio is None:
        raise ValueError("No audio input detected. Please ensure the microphone is functioning correctly.")
    # Step 1: Transcribe the audio captured from the microphone
    transcribed_text = transcribe_audio(audio)

    # Step 2: Pass the transcription to the chat function
    response = chat(transcribed_text, history)

    return response


def chat(message, history):
    messages = [{"role": "system", "content": system_message}] + history + [{"role": "user", "content": message}]
    response = openai.chat.completions.create(model=MODEL, messages=messages)
    reply = response.choices[0].message.content

    print(f"History: {history}")
    print(f"Message: {message}")
    print(f"Messages: {messages}")

    return talker(reply)


# Gradio interface for microphone input
interface = gr.Interface(
    fn=process_microphone_input,
    inputs=[gr.Audio(sources="microphone", type="filepath")],  # Microphone as input
    outputs=["text", "html"],  # Keep text + autoplaying HTML audio
    title="Speech-to-Chatbot-to-Speech Language Tutor",
    description="Speak into the microphone to chat with GPT-4. Wait a couple of seconds before you submit your message."
)

if __name__ == "__main__":
    interface.launch()