import openai
import base64
import cv2
import gradio as gr
# import os

# api_key = os.getenv("OPENAI_API_KEY")
# openai.api_key = api_key
# Function to encode a numpy array image
def encode_image(image_array):
    _, buffer = cv2.imencode('.jpg', image_array)
    return base64.b64encode(buffer).decode("utf-8")

    
def respond_with_audio(image_array, text_prompt, voice='alloy'):
    # Encode the image
    base64_image = encode_image(image_array)

    # Step 1: Get response from GPT-4o
    response = openai.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": "Describe and respond to the user's input."},
            {"role": "user", "content": [
                {"type": "text", "text": text_prompt},
                {"type": "image_url", "image_url": {
                    "url": f"data:image/jpeg;base64,{base64_image}",
                }},
            ]}
        ],
        max_tokens=300
    )

    message_text = response.choices[0].message.content

    # Step 2: Convert text to speech
    tts_response = openai.audio.speech.create(
        model="tts-1",
        voice=voice,  # e.g., 'alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'
        input=message_text
    )
    tts_response_ = tts_response.content
    return message_text, tts_response_

with gr.Blocks() as block:
    with gr.Row():
        gr.HTML(
            f"""
            <h1 style='text-align: center;'> Kid's Bot 👼</h1>
            <h3 style='text-align: center;'> Hint: change the question </h3>
            <p style='text-align: center;'> Created by <a href="https://github.com/pacomesimon"> Pacome</a>
            """
        )
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(label="Capture Image", sources=["webcam"], type="numpy")
            examples = gr.Examples(
                    examples=["small_image.jpg"],
                    inputs=[image_input],
                    )
            text_prompt = gr.Textbox(label="Enter your question",
                                    value="what is this image showing? explain it to a 5 years old kid", lines=1)
        with gr.Column():
            # audio_out = gr.Audio(label="Spoken Answer", streaming=True, autoplay=True, loop=False)
            audio_out = gr.Audio(label="Spoken Answer", autoplay=True)

            answer = gr.Textbox(label="Answer")
    
    def process_input(image, text):
        # print(f"Processing input Q: {text}")
        message_text, tts_content = respond_with_audio(image, text)
        # print(f"Answer: {message_text}")
        # print(f"Audio: {[tts_content]}")
        return message_text, tts_content

    image_input.change(fn=process_input, inputs=[image_input, text_prompt], outputs=[answer, audio_out])

block.launch(
    debug=True,
    share=True,
)