app.py

import gradio as gr
import openai
import fitz  # PyMuPDF for PDF processing
import base64
import io

# Variable to store API key
api_key = ""

# Function to update API key
def set_api_key(key):
    global api_key
    api_key = key
    return "API Key Set Successfully!"

# Function to interact with OpenAI API
def query_openai(messages, temperature, top_p, max_output_tokens):
    if not api_key:
        return "Please enter your OpenAI API key first."

    try:
        openai.api_key = api_key  # Set API key dynamically

        # Ensure numeric values for OpenAI parameters
        temperature = float(temperature) if temperature else 1.0
        top_p = float(top_p) if top_p else 1.0
        max_output_tokens = int(max_output_tokens) if max_output_tokens else 2048

        response = openai.ChatCompletion.create(
            model="gpt-4.5-preview",
            messages=messages,
            temperature=temperature,
            top_p=top_p,
            max_tokens=max_output_tokens
        )
        return response["choices"][0]["message"]["content"]
    except Exception as e:
        return f"Error: {str(e)}"

# Function to process image URL input
def image_url_chat(image_url, text_query, temperature, top_p, max_output_tokens):
    if not image_url or not text_query:
        return "Please provide an image URL and a query."

    messages = [
        {"role": "user", "content": [
            {"type": "image_url", "image_url": {"url": image_url}},  # Corrected format
            {"type": "text", "text": text_query}
        ]},
    ]
    return query_openai(messages, temperature, top_p, max_output_tokens)

# Function to process text input
def text_chat(text_query, temperature, top_p, max_output_tokens):
    if not text_query:
        return "Please enter a query."

    messages = [{"role": "user", "content": [{"type": "text", "text": text_query}]}]
    return query_openai(messages, temperature, top_p, max_output_tokens)

# Function to process uploaded image input
def image_chat(image_file, text_query, temperature, top_p, max_output_tokens):
    if image_file is None or not text_query:
        return "Please upload an image and provide a query."

    # Encode image as base64
    with open(image_file, "rb") as img:
        base64_image = base64.b64encode(img.read()).decode("utf-8")

    image_data = f"data:image/jpeg;base64,{base64_image}"

    messages = [
        {"role": "user", "content": [
            {"type": "image_url", "image_url": {"url": image_data}},  # Fixed format
            {"type": "text", "text": text_query}
        ]},
    ]
    return query_openai(messages, temperature, top_p, max_output_tokens)

# Function to process uploaded PDF input
def pdf_chat(pdf_file, text_query, temperature, top_p, max_output_tokens):
    if pdf_file is None or not text_query:
        return "Please upload a PDF and provide a query."

    try:
        # Extract text from all pages of the PDF
        doc = fitz.open(pdf_file.name)
        text = "\n".join([page.get_text("text") for page in doc])  # Extract text from all pages

        # If no text found, return an error
        if not text.strip():
            return "No text found in the PDF."

        # Create the query message with the extracted text and the user's query
        messages = [
            {"role": "user", "content": [
                {"type": "text", "text": text},  # The extracted text from the PDF
                {"type": "text", "text": text_query}
            ]},
        ]
        return query_openai(messages, temperature, top_p, max_output_tokens)
    
    except Exception as e:
        return f"Error processing the PDF: {str(e)}"

# Function to transcribe audio to text using OpenAI Whisper API
def transcribe_audio(audio_binary, openai_api_key):
    if not openai_api_key:
        return "Error: No API key provided."
    
    openai.api_key = openai_api_key
    
    try:
        # Use the correct transcription API call
        audio_file_obj = io.BytesIO(audio_binary)
        audio_file_obj.name = 'audio.wav'  # Set a name for the file object (as OpenAI expects it)

        # Transcribe the audio to text using OpenAI's whisper model
        audio_file_transcription = openai.Audio.transcribe(file=audio_file_obj, model="whisper-1")
        return audio_file_transcription.text
    except Exception as e:
        return f"Error transcribing audio: {str(e)}"

# Function to clear the chat (Fix: Returns the correct number of outputs)
def clear_chat():
    return "", "", "", "", "", "", "", None, "", None, "", 1.0, 1.0, 2048

# Gradio UI Layout
with gr.Blocks() as demo:
    gr.Markdown("## GPT-4.5 Preview Chatbot")

    # Accordion for explaining hyperparameters
    with gr.Accordion("Hyperparameters", open=False):
        gr.Markdown("""
        ### Temperature: 
        Controls the randomness of the model's output. A lower temperature makes the model more deterministic, while a higher temperature makes it more creative and varied.
        ### Top-P (Nucleus Sampling): 
        Controls the cumulative probability distribution from which the model picks the next word. A lower value makes the model more focused and deterministic, while a higher value increases randomness.
        ### Max Output Tokens: 
        Limits the number of tokens (words or subwords) the model can generate in its response. You can use this to control the length of the response.
        """)

    gr.HTML("""
    <style>
        #api_key_button {
            margin-top: 27px; /* Add margin-top to the button */
            background: linear-gradient(135deg, #4a00e0 0%, #8e2de2 100%); /* Purple gradient */
        }
        #api_key_button:hover {
            background: linear-gradient(135deg, #5b10f1 0%, #9f3ef3 100%); /* Slightly lighter */
        }
        #clear_chat_button {
            background: linear-gradient(135deg, #e53e3e 0%, #f56565 100%); /* Red gradient */
        }
        #clear_chat_button:hover {
            background: linear-gradient(135deg, #c53030 0%, #e53e3e 100%); /* Slightly darker red gradient on hover */
        }
        #ask_button {
            background: linear-gradient(135deg, #fbd38d 0%, #f6e05e 100%); /* Yellow gradient */
        }
        #ask_button:hover {
            background: linear-gradient(135deg, #ecc94b 0%, #fbd38d 100%); /* Slightly darker yellow gradient on hover */
        }
    </style>
    """)
    
    # API Key Input
    with gr.Row():
        api_key_input = gr.Textbox(label="Enter OpenAI API Key", type="password")
        api_key_button = gr.Button("Set API Key", elem_id="api_key_button")
        api_key_output = gr.Textbox(label="API Key Status", interactive=False)

    with gr.Row():
        temperature = gr.Slider(0, 2, value=1.0, step=0.1, label="Temperature")
        top_p = gr.Slider(0, 1, value=1.0, step=0.1, label="Top-P")
        max_output_tokens = gr.Slider(0, 16384, value=2048, step=512, label="Max Output Tokens")  # Changed default to 2048
    
    with gr.Tabs():
        with gr.Tab("Image URL Chat"):
            image_url = gr.Textbox(label="Enter Image URL")
            image_query = gr.Textbox(label="Ask about the Image")
            image_url_output = gr.Textbox(label="Response", interactive=False)
            image_url_button = gr.Button("Ask",elem_id="ask_button")
        
        with gr.Tab("Text Chat"):
            text_query = gr.Textbox(label="Enter your query")
            text_output = gr.Textbox(label="Response", interactive=False)
            text_button = gr.Button("Ask",elem_id="ask_button")
        
        with gr.Tab("Image Chat"):
            image_upload = gr.File(label="Upload an Image", type="filepath")
            image_text_query = gr.Textbox(label="Ask about the uploaded image")
            image_output = gr.Textbox(label="Response", interactive=False)
            image_button = gr.Button("Ask",elem_id="ask_button")
        
        with gr.Tab("PDF Chat"):
            pdf_upload = gr.File(label="Upload a PDF", type="filepath")
            pdf_text_query = gr.Textbox(label="Ask about the uploaded PDF")
            pdf_output = gr.Textbox(label="Response", interactive=False)
            pdf_button = gr.Button("Ask",elem_id="ask_button")

        with gr.Tab("Voice Chat"):
            audio_upload = gr.File(label="Upload an Audio File", type="binary")
            audio_query = gr.Textbox(label="Ask about the transcription")
            audio_output = gr.Textbox(label="Response", interactive=False)
            audio_button = gr.Button("Ask",elem_id="ask_button")

    # Clear chat button
    clear_button = gr.Button("Clear Chat",elem_id="clear_chat_button")

    # Button Click Actions
    api_key_button.click(set_api_key, inputs=[api_key_input], outputs=[api_key_output])
    image_url_button.click(image_url_chat, [image_url, image_query, temperature, top_p, max_output_tokens], image_url_output)
    text_button.click(text_chat, [text_query, temperature, top_p, max_output_tokens], text_output)
    image_button.click(image_chat, [image_upload, image_text_query, temperature, top_p, max_output_tokens], image_output)
    pdf_button.click(pdf_chat, [pdf_upload, pdf_text_query, temperature, top_p, max_output_tokens], pdf_output)
    
    # For Voice Chat
    audio_button.click(
        lambda audio_binary, query, temperature, top_p, max_output_tokens: query_openai(
            [{"role": "user", "content": [{"type": "text", "text": transcribe_audio(audio_binary, api_key)}, {"type": "text", "text": query}]}],
            temperature, top_p, max_output_tokens
        ), [audio_upload, audio_query, temperature, top_p, max_output_tokens], audio_output
    )

    # Fix: Clear button resets all necessary fields correctly
    clear_button.click(
        clear_chat,
        outputs=[
            image_url, image_query, image_url_output, 
            text_query, text_output, 
            image_text_query, image_output, 
            pdf_upload, pdf_text_query, pdf_output, 
            temperature, top_p, max_output_tokens
        ]
    )

# Launch Gradio App
if __name__ == "__main__":
    demo.launch()