app.py

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

# Variable to store the API key
api_key = ""

# Function to update the 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
        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."

    # Extract text from the first few pages
    doc = fitz.open(pdf_file)
    text = "\n".join([page.get_text("text") for page in doc][:5])  # Limit extraction for performance

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

# Function to clear the chat (Resets input and output)
def clear_chat():
    return "", "", "", "", "", "", "", 1.0, 1.0, 1024

# Gradio UI Layout
with gr.Blocks() as demo:
    gr.Markdown("## GPT-4.5 Preview Chatbot")
    
    # 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")
        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=1024, step=512, label="Max Output Tokens")
    
    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")
        
        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")
        
        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")
        
        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")

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

    # 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)

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

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