app.py

import gradio as gr
from PIL import Image
import requests
import os
from together import Together
import base64
import io

# Initialize Together client
client = None

def initialize_client(api_key=None):
    global client
    if api_key:
        os.environ["TOGETHER_API_KEY"] = api_key
    if "TOGETHER_API_KEY" in os.environ:
        client = Together()
    else:
        raise ValueError("Please provide a Together API Key")

def encode_image(image_path):
    with Image.open(image_path) as img:
        buffered = io.BytesIO()
        img.save(buffered, format="PNG")
        return base64.b64encode(buffered.getvalue()).decode('utf-8')

def bot_streaming(message, history, together_api_key, max_new_tokens=250, temperature=0.7):
    if client is None:
        try:
            initialize_client(together_api_key)
        except Exception as e:
            yield f"Error initializing client: {str(e)}"
            return

    prompt = "You are a helpful AI assistant. Analyze the image provided (if any) and respond to the user's query or comment."
    
    messages = [{"role": "system", "content": prompt}]
    
    # Add history to messages
    for user_msg, assistant_msg in history:
        if isinstance(user_msg, tuple):  # Handle image messages
            text = user_msg[1] if len(user_msg) > 1 else ""
            messages.append({"role": "user", "content": [
                {"type": "text", "text": text},
                {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{encode_image(user_msg[0])}"}}
            ]})
        else:
            messages.append({"role": "user", "content": [{"type": "text", "text": user_msg}]})
        messages.append({"role": "assistant", "content": [{"type": "text", "text": assistant_msg}]})
    
    # Prepare the current message
    current_message = {"role": "user", "content": []}
    
    # Add text content
    if isinstance(message, dict) and message.get("text"):
        current_message["content"].append({"type": "text", "text": message["text"]})
    elif isinstance(message, str):
        current_message["content"].append({"type": "text", "text": message})
    
    # Add image content if present
    if isinstance(message, dict) and message.get("files") and len(message["files"]) > 0:
        image_path = message["files"][0]["path"] if isinstance(message["files"][0], dict) else message["files"][0]
        image_base64 = encode_image(image_path)
        current_message["content"].append({
            "type": "image_url",
            "image_url": {"url": f"data:image/png;base64,{image_base64}"}
        })
    
    messages.append(current_message)

    try:
        stream = client.chat.completions.create(
            model="meta-llama/Llama-Vision-Free",
            messages=messages,
            max_tokens=max_new_tokens,
            temperature=temperature,
            stream=True,
        )

        response = ""
        for chunk in stream:
            response += chunk.choices[0].delta.content or ""
            yield response

        if not response:
            yield "No response generated. Please try again."

    except Exception as e:
        if "Request Entity Too Large" in str(e):
            yield "The image is too large. Please try with a smaller image or compress the existing one."
        else:
            yield f"An error occurred: {str(e)}"

with gr.Blocks() as demo:
    gr.Markdown("# Meta Llama-3.2-11B-Vision-Instruct (FREE)")
    gr.Markdown("Try the new Llama 3.2 11B Vision API by Meta for free through Together AI. Upload an image, and start chatting about it. Just paste in your Together AI API key and get started!")
    
    with gr.Row():
        together_api_key = gr.Textbox(
            label="Together API Key",
            placeholder="Enter your TOGETHER_API_KEY here",
            type="password"
        )
    
    with gr.Row():
        max_new_tokens = gr.Slider(
            minimum=10,
            maximum=500,
            value=250,
            step=10,
            label="Maximum number of new tokens",
        )
        temperature = gr.Number(
            value=0.7,
            minimum=0,
            maximum=1,
            step=0.1,
            label="Temperature"
        )
    
    chatbot = gr.Chatbot()
    msg = gr.MultimodalTextbox(label="Enter text or upload an image")
    clear = gr.Button("Clear")

    msg.submit(bot_streaming, [msg, chatbot, together_api_key, max_new_tokens, temperature], chatbot)
    clear.click(lambda: None, None, chatbot, queue=False)

if __name__ == "__main__":
    demo.launch(debug=True)