app.py

import gradio as gr
from PIL import Image
import requests
import os
from together import Together
import base64
from threading import Thread
import time
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, max_size=(800, 800), quality=85):
    with Image.open(image_path) as img:
        img.thumbnail(max_size)
        if img.mode in ('RGBA', 'LA'):
            background = Image.new(img.mode[:-1], img.size, (255, 255, 255))
            background.paste(img, mask=img.split()[-1])
            img = background
        buffered = io.BytesIO()
        img.save(buffered, format="JPEG", quality=quality)
        return base64.b64encode(buffered.getvalue()).decode('utf-8')

def bot_streaming(message, history, together_api_key, max_new_tokens=250, max_history=5):
    if client is None:
        initialize_client(together_api_key)

    txt = message["text"]
    messages = []
    images = []

    for i, msg in enumerate(history[-max_history:]):
        if isinstance(msg[0], tuple):
            messages.append({"role": "user", "content": [{"type": "text", "text": history[i+1][0]}, {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{encode_image(msg[0][0])}"}}]})
            messages.append({"role": "assistant", "content": [{"type": "text", "text": history[i+1][1]}]})
        elif isinstance(history[i-1], tuple) and isinstance(msg[0], str):
            pass
        elif isinstance(history[i-1][0], str) and isinstance(msg[0], str):
            messages.append({"role": "user", "content": [{"type": "text", "text": msg[0]}]})
            messages.append({"role": "assistant", "content": [{"type": "text", "text": msg[1]}]})

    if len(message["files"]) == 1:
        if isinstance(message["files"][0], str):  # examples
            image_path = message["files"][0]
        else:  # regular input
            image_path = message["files"][0]["path"]
        messages.append({"role": "user", "content": [{"type": "text", "text": txt}, {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{encode_image(image_path)}"}}]})
    else:
        messages.append({"role": "user", "content": [{"type": "text", "text": txt}]})

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

        buffer = ""
        for chunk in stream:
            if chunk.choices[0].delta.content is not None:
                buffer += chunk.choices[0].delta.content
                time.sleep(0.01)
                yield buffer

    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](https://api.together.xyz/settings/api-keys) and get started!")
    
    together_api_key = gr.Textbox(
        label="Together API Key",
        placeholder="Enter your TOGETHER_API_KEY here",
        type="password"
    )
    
    chatbot = gr.ChatInterface(
        fn=bot_streaming,
        textbox=gr.MultimodalTextbox(),
        additional_inputs=[
            gr.Slider(
                minimum=10,
                maximum=500,
                value=250,
                step=10,
                label="Maximum number of new tokens to generate",
            )
        ],
        cache_examples=False,
        stop_btn="Stop Generation",
        fill_height=True,
        multimodal=True
    )
    
    together_api_key.change(lambda x: x, inputs=[together_api_key], outputs=[chatbot.additional_inputs[0]])

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