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

# Initialize Together client
client = Together()

# Ensure API key is set
if "TOGETHER_API_KEY" not in os.environ:
    raise ValueError("Please set the TOGETHER_API_KEY environment variable")

def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode('utf-8')

def bot_streaming(message, history, max_new_tokens=250):
    txt = message["text"]
    messages = []
    images = []

    for i, msg in enumerate(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}]})

    stream = client.chat.completions.create(
        model="meta-llama/Llama-3.2-11B-Vision-Instruct-Turbo",
        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

demo = gr.ChatInterface(
    fn=bot_streaming,
    title="Multimodal Llama",
    examples=[
        [{"text": "Which era does this piece belong to? Give details about the era.", "files":["./examples/rococo.jpg"]}, 200],
        [{"text": "Where do the droughts happen according to this diagram?", "files":["./examples/weather_events.png"]}, 250],
        [{"text": "What happens when you take out white cat from this chain?", "files":["./examples/ai2d_test.jpg"]}, 250],
        [{"text": "Which company was this invoice addressed to?", "files":["./examples/invoice.png"]}, 250],
        [{"text": "Where to find this monument? Can you give me other recommendations around the area?", "files":["./examples/wat_arun.jpg"]}, 250],
    ],
    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,
    description="Try Multimodal Llama by Meta with the Together API in this demo. Upload an image, and start chatting about it, or simply try one of the examples below.",
    stop_btn="Stop Generation",
    fill_height=True,
    multimodal=True
)

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