app.py

import gradio as gr
from huggingface_hub import InferenceClient
import base64

client = InferenceClient("openbmb/MiniCPM-Llama3-V-2_5-int4",trust_remote_code=True)

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

def respond(
    message,
    image,
    history,
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]
    for user_msg, bot_msg in history:
        messages.append({"role": "user", "content": user_msg})
        messages.append({"role": "assistant", "content": bot_msg})
    
    if image:
        base64_image = encode_image(image)
        image_message = f"<image>{base64_image}</image>"
        message = image_message + "\n" + message

    messages.append({"role": "user", "content": message})
    
    response = ""
    for message in client.text_generation(
        prompt=f"{messages}",
        max_new_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.token.text
        response += token
        yield response, history + [(message, response)]

demo = gr.Interface(
    respond,
    inputs=[
        gr.Textbox(label="Message"),
        gr.Image(type="filepath", label="Upload Image"),
        gr.State([]),  # for history
        gr.Textbox(value="You are a friendly AI assistant capable of understanding images and text.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
    ],
    outputs=[
        gr.Textbox(label="Response"),
        gr.State()  # for updated history
    ],
    title="MiniCPM-Llama3-V-2_5 Image and Text Chat",
    description="Upload an image and ask questions about it, or just chat without an image.",
    allow_flagging="never"
)

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