app_paligemma.py

import gradio as gr
from gradio_client import Client, handle_file

MODELS = {"Paligemma-10B": "akhaliq/paligemma2-10b-ft-docci-448"}


def create_chat_fn(client, system_prompt, temperature, max_tokens, top_k, rep_penalty, top_p):
    def chat(message, history):
        text = message.get("text", "")
        files = message.get("files", [])
        processed_files = [handle_file(f) for f in files]

        response = client.predict(
            message={"text": text, "files": processed_files},
            system_prompt=system_prompt,
            temperature=temperature,
            max_new_tokens=max_tokens,
            top_k=top_k,
            repetition_penalty=rep_penalty,
            top_p=top_p,
            api_name="/chat",
        )
        return response

    return chat


def set_client_for_session(model_name, request: gr.Request):
    headers = {}
    if request and hasattr(request, "headers"):
        x_ip_token = request.headers.get("x-ip-token")
        if x_ip_token:
            headers["X-IP-Token"] = x_ip_token

    return Client(MODELS[model_name], headers=headers)


def safe_chat_fn(message, history, client, system_prompt, temperature, max_tokens, top_k, rep_penalty, top_p):
    if client is None:
        return "Error: Client not initialized. Please refresh the page."
    try:
        return create_chat_fn(client, system_prompt, temperature, max_tokens, top_k, rep_penalty, top_p)(
            message, history
        )
    except Exception as e:
        print(f"Error during chat: {e!s}")
        return f"Error during chat: {e!s}"


with gr.Blocks() as demo:
    client = gr.State()

    with gr.Accordion("Advanced Settings", open=False):
        system_prompt = gr.Textbox(value="You are a helpful AI assistant.", label="System Prompt")
        with gr.Row():
            temperature = gr.Slider(minimum=0.0, maximum=2.0, value=0.7, label="Temperature")
            top_p = gr.Slider(minimum=0.0, maximum=1.0, value=0.95, label="Top P")
        with gr.Row():
            top_k = gr.Slider(minimum=1, maximum=100, value=40, step=1, label="Top K")
            rep_penalty = gr.Slider(minimum=1.0, maximum=2.0, value=1.1, label="Repetition Penalty")
        max_tokens = gr.Slider(minimum=64, maximum=4096, value=1024, step=64, label="Max Tokens")

    chat_interface = gr.ChatInterface(
        fn=safe_chat_fn,
        additional_inputs=[client, system_prompt, temperature, max_tokens, top_k, rep_penalty, top_p],
        multimodal=True,
    )

    # Initialize client on page load with default model
    demo.load(fn=set_client_for_session, inputs=[gr.State("Paligemma-10B")], outputs=[client])  # Using default model

# Move the API access check here, after demo is defined
if hasattr(demo, "fns"):
    for fn in demo.fns.values():
        fn.api_name = False

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