app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient("google/med-gemma-7b", token=os.getenv("HF_TOKEN"))

def format_prompt(history, user_input, system_message):
    prompt = f"### System:\n{system_message.strip()}\n\n"
    for user_msg, bot_msg in history:
        prompt += f"### User:\n{user_msg.strip()}\n\n"
        prompt += f"### Assistant:\n{bot_msg.strip()}\n\n"
    prompt += f"### User:\n{user_input.strip()}\n\n### Assistant:\n"
    return prompt

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    prompt = format_prompt(history, message, system_message)
    response = ""
    
    for token in client.text_generation(
        prompt=prompt,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        stream=True,
    ):
        response += token.token
        yield response


demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a helpful medical assistant.", 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)"),
    ],
)

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