app.py

import spaces
import gradio as gr
from transformers import pipeline
import torch

# Initialize the pipeline
pipe = pipeline(
    "text-generation",
    model="google/medgemma-27b-text-it",
    torch_dtype=torch.bfloat16,
    device="cuda",
)

@spaces.GPU
def generate_response(system_prompt, user_prompt):
    messages = [
        {"role": "system", "content": system_prompt},
        {"role": "user", "content": user_prompt},
    ]
    
    output = pipe(messages, max_new_tokens=2048)
    return output[0]["generated_text"][-1]["content"]

# Create Gradio UI
demo = gr.Interface(
    fn=generate_response,
    inputs=[
        gr.Textbox(label="System Prompt", value="You are a helpful medical assistant."),
        gr.Textbox(label="User Prompt", placeholder="Enter your question here..."),
    ],
    outputs=gr.Textbox(label="Generated Response"),
    title="MedGemma Medical Assistant",
    description="Enter a system and user prompt to generate a medically-informed response."
)

# Launch the app
demo.launch()