app.py

import gradio as gr
from transformers import AutoProcessor, AutoModelForImageTextToText
from PIL import Image
import torch

model_id = "google/medgemma-4b-it"

# Load model and processor
model = AutoModelForImageTextToText.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",  # Requires 'accelerate'
)
processor = AutoProcessor.from_pretrained(model_id)

def generate_report(image, clinical_info):
    if image is None:
        return "Please upload a medical image."

    # Create message list for chat-style input
    user_content = []
    if clinical_info:
        user_content.append({"type": "text", "text": f"Patient info: {clinical_info}"})
    user_content.append({"type": "text", "text": "Please describe the medical image in a radiology report style."})
    user_content.append({"type": "image", "image": image})

    messages = [
        {"role": "system", "content": [{"type": "text", "text": "You are an expert radiologist."}]},
        {"role": "user", "content": user_content}
    ]

    # Process input
    inputs = processor.apply_chat_template(
        messages, add_generation_prompt=True, tokenize=True,
        return_dict=True, return_tensors="pt"
    ).to(model.device, dtype=torch.bfloat16)

    input_len = inputs["input_ids"].shape[-1]

    with torch.inference_mode():
        output = model.generate(**inputs, max_new_tokens=512, do_sample=True, top_p=0.9, top_k=50)

    generated_ids = output[0]
    decoded = processor.decode(generated_ids[input_len:], skip_special_tokens=True)

    return decoded.strip()

# Gradio interface
gr.Interface(
    fn=generate_report,
    inputs=[
        gr.Image(type="pil", label="Upload Medical Image (X-ray, etc)"),
        gr.Textbox(lines=2, placeholder="e.g. Prior diagnosis: pneumonia. 65-year-old male with cough...", label="Optional Clinical Info")
    ],
    outputs=gr.Textbox(label="Generated Radiology Report"),
    title="🧠 MedGemma Radiology Report Generator",
    description="Upload a medical image and optionally include clinical info (like prior findings or diagnosis). Powered by Google's MedGemma-4B model.",
    allow_flagging="never"
).launch()