app.py

import gradio as gr
from src.pipeline import generate_report

# ------------------------------------------------------------------
# 1. Pre-load models on Space start-up
# ------------------------------------------------------------------
print("Pre-loading models for fast inference …")
try:
    from src.tools_loader import get_tools          # downloads BiomedCLIP + SPECTER-2
    _ = get_tools()
    print("Models pre-loaded successfully!")
except Exception as e:
    print(f"Model pre-loading failed: {e}")

# ------------------------------------------------------------------
# 2. Inference wrapper
# ------------------------------------------------------------------
def process_upload(image_path: str):
    """Run the multi-agent pipeline on an uploaded chest X-ray."""
    if image_path is None:
        return "Please upload a chest X-ray image."

    try:
        report = generate_report(image_path)
        return report
    except Exception as e:
        return f"Error processing image: {e}"

# ------------------------------------------------------------------
# 3. Gradio UI
# ------------------------------------------------------------------
with gr.Blocks(title="Multi-Agent Radiology Assistant") as demo:
    gr.Markdown(
        """
        # Multi-Agent Radiology Assistant  
        Upload a chest X-ray and receive an AI-generated report produced by a multi-agent pipeline.
        """
    )

    # --- Upload widget + button ------------------------------------------------
    with gr.Column():
        input_image = gr.Image(
            type="filepath",
            label="Upload Chest X-ray",
            height=400
        )
        process_btn = gr.Button("Generate Report", variant="primary")

    # --- Report output ---------------------------------------------------------
    output_report = gr.Markdown(label="Radiology Report", show_label=True)

    # --- Wire everything together ---------------------------------------------
    process_btn.click(
        fn=process_upload,
        inputs=input_image,
        outputs=output_report
    )

    gr.Markdown("### Need an example?  \nUse any frontal CXR PNG file and click **Generate Report**.")

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