run_dummy.py

#!/usr/bin/env python3
from __future__ import annotations

import json
import os
import sys
from pathlib import Path

import pandas as pd

PROJECT_ROOT = Path(__file__).resolve().parent
SRC_ROOT = PROJECT_ROOT / "src"
sys.path.insert(0, str(SRC_ROOT))

from demo_backend.assets import ensure_assets_from_hub
from demo_backend.paths import (
    DEFAULT_OUTPUT_DIR,
    EXAMPLE_EHR,
    resolve_default_avra_script,
    resolve_default_checkpoint,
    resolve_default_stats,
)
from demo_backend.pipeline import run_full_inference



def main() -> None:
    sample_nifti = PROJECT_ROOT.parent / "avra_public-master" / "inference" / "Oasis4" / "sub-OAS42000_sess-d3016_run-01_T1w_mni_dof_6.nii"
    if not sample_nifti.exists():
        sample_nifti = PROJECT_ROOT.parent / "inference" / "Oasis4" / "sub-OAS42000_sess-d3016_run-01_T1w_mni_dof_6.nii"

    if not sample_nifti.exists():
        raise FileNotFoundError(
            "No sample NIfTI found. Set up a local test NIfTI and update run_dummy.py path."
        )
    if not EXAMPLE_EHR.exists():
        raise FileNotFoundError(f"Sample EHR CSV not found: {EXAMPLE_EHR}")

    assets_repo_id = os.getenv("HF_ASSETS_REPO_ID", "")
    if assets_repo_id:
        ensure_assets_from_hub(repo_id=assets_repo_id, revision=os.getenv("HF_ASSETS_REVISION", "main"))

    ehr_row = pd.read_csv(EXAMPLE_EHR).iloc[0].to_dict()

    out_dir = DEFAULT_OUTPUT_DIR / "dummy_cli"
    out_dir.mkdir(parents=True, exist_ok=True)

    result = run_full_inference(
        nifti_path=sample_nifti,
        ehr_row=ehr_row,
        output_dir=out_dir,
        checkpoint_path=resolve_default_checkpoint(),
        stats_path=resolve_default_stats(),
        avra_script_path=resolve_default_avra_script(),
        device="auto",
        allow_foundation_backbones=False,
        reuse_registration=True,
        use_fsl=False,
        enable_remote_medgemma_report=False,
    )

    print("Predicted:", result["prediction"]["predicted_class_name"])
    print("Confidence:", max(result["prediction"]["class_probabilities"].values()))
    print("Output JSON:", result["output_json"])

    pretty = out_dir / "dummy_cli_summary.json"
    pretty.write_text(json.dumps(result, indent=2), encoding="utf-8")
    print("Saved summary:", pretty)


if __name__ == "__main__":
    main()