operating-room-monitor-server/backend/app/algo_host/subprocess_runner.py

"""Spawn reference bundle child processes (main.py, visualize_result_video.py)."""

from __future__ import annotations

import os
import signal
import subprocess
import sys
from pathlib import Path

from loguru import logger

from app.algo_host.bundle import load_reference_default_config, resolve_bundle_relative_path
from app.algo_host.transcode import VISUALIZATION_MAX_WIDTH


def build_reference_env() -> dict[str, str]:
    env = os.environ.copy()
    env["PYTHONFAULTHANDLER"] = "1"
    env["PYTHONUNBUFFERED"] = "1"
    return env


def build_batch_main_command(*, bundle_dir: Path, config_path: Path) -> list[str]:
    return [
        "uv",
        "run",
        "python",
        "-X",
        "faulthandler",
        str(bundle_dir / "main.py"),
        "--config",
        str(config_path),
    ]


def build_visualization_command(
    *,
    bundle_dir: Path,
    video_path: Path,
    result_path: Path,
    output_video_path: Path,
) -> list[str]:
    cfg = load_reference_default_config(bundle_dir)
    weights = cfg.get("weights") if isinstance(cfg.get("weights"), dict) else {}
    device_cfg = cfg.get("device") if isinstance(cfg.get("device"), dict) else {}
    hand_raw = str((weights or {}).get("hand") or "weights/hand_detect.pt").strip()
    hand_model = resolve_bundle_relative_path(bundle_dir, hand_raw)
    return [
        sys.executable,
        "-X",
        "faulthandler",
        str((bundle_dir / "visualize_result_video.py").resolve()),
        "--video",
        str(video_path.resolve()),
        "--result-txt",
        str(result_path.resolve()),
        "--hand-model",
        str(hand_model),
        "--out-video",
        str(output_video_path.resolve()),
        "--device",
        str(device_cfg.get("type") or "cuda"),
        "--max-width",
        str(VISUALIZATION_MAX_WIDTH),
    ]


def _signal_name(signum: int) -> str:
    try:
        return signal.Signals(signum).name
    except ValueError:
        return f"signal {signum}"


def describe_batch_returncode(returncode: int) -> str:
    if returncode < 0:
        signum = -returncode
        return f"terminated by {_signal_name(signum)} ({signum})"
    if returncode > 128:
        wrapped = returncode - 256
        if wrapped < 0:
            signum = -wrapped
            return f"exit={returncode} (possibly propagated {wrapped}/{_signal_name(signum)})"
    return f"exit={returncode}"


def format_batch_failure(
    returncode: int,
    *,
    stdout: str,
    stderr: str,
    work_dir: Path,
    output_path: Path,
) -> str:
    chunks: list[str] = [
        describe_batch_returncode(returncode),
        f"work_dir={work_dir}",
        f"output={output_path}",
    ]
    stdout = stdout.strip()
    stderr = stderr.strip()
    if stdout:
        chunks.append(f"stdout:\n{stdout[-3000:]}")
    if stderr:
        chunks.append(f"stderr:\n{stderr[-3000:]}")
    return "\n".join(chunks)


def _log_subprocess_output(prefix: str, stdout: str, stderr: str, *, max_lines: int = 40) -> None:
    for label, text in (("stdout", stdout), ("stderr", stderr)):
        lines = [ln for ln in (text or "").splitlines() if ln.strip()]
        if not lines:
            continue
        tail = lines[-max_lines:] if len(lines) > max_lines else lines
        for line in tail:
            logger.info("{} {}", prefix, line)


def run_subprocess(
    cmd: list[str],
    *,
    cwd: Path,
    work_dir: Path,
    output_path: Path,
    log_label: str,
) -> None:
    proc = subprocess.run(
        cmd,
        cwd=str(cwd),
        check=False,
        text=True,
        capture_output=True,
        env=build_reference_env(),
    )
    if proc.returncode != 0:
        msg = format_batch_failure(
            proc.returncode,
            stdout=proc.stdout or "",
            stderr=proc.stderr or "",
            work_dir=work_dir,
            output_path=output_path,
        )
        raise RuntimeError(f"{log_label} failed {msg}")
    _log_subprocess_output(log_label, proc.stdout or "", proc.stderr or "")


def run_batch_main(*, bundle_dir: Path, config_path: Path, work_dir: Path, output_path: Path) -> None:
    cmd = build_batch_main_command(bundle_dir=bundle_dir, config_path=config_path)
    logger.info("reference batch starting: {}", " ".join(cmd))
    run_subprocess(
        cmd,
        cwd=bundle_dir,
        work_dir=work_dir,
        output_path=output_path,
        log_label="reference batch",
    )


def run_visualization_script(
    *,
    bundle_dir: Path,
    video_path: Path,
    result_path: Path,
    raw_output_video_path: Path,
) -> None:
    cmd = build_visualization_command(
        bundle_dir=bundle_dir,
        video_path=video_path,
        result_path=result_path,
        output_video_path=raw_output_video_path,
    )
    logger.info("reference visualization starting: {}", " ".join(cmd))
    run_subprocess(
        cmd,
        cwd=bundle_dir,
        work_dir=raw_output_video_path.parent,
        output_path=raw_output_video_path,
        log_label="visualize",
    )