"""
Transcribe audio/<module>/*.mp3 -> transcripts/<module>/*.txt using
faster-whisper (CPU, CTranslate2). Deletes the source audio after a
successful transcript (nobody wants the audio kept — text only).
Resumable: skips lectures with an existing transcript.
"""

import argparse
import logging
import sys
from pathlib import Path

from faster_whisper import WhisperModel

logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
log = logging.getLogger(__name__)


def transcribe_file(model: WhisperModel, audio_path: Path, txt_path: Path) -> None:
    segments, info = model.transcribe(str(audio_path), language="ro", beam_size=5, vad_filter=True)
    txt_path.parent.mkdir(parents=True, exist_ok=True)
    with open(txt_path, "w", encoding="utf-8") as f:
        for seg in segments:
            f.write(seg.text.strip() + "\n")
    log.info(f"  Transcribed: {txt_path.name} (lang={info.language}, duration={info.duration:.0f}s)")


def main():
    p = argparse.ArgumentParser(description="Transcribe ATM course audio")
    p.add_argument("--audio-dir", default="audio")
    p.add_argument("--out", default="transcripts")
    p.add_argument("--model", default="medium", help="faster-whisper model size")
    p.add_argument("--only", default=None, help="Limit to one module dir name (for piloting)")
    p.add_argument("--keep-audio", action="store_true", help="Don't delete source mp3 after transcribing")
    args = p.parse_args()

    audio_dir = Path(args.audio_dir)
    out_dir = Path(args.out)
    mp3_files = sorted(audio_dir.glob("*/*.mp3"))
    if args.only:
        mp3_files = [f for f in mp3_files if f.parent.name == args.only]
    if not mp3_files:
        log.error(f"No audio files found under {audio_dir}")
        sys.exit(1)

    log.info(f"Loading faster-whisper model '{args.model}' (CPU, int8)...")
    model = WhisperModel(args.model, device="cpu", compute_type="int8")

    done = skipped = 0
    for audio_path in mp3_files:
        txt_path = out_dir / audio_path.parent.name / (audio_path.stem + ".txt")
        if txt_path.exists() and txt_path.stat().st_size > 50:
            log.info(f"  Skipping (exists): {txt_path.name}")
            skipped += 1
            continue

        log.info(f"Transcribing: {audio_path}")
        transcribe_file(model, audio_path, txt_path)
        if not args.keep_audio:
            audio_path.unlink()
        done += 1

    log.info("=" * 60)
    log.info(f"Transcribed {done}, skipped {skipped}.")


if __name__ == "__main__":
    main()