#!/usr/bin/env python3
import argparse
import json
import subprocess
import sys
from pathlib import Path
from typing import Dict

import orjson
from tqdm import tqdm


def run_cmd(template: str, audio_path: str) -> str:
	cmd = template.replace("{audio}", audio_path)
	proc = subprocess.Popen(cmd, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
	out, err = proc.communicate()
	if proc.returncode != 0:
		sys.stderr.write(f"[ASR ERROR] {audio_path}\n{err}\n")
		return ""
	# Take first line as transcript
	text = out.strip().splitlines()[0].strip() if out.strip() else ""
	return text


def main() -> None:
	parser = argparse.ArgumentParser(description="Run external ASR command per audio from a JSONL and write hypothesis JSONL.")
	parser.add_argument("--jsonl", required=True, help="Input JSONL with fields id,audio,...")
	parser.add_argument("--cmd", required=True, help="Command template with {audio} placeholder")
	parser.add_argument("--out", required=True, help="Output hypothesis JSONL (id,text)")
	parser.add_argument("--limit", type=int, default=0, help="Limit number of items for quick tests")
	args = parser.parse_args()

	items = []
	with open(args.jsonl, "r", encoding="utf-8") as f:
		for line in f:
			if not line.strip():
				continue
			items.append(json.loads(line))
	if args.limit:
		items = items[: args.limit]

	with open(args.out, "w", encoding="utf-8") as out_f:
		for it in tqdm(items, desc="Transcribing"):
			audio = it.get("audio") or it.get("audio_filepath")
			text = run_cmd(args.cmd, audio)
			rec = {"id": it["id"], "text": text}
			out_f.write(orjson.dumps(rec).decode("utf-8") + "\n")
	print(f"Wrote hypotheses: {args.out}")


if __name__ == "__main__":
	main()