"""
Prepare training data for MLX-LM.

MLX-LM expects a directory containing:
  train.jsonl   - chat-format training data
  valid.jsonl   - chat-format validation data (optional)
  test.jsonl    - chat-format test data (optional)

Each line is a JSON object with one of:
  {"messages": [...]}        (chat format, our existing format)
  {"prompt": ..., "completion": ...}
  {"text": "..."}

This script splits our existing corpus into train/valid splits and writes
them into the MLX-LM expected directory layout.

Usage:
    python scripts/prepare_mlx_data.py \
        --input data/synthetic/corpus_50k.jsonl \
        --output data/mlx_dataset \
        --valid-count 100 \
        --seed 42
"""

import argparse
import json
import random
from pathlib import Path


def main():
    parser = argparse.ArgumentParser(description="Prepare MLX-LM training data")
    parser.add_argument("--input", type=Path, required=True)
    parser.add_argument("--output", type=Path, required=True)
    parser.add_argument("--valid-count", type=int, default=100)
    parser.add_argument("--max-samples", type=int, default=None)
    parser.add_argument("--seed", type=int, default=42)
    args = parser.parse_args()

    random.seed(args.seed)

    rows = []
    with args.input.open("r", encoding="utf-8") as f:
        for line in f:
            line = line.strip()
            if not line:
                continue
            rows.append(line)

    if args.max_samples:
        rows = rows[: args.max_samples]

    random.shuffle(rows)
    valid_rows = rows[: args.valid_count]
    train_rows = rows[args.valid_count :]

    args.output.mkdir(parents=True, exist_ok=True)

    train_path = args.output / "train.jsonl"
    valid_path = args.output / "valid.jsonl"

    with train_path.open("w", encoding="utf-8") as f:
        f.write("\n".join(train_rows) + "\n")
    with valid_path.open("w", encoding="utf-8") as f:
        f.write("\n".join(valid_rows) + "\n")

    print(f"Wrote {len(train_rows)} train rows to {train_path}")
    print(f"Wrote {len(valid_rows)} valid rows to {valid_path}")


if __name__ == "__main__":
    main()