#!/usr/bin/env python3
"""Completion CLI: continue a prompt with a trained Notio checkpoint.

Usage:
    src/complete.py "Once upon a time" --checkpoint out/notio.pt --n 2
    src/complete.py "The little cat" --temp 0.8 --top-k 40 --max-tokens 200
"""
import argparse
import sys
from pathlib import Path

sys.path.insert(0, str(Path(__file__).resolve().parent))
import sampler


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("prompt", nargs="?", default="", help="text to complete (default: story start)")
    ap.add_argument("--checkpoint", default="out/notio.pt")
    ap.add_argument("--temp", type=float, default=0.9)
    ap.add_argument("--top-k", type=int, default=0, help="0 = off")
    ap.add_argument("--max-tokens", type=int, default=900)
    ap.add_argument("--n", type=int, default=1, help="number of completions")
    ap.add_argument("--device", default=None)
    a = ap.parse_args()

    m, device, ck = sampler.load_model(a.checkpoint, a.device)
    print(f"checkpoint: {a.checkpoint} | step {ck['step']} | "
          f"train {ck.get('loss', float('nan')):.3f} | val {ck.get('val_loss', float('nan')):.3f} | {device}")

    prompt = a.prompt or "Once upon a time"
    prompt_ids = sampler.encode_prompt(prompt)
    block = m.cfg.layer1.block_size
    if len(prompt_ids) >= block:
        sys.exit(f"prompt too long: {len(prompt_ids)} ids, block_size {block}")
    print(f"prompt: {prompt!r} ({len(prompt_ids)} ids) | temp {a.temp} | top-k {a.top_k}")
    for i in range(a.n):
        ids = sampler.generate(m, device, prompt_ids, a.max_tokens, a.temp, a.top_k)
        print(f"\n=== completion {i} ({len(ids)} ids) ===\n{sampler.ids_to_display(ids)}")


if __name__ == "__main__":
    main()