"""
evaluate.py

Reproduce the demo test accuracy on the hand-crafted 18-query set.

Run:
    python evaluate.py
"""

from hv_router import HVRouter


DEMO_QUERIES = [
    ("what is the weather in tokyo", "weather"),
    ("remind me to water the plants", "reminder"),
    ("play something relaxing", "music"),
    ("how do i get to the train station", "directions"),
    ("book a table for four", "restaurant"),
    ("show me my recent purchases", "shopping"),
    ("what happened in the news today", "news"),
    ("draft an email to sarah", "email"),
    ("what is on my calendar tomorrow", "calendar"),
    ("how many calories did i eat today", "health"),
    ("transfer fifty dollars to checking", "finance"),
    ("tell me a joke", "out_of_scope"),
    ("what did i buy last week", "shopping"),
    ("send a message to the team", "email"),
    ("how many steps did i take", "health"),
    ("what is the capital of france", "out_of_scope"),
    ("who are you", "out_of_scope"),
    ("sing me a song", "out_of_scope"),
]


def main(model_path="router_model"):
    router = HVRouter.load(model_path)

    print(f"Model: {len(router.domain_labels)} domains, "
          f"{len(router.vocab_index)} vocabulary")
    print()

    correct = 0
    print(f"{'query':<40} {'predicted':<14} {'expected':<14} "
          f"{'peak':>6}  ok")
    print("-" * 80)

    for phrase, expected in DEMO_QUERIES:
        vote, peak = router.scores_and_peak(phrase)
        pred = router.route(phrase)
        ok = pred == expected
        if ok:
            correct += 1
        mark = "OK" if ok else "ERR"
        print(f"{phrase:<40} {pred:<14} {expected:<14} "
              f"{peak:>6.2f}  {mark}")

    print()
    n = len(DEMO_QUERIES)
    print(f"Accuracy: {correct}/{n} = {correct/n:.1%}")


if __name__ == "__main__":
    import sys
    path = sys.argv[1] if len(sys.argv) > 1 else "router_model"
    main(path)