#!/usr/bin/env python3
"""
==============================================================================
LEGALBENCH-ID — OFFICIAL EVALUATION CLI RUNNER
==============================================================================
Dahono Labs (PT Dahono Consulting Agency)
Contoh Penggunaan:
  # Evaluasi via Ollama lokal (50 butir cepat atau 500 penuh):
  python run_eval.py --backend ollama --model dahono-4b:v0.8 --samples 50

  # Evaluasi via endpoint vLLM / OpenAI API:
  python run_eval.py --backend openai --api-url http://localhost:8000/v1 --model DahonoLabs/Dahono-4B

  # Uji Coba Cepat Tanpa Server (Mock):
  python run_eval.py --backend mock --samples 5
==============================================================================
"""

import os
import sys
import json
import time
import argparse
from typing import List, Dict, Any

try:
    sys.stdout.reconfigure(encoding='utf-8')
    sys.stderr.reconfigure(encoding='utf-8')
except Exception:
    pass

BASE_DIR = os.path.dirname(os.path.abspath(__file__))
sys.path.insert(0, BASE_DIR)

from harness.evaluator import LegalBenchEvaluator


def load_dataset(dataset_path_or_name: str) -> List[Dict[str, Any]]:
    """Memuat berkas soal uji dari berkas lokal atau Hugging Face Hub."""
    if os.path.exists(dataset_path_or_name):
        print(f"[Dataset] Memuat sampel dari berkas lokal: {dataset_path_or_name}")
        records = []
        if dataset_path_or_name.endswith(".parquet"):
            import pandas as pd
            df = pd.read_parquet(dataset_path_or_name)
            return df.to_dict(orient="records")
        with open(dataset_path_or_name, "r", encoding="utf-8") as f:
            for line in f:
                if line.strip():
                    records.append(json.loads(line))
        return records

    try:
        from datasets import load_dataset as hf_load_dataset
        print(f"[Dataset] Mengunduh dataset dari Hugging Face: {dataset_path_or_name}")
        ds = hf_load_dataset(dataset_path_or_name, split="test")
        return [dict(x) for x in ds]
    except Exception as e:
        raise RuntimeError(f"Gagal memuat dataset '{dataset_path_or_name}': {e}")


def main():
    parser = argparse.ArgumentParser(description="LegalBench-ID Evaluation CLI")
    parser.add_argument("--backend", choices=["ollama", "openai", "mock"], default="ollama", help="Backend inferensi (ollama, openai, mock)")
    parser.add_argument("--model", default="dahono-4b:v0.8", help="Nama model atau tag lokal")
    parser.add_argument("--api-url", default="http://127.0.0.1:11434", help="URL API inferensi")
    parser.add_argument("--api-key", default=None, help="Kunci API jika backend membutuhkan autentikasi")
    parser.add_argument("--dataset", default=os.path.join(BASE_DIR, "data", "test.jsonl"), help="Path dataset lokal atau nama Hugging Face repo")
    parser.add_argument("--samples", type=int, default=50, help="Jumlah soal uji (maksimal 500)")
    parser.add_argument("--citation-mode", choices=["fuzzy", "strict"], default="fuzzy", help="Metode penilaian rujukan pasal")
    parser.add_argument("--fewshot", action="store_true", help="Gunakan mode 5-shot prompt")
    parser.add_argument("--output-json", default="eval_results.json", help="Berkas hasil JSON")
    parser.add_argument("--output-report", default="eval_report.md", help="Berkas laporan Markdown")
    args = parser.parse_args()

    print("=" * 80)
    print("⚖️  LEGALBENCH-ID: THE SOVEREIGN INDONESIAN LEGAL AI BENCHMARK")
    print("=" * 80)
    print(f"Target Model      : {args.model}")
    print(f"Backend           : {args.backend} ({args.api_url if args.backend != 'mock' else 'Mock Mode'})")
    print(f"Dataset Path      : {args.dataset}")
    print(f"Jumlah Sampel     : {args.samples}")
    print(f"Metode Sitasi     : {args.citation_mode.upper()}")
    print(f"Few-Shot Mode     : {'Aktif (5-Shot)' if args.fewshot else 'Nonaktif (Zero-Shot)'}")
    print("=" * 80)

    # 1. Muat Dataset
    dataset = load_dataset(args.dataset)
    print(f"[Dataset] Sukses memuat {len(dataset)} soal.")

    fewshot_path = os.path.join(BASE_DIR, "data", "fewshot_examples.jsonl") if args.fewshot else None

    # 2. Inisialisasi Evaluator
    evaluator = LegalBenchEvaluator(
        backend=args.backend,
        model=args.model,
        api_url=args.api_url,
        api_key=args.api_key,
        citation_mode=args.citation_mode,
        fewshot_path=fewshot_path
    )

    # 3. Jalankan Evaluasi
    summary = evaluator.run(dataset, max_samples=args.samples, verbose=True)

    metrics = summary["aggregate_metrics"]
    print("\n" + "=" * 80)
    print("HASIL AKHIR EVALUASI HUKUM:")
    print("=" * 80)
    print(f"1. Akurasi Doktrinal & Analisis   : {metrics['doctrinal_legal_reasoning']}%")
    print(f"2. Presisi Sitasi Regulasi        : {metrics['citation_precision']}% (Mode: {args.citation_mode})")
    print(f"3. Kemurnian Bahasa Indonesia      : {metrics['indonesian_language_purity']}%")
    print(f"4. Kepatuhan Struktur Nalar (CoT) : {metrics['cot_structural_compliance']}%")
    print("-" * 80)
    print(f"🏆 COMPOSITE LEGAL SCORE           : {metrics['composite_legal_score']}%")
    print("=" * 80)

    # 4. Simpan Berkas
    out_json = os.path.join(BASE_DIR, args.output_json)
    with open(out_json, "w", encoding="utf-8") as f:
        json.dump(summary, f, indent=2, ensure_ascii=False)
    print(f"[Simpan] Berkas JSON tersimpan di: {out_json}")

    out_md = os.path.join(BASE_DIR, args.output_report)
    with open(out_md, "w", encoding="utf-8") as f:
        f.write(f"# Laporan Evaluasi LegalBench-ID: {args.model}\n\n")
        f.write(f"- Tanggal Evaluasi: {summary['metadata']['timestamp']}\n")
        f.write(f"- Backend: `{args.backend}`\n")
        f.write(f"- Sampel Teruji: {summary['metadata']['samples_evaluated']} soal\n")
        f.write(f"- Durasi: {summary['metadata']['total_duration_sec']} detik\n\n")
        f.write("## Ringkasan Metrik\n\n")
        f.write("| Dimensi Yuridis | Skor Hasil Uji |\n")
        f.write("|---|:---:|\n")
        f.write(f"| **Composite Legal Score** | **{metrics['composite_legal_score']}%** |\n")
        f.write(f"| Akurasi Doktrinal Hukum | {metrics['doctrinal_legal_reasoning']}% |\n")
        f.write(f"| Presisi Sitasi Regulasi | {metrics['citation_precision']}% |\n")
        f.write(f"| Kemurnian Bahasa Indonesia | {metrics['indonesian_language_purity']}% |\n")
        f.write(f"| Kepatuhan Struktur CoT | {metrics['cot_structural_compliance']}% |\n")
    print(f"[Simpan] Berkas Markdown tersimpan di: {out_md}")


if __name__ == "__main__":
    main()