"""One-off: load a CSV via the real reader, run the detector with
use_model=True, and print:
  - the new categorised method names + per-row attribution,
  - the aggregate `model_feature_drivers` map from get_statistics()
    (the shape the frontend will chart).
"""
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).parent))

from telecom_anomaly.core.detector import TelecomAnomalyDetector
from telecom_anomaly.core.models import TelecomConfig

CSV = (
    r"C:\my_project\work\GH_Police\main project\backend"
    r"\telecom_anomaly_detection\data\strictly_cleaned_data"
    r"\591296805 calls (1)_strictly_cleaned.csv"
)

cfg = TelecomConfig()
det = TelecomAnomalyDetector(cfg)
det.load_data([CSV])
det.detect_anomalies(use_model=True)

stats = det.get_statistics()
print(f"\nTotal anomalies: {stats.total_anomalies_detected}")
print(f"Rows with anomalies: {stats.rows_with_anomalies}")
print(f"\nanomaly_types (rule + model categories combined):")
for k, v in sorted(stats.anomaly_types.items(), key=lambda x: -x[1]):
    print(f"  {v:4d}  {k}")

print(f"\nmodel_feature_drivers (top-driver counts across all AI flags):")
for k, v in sorted(stats.model_feature_drivers.items(), key=lambda x: -x[1]):
    print(f"  {v:4d}  {k}")

print(f"\nSample AI-flagged rows with category + reason + structured attribution:")
shown = 0
for orig_idx, methods in det.row_anomaly_map.items():
    ai = [(i, m) for i, m in enumerate(methods) if m.startswith('AI:')]
    if not ai or shown >= 5:
        continue
    i, method = ai[0]
    reason = det.row_anomaly_reasons[orig_idx][i]
    attribution = det.row_anomaly_attribution[orig_idx][i]
    row = det.original_data[orig_idx]
    print(f"\n  row {orig_idx}: {row.get('event_date')} {row.get('event_time')} "
          f"dur={row.get('duration')}")
    print(f"    method   : {method}")
    print(f"    reason   : {reason}")
    print(f"    attribution: {attribution}")
    shown += 1