#!/bin/bash
# prediction 알고리즘 진단 스냅샷 수집기 (5분 주기, 수동 종료까지 연속 실행)
#
# 용도: DAR-03 G코드 + 쌍끌이 + 어구 위반 포함 알고리즘 동작 검증
# 실행: nohup bash /home/apps/kcg-ai-prediction/scripts/diagnostic-snapshot.sh &
# 종료: kill $(cat /home/apps/kcg-ai-prediction/data/diag/diag.pid)
# 출력: /home/apps/kcg-ai-prediction/data/diag/YYYYMMDD-HHMM.txt

set -u

OUTDIR=/home/apps/kcg-ai-prediction/data/diag
mkdir -p "$OUTDIR"
echo $$ > "$OUTDIR/diag.pid"

export PGPASSWORD=Kcg2026ai
PSQL="psql -U kcg-app -d kcgaidb -h 211.208.115.83 -P pager=off -x"
PSQL_TABLE="psql -U kcg-app -d kcgaidb -h 211.208.115.83 -P pager=off"

INTERVAL_SEC=300  # 5분

while true; do

STAMP=$(date '+%Y%m%d-%H%M')
OUT="$OUTDIR/$STAMP.txt"

{
echo "###################################################################"
echo "# PREDICTION DIAGNOSTIC SNAPSHOT (DAR-03 enhanced)"
echo "# generated: $(date '+%Y-%m-%d %H:%M:%S %Z')"
echo "# host: $(hostname)"
echo "# interval: ${INTERVAL_SEC}s"
echo "###################################################################"

#===================================================================
# PART 1: 종합 지표
#===================================================================
echo ""
echo "================================================================="
echo "PART 1: 종합 지표 (last 5min)"
echo "================================================================="

$PSQL_TABLE << 'SQL'
SELECT count(*) total,
       count(*) FILTER (WHERE vessel_type != 'UNKNOWN') pipeline,
       count(*) FILTER (WHERE vessel_type = 'UNKNOWN') lightweight,
       count(*) FILTER (WHERE is_dark) dark,
       count(*) FILTER (WHERE transship_suspect) transship,
       count(*) FILTER (WHERE gear_judgment IS NOT NULL AND gear_judgment != '') gear_violation,
       count(*) FILTER (WHERE risk_level='CRITICAL') crit,
       count(*) FILTER (WHERE risk_level='HIGH') high,
       round(avg(risk_score)::numeric, 1) avg_risk,
       max(risk_score) max_risk,
       round(count(*) FILTER (WHERE is_dark)::numeric / NULLIF(count(*), 0) * 100, 1) AS dark_pct
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes';
SQL

#===================================================================
# PART 2: 다크베셀 심층 진단
#===================================================================
echo ""
echo "================================================================="
echo "PART 2: DARK VESSEL 심층 진단"
echo "================================================================="

echo ""
echo "--- 2-1. dark_suspicion_score 히스토그램 ---"
$PSQL_TABLE << 'SQL'
SELECT CASE
  WHEN (features->>'dark_suspicion_score')::int >= 90 THEN 'a_90-100 (CRITICAL_HIGH)'
  WHEN (features->>'dark_suspicion_score')::int >= 70 THEN 'b_70-89  (CRITICAL)'
  WHEN (features->>'dark_suspicion_score')::int >= 50 THEN 'c_50-69  (HIGH)'
  WHEN (features->>'dark_suspicion_score')::int >= 30 THEN 'd_30-49  (WATCH)'
  WHEN (features->>'dark_suspicion_score')::int >= 1  THEN 'e_1-29   (NONE_SCORED)'
  ELSE 'f_0      (NOT_DARK)'
  END bucket,
  count(*) cnt,
  round(avg(gap_duration_min)::numeric, 0) avg_gap_min,
  round(avg(risk_score)::numeric, 1) avg_risk
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND is_dark = true
GROUP BY bucket ORDER BY bucket;
SQL

echo ""
echo "--- 2-2. dark_patterns 발동 빈도 ---"
$PSQL_TABLE << 'SQL'
SELECT pattern,
       count(*) cnt,
       round(count(*)::numeric / NULLIF((SELECT count(*) FROM kcg.vessel_analysis_results
         WHERE analyzed_at > now() - interval '5 minutes' AND is_dark), 0) * 100, 1) AS pct
FROM kcg.vessel_analysis_results,
     LATERAL jsonb_array_elements_text(features->'dark_patterns') AS pattern
WHERE analyzed_at > now() - interval '5 minutes'
  AND is_dark = true
GROUP BY pattern ORDER BY cnt DESC;
SQL

echo ""
echo "--- 2-3. P9/P10/P11 + coverage 요약 ---"
$PSQL_TABLE << 'SQL'
WITH dark AS (
  SELECT features FROM kcg.vessel_analysis_results
  WHERE analyzed_at > now() - interval '5 minutes' AND is_dark = true
)
SELECT count(*) total_dark,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%fishing_vessel_dark%'
                      OR features->>'dark_patterns' LIKE '%cargo_natural_gap%') p9,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%underway_deliberate_off%'
                      OR features->>'dark_patterns' LIKE '%anchored_natural_gap%') p10,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%heading_cog_mismatch%') p11,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%out_of_coverage%') coverage
FROM dark;
SQL

echo ""
echo "--- 2-4. CRITICAL dark 상위 10건 ---"
$PSQL_TABLE << 'SQL'
SELECT mmsi, gap_duration_min, zone_code, activity_state,
       (features->>'dark_suspicion_score')::int AS score,
       features->>'dark_tier' AS tier,
       features->>'dark_patterns' AS patterns,
       risk_score, risk_level
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND features->>'dark_tier' = 'CRITICAL'
ORDER BY (features->>'dark_suspicion_score')::int DESC
LIMIT 10;
SQL

#===================================================================
# PART 3: 환적 탐지
#===================================================================
echo ""
echo "================================================================="
echo "PART 3: TRANSSHIPMENT 진단"
echo "================================================================="

$PSQL_TABLE << 'SQL'
SELECT count(*) suspects,
       count(*) FILTER (WHERE (features->>'transship_score')::numeric >= 70) critical,
       count(*) FILTER (WHERE (features->>'transship_score')::numeric >= 50
                          AND (features->>'transship_score')::numeric < 70) high,
       round(avg((features->>'transship_score')::numeric)::numeric, 1) avg_score,
       round(avg(transship_duration_min)::numeric, 0) avg_dur
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes' AND transship_suspect = true;
SQL

echo ""
echo "--- 3-1. 환적 의심 개별 건 ---"
$PSQL_TABLE << 'SQL'
SELECT mmsi, transship_pair_mmsi pair, transship_duration_min dur,
       (features->>'transship_score')::numeric score,
       features->>'transship_tier' tier, zone_code
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes' AND transship_suspect = true
ORDER BY (features->>'transship_score')::numeric DESC;
SQL

#===================================================================
# PART 4: G코드 어구 위반 진단 (DAR-03 신규)
#===================================================================
echo ""
echo "================================================================="
echo "PART 4: G코드 어구 위반 진단 (DAR-03)"
echo "================================================================="

echo ""
echo "--- 4-1. gear_judgment 분포 ---"
$PSQL_TABLE << 'SQL'
SELECT coalesce(NULLIF(gear_judgment, ''), '(none)') AS judgment,
       count(*) cnt,
       round(avg(risk_score)::numeric, 1) avg_risk
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
GROUP BY judgment ORDER BY cnt DESC;
SQL

echo ""
echo "--- 4-2. G코드별 발동 빈도 ---"
$PSQL_TABLE << 'SQL'
SELECT gcode,
       count(*) cnt,
       round(avg(risk_score)::numeric, 1) avg_risk
FROM kcg.vessel_analysis_results,
     LATERAL jsonb_array_elements_text(features->'g_codes') AS gcode
WHERE analyzed_at > now() - interval '5 minutes'
GROUP BY gcode ORDER BY cnt DESC;
SQL

echo ""
echo "--- 4-3. G-01 수역-어구 위반 상세 (상위 20건) ---"
$PSQL_TABLE << 'SQL'
SELECT mmsi, zone_code, vessel_type, risk_score, gear_judgment,
       (features->>'gear_violation_score')::int AS gv_score,
       (features->'gear_violation_evidence'->'G-01'->>'allowed')::text AS allowed
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND features->>'g_codes' LIKE '%G-01%'
ORDER BY risk_score DESC LIMIT 20;
SQL

echo ""
echo "--- 4-4. G-06 쌍끌이 공조 탐지 ---"
$PSQL_TABLE << 'SQL'
SELECT mmsi, zone_code, vessel_type, risk_score,
       (features->'gear_violation_evidence'->'G-06'->>'sync_duration_min') sync_min,
       (features->'gear_violation_evidence'->'G-06'->>'mean_separation_nm') sep_nm,
       (features->'gear_violation_evidence'->'G-06'->>'pair_mmsi') pair_mmsi,
       features->>'pair_trawl_detected' pt
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND (features->>'pair_trawl_detected' = 'true' OR features->>'g_codes' LIKE '%G-06%')
ORDER BY risk_score DESC LIMIT 20;
SQL

echo ""
echo "--- 4-5. G-04 MMSI 조작 + G-05 어구 이동 ---"
$PSQL_TABLE << 'SQL'
SELECT mmsi, zone_code, vessel_type, risk_score,
       features->>'g_codes' g_codes,
       (features->'gear_violation_evidence'->'G-04'->>'cycling_count') g04_cycle,
       (features->'gear_violation_evidence'->'G-05'->>'drift_nm') g05_drift
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND (features->>'g_codes' LIKE '%G-04%' OR features->>'g_codes' LIKE '%G-05%')
ORDER BY risk_score DESC LIMIT 10;
SQL

echo ""
echo "--- 4-6. GEAR_ILLEGAL 이벤트 ---"
$PSQL_TABLE << 'SQL'
SELECT category, level, title, count(*) cnt
FROM kcg.prediction_events
WHERE created_at > now() - interval '5 minutes'
  AND category IN ('GEAR_ILLEGAL', 'MMSI_TAMPERING')
GROUP BY category, level, title ORDER BY cnt DESC;
SQL

echo ""
echo "--- 4-7. violation_categories ILLEGAL_GEAR ---"
$PSQL_TABLE << 'SQL'
SELECT count(*) total,
       count(*) FILTER (WHERE gear_judgment = 'ZONE_VIOLATION') zone_viol,
       count(*) FILTER (WHERE gear_judgment = 'PAIR_TRAWL') pair_trawl,
       count(*) FILTER (WHERE gear_judgment = 'GEAR_MISMATCH') mismatch
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND 'ILLEGAL_GEAR' = ANY(violation_categories);
SQL

#===================================================================
# PART 5: 수역 × 어구 타입 교차 (G-01 검증 핵심)
#===================================================================
echo ""
echo "================================================================="
echo "PART 5: 수역 x 어구 타입 교차 (G-01 검증)"
echo "================================================================="

$PSQL_TABLE << 'SQL'
SELECT zone_code, vessel_type, count(*) total,
       count(*) FILTER (WHERE features->>'g_codes' LIKE '%G-01%') g01
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND vessel_type != 'UNKNOWN' AND zone_code LIKE 'ZONE_%'
GROUP BY zone_code, vessel_type ORDER BY zone_code, vessel_type;
SQL

#===================================================================
# PART 6: 이벤트 + KPI
#===================================================================
echo ""
echo "================================================================="
echo "PART 6: 이벤트 + KPI"
echo "================================================================="

$PSQL_TABLE << 'SQL'
SELECT category, level, count(*) cnt
FROM kcg.prediction_events
WHERE created_at > now() - interval '5 minutes'
GROUP BY category, level ORDER BY cnt DESC;
SQL

$PSQL_TABLE << 'SQL'
SELECT kpi_key, value, trend, delta_pct, updated_at
FROM kcg.prediction_kpi_realtime ORDER BY kpi_key;
SQL

#===================================================================
# PART 7: 사이클 로그 + 에러
#===================================================================
echo ""
echo "================================================================="
echo "PART 7: 사이클 로그 (최근 6분)"
echo "================================================================="
journalctl -u kcg-ai-prediction --since '6 minutes ago' --no-pager 2>/dev/null | \
  grep -E 'analysis cycle:|lightweight|pipeline dark:|event_generator:|pair_trawl|gear_violation|GEAR_ILLEGAL|ERROR|Traceback' | \
  tail -20

#===================================================================
# PART 8: 해역별 종합 교차
#===================================================================
echo ""
echo "================================================================="
echo "PART 8: 해역별 종합 교차표"
echo "================================================================="
$PSQL_TABLE << 'SQL'
SELECT zone_code, count(*) total,
       count(*) FILTER (WHERE is_dark) dark,
       count(*) FILTER (WHERE transship_suspect) transship,
       count(*) FILTER (WHERE gear_judgment IS NOT NULL AND gear_judgment != '') gear_viol,
       count(*) FILTER (WHERE risk_level='CRITICAL') crit,
       count(*) FILTER (WHERE risk_level='HIGH') high,
       round(avg(risk_score)::numeric, 1) avg_risk
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
GROUP BY zone_code ORDER BY total DESC;
SQL

echo ""
echo "================================================================="
echo "END OF SNAPSHOT $STAMP"
echo "================================================================="

} > "$OUT" 2>&1

echo "[diag] $(date '+%H:%M:%S') saved: $OUT ($(wc -l < "$OUT") lines)"

sleep $INTERVAL_SEC
done