#!/bin/bash
# prediction 알고리즘 진단 스냅샷 수집기 (5분 주기, 수동 종료까지 연속 실행)
#
# 용도: 알고리즘 재설계 후 동작 검증용. 단순 집계가 아닌 개별 판정 과정 추적.
# 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"
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 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 (어떤 규칙이 얼마나 적용되는지) ---"
$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 선종별 dark 분포 (신규 패턴 검증) ---"
$PSQL_TABLE << 'SQL'
SELECT
  CASE WHEN features->>'dark_patterns' LIKE '%fishing_vessel_dark%' THEN 'FISHING(+10)'
       WHEN features->>'dark_patterns' LIKE '%cargo_natural_gap%' THEN 'CARGO(-10)'
       ELSE 'NO_KIND_EFFECT' END AS p9_effect,
  count(*) cnt,
  round(avg((features->>'dark_suspicion_score')::int)::numeric, 1) avg_score,
  round(avg(gap_duration_min)::numeric, 0) avg_gap
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND is_dark = true
GROUP BY p9_effect ORDER BY cnt DESC;
SQL

echo ""
echo "--- 2-4. P10 항해상태 dark 분포 (신규 패턴 검증) ---"
$PSQL_TABLE << 'SQL'
SELECT
  CASE WHEN features->>'dark_patterns' LIKE '%underway_deliberate_off%' THEN 'UNDERWAY_OFF(+20)'
       WHEN features->>'dark_patterns' LIKE '%anchored_natural_gap%' THEN 'ANCHORED(-15)'
       ELSE 'NO_STATUS_EFFECT' END AS p10_effect,
  count(*) cnt,
  round(avg((features->>'dark_suspicion_score')::int)::numeric, 1) avg_score
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND is_dark = true
GROUP BY p10_effect ORDER BY cnt DESC;
SQL

echo ""
echo "--- 2-5. P11 heading/COG 불일치 (신규 패턴 검증) ---"
$PSQL_TABLE << 'SQL'
SELECT
  CASE WHEN features->>'dark_patterns' LIKE '%heading_cog_mismatch%' THEN 'MISMATCH(+15)'
       ELSE 'NO_MISMATCH' END AS p11_effect,
  count(*) cnt,
  round(avg((features->>'dark_suspicion_score')::int)::numeric, 1) avg_score
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND is_dark = true
GROUP BY p11_effect ORDER BY cnt DESC;
SQL

echo ""
echo "--- 2-6. GAP 구간별 dark_tier 교차표 (임계값 100분 검증) ---"
$PSQL_TABLE << 'SQL'
SELECT CASE
  WHEN gap_duration_min < 100 THEN 'a_lt100 (NOT_DARK 예상)'
  WHEN gap_duration_min < 180 THEN 'b_100-179'
  WHEN gap_duration_min < 360 THEN 'c_180-359'
  WHEN gap_duration_min < 720 THEN 'd_360-719'
  ELSE 'e_gte720' END gap_bucket,
  count(*) total,
  count(*) FILTER (WHERE is_dark) dark,
  count(*) FILTER (WHERE features->>'dark_tier' = 'CRITICAL') crit,
  count(*) FILTER (WHERE features->>'dark_tier' = 'HIGH') high,
  count(*) FILTER (WHERE features->>'dark_tier' = 'WATCH') watch,
  count(*) FILTER (WHERE features->>'dark_tier' = 'NONE') tier_none
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
GROUP BY gap_bucket ORDER BY gap_bucket;
SQL

echo ""
echo "--- 2-7. CRITICAL dark 상위 10건 (개별 판정 상세) ---"
$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 심층 진단
#===================================================================
echo ""
echo "================================================================="
echo "PART 3: TRANSSHIPMENT 심층 진단"
echo "================================================================="

echo ""
echo "--- 3-1. 환적 의심 건수 + 점수 분포 ---"
$PSQL_TABLE << 'SQL'
SELECT count(*) total_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_duration_min
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND transship_suspect = true;
SQL

echo ""
echo "--- 3-2. 환적 의심 개별 건 상세 (전체) ---"
$PSQL_TABLE << 'SQL'
SELECT mmsi, transship_pair_mmsi AS pair_mmsi,
       transship_duration_min AS dur_min,
       (features->>'transship_score')::numeric AS score,
       features->>'transship_tier' AS tier,
       zone_code,
       activity_state,
       risk_score
FROM kcg.vessel_analysis_results
WHERE analyzed_at > now() - interval '5 minutes'
  AND transship_suspect = true
ORDER BY (features->>'transship_score')::numeric DESC;
SQL

echo ""
echo "--- 3-3. 환적 후보 선종 분포 (Stage 1 이종 쌍 검증) ---"
echo "   (이 쿼리는 journalctl 로그에서 추출)"
journalctl -u kcg-ai-prediction --since '6 minutes ago' --no-pager 2>/dev/null | \
  grep -o 'transshipment:.*' | tail -1

#===================================================================
# PART 4: 이벤트 + KPI
#===================================================================
echo ""
echo "================================================================="
echo "PART 4: 이벤트 + KPI (시스템 출력 검증)"
echo "================================================================="

echo ""
echo "--- 4-1. prediction_events (last 5min) ---"
$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

echo ""
echo "--- 4-2. KPI 실시간 ---"
$PSQL_TABLE << 'SQL'
SELECT kpi_key, value, trend, delta_pct, updated_at
FROM kcg.prediction_kpi_realtime ORDER BY kpi_key;
SQL

#===================================================================
# PART 5: signal-batch 정적정보 보강 검증
#===================================================================
echo ""
echo "================================================================="
echo "PART 5: signal-batch 정적정보 보강 검증"
echo "================================================================="

echo ""
echo "--- 5-1. 직전 사이클 enrich 로그 ---"
journalctl -u kcg-ai-prediction --since '6 minutes ago' --no-pager 2>/dev/null | \
  grep -E 'signal-batch enrich|fetch_recent_detail' | tail -2

echo ""
echo "--- 5-2. features 내 신규 패턴(P9/P10/P11) 적용 비율 ---"
$PSQL_TABLE << 'SQL'
WITH dark_vessels AS (
  SELECT features FROM kcg.vessel_analysis_results
  WHERE analyzed_at > now() - interval '5 minutes' AND is_dark = true
)
SELECT
  count(*) AS total_dark,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%fishing_vessel_dark%'
                      OR features->>'dark_patterns' LIKE '%cargo_natural_gap%') AS p9_applied,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%underway_deliberate_off%'
                      OR features->>'dark_patterns' LIKE '%anchored_natural_gap%') AS p10_applied,
  count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%heading_cog_mismatch%') AS p11_applied,
  round(count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%fishing_vessel_dark%'
                            OR features->>'dark_patterns' LIKE '%cargo_natural_gap%')::numeric
        / NULLIF(count(*), 0) * 100, 1) AS p9_pct,
  round(count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%underway_deliberate_off%'
                            OR features->>'dark_patterns' LIKE '%anchored_natural_gap%')::numeric
        / NULLIF(count(*), 0) * 100, 1) AS p10_pct,
  round(count(*) FILTER (WHERE features->>'dark_patterns' LIKE '%heading_cog_mismatch%')::numeric
        / NULLIF(count(*), 0) * 100, 1) AS p11_pct
FROM dark_vessels;
SQL

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

#===================================================================
# PART 7: 해역별 + 위험도 교차 (운영 지표)
#===================================================================
echo ""
echo "================================================================="
echo "PART 7: 해역별 × 위험도 교차표"
echo "================================================================="
$PSQL_TABLE << 'SQL'
SELECT zone_code,
       count(*) total,
       count(*) FILTER (WHERE is_dark) dark,
       count(*) FILTER (WHERE risk_level='CRITICAL') crit,
       count(*) FILTER (WHERE risk_level='HIGH') high,
       round(avg(risk_score)::numeric, 1) avg_risk,
       count(*) FILTER (WHERE transship_suspect) transship
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