Unify runtime state in SQLite and add rollout observability

2026-03-20 23:00:07 +08:00
parent 6b76290cff
commit 43749fff7c
24 changed files with 1875 additions and 15 deletions
@@ -17,6 +17,7 @@ OPEN_METEO_DISK_CACHE_PATH=/var/lib/polyweather/open_meteo_cache.json
 # Windows / macOS can usually keep the defaults.
 UID=1000
 GID=1000
+POLYWEATHER_STATE_STORAGE_MODE=dual

 ########################################
 # 2) Telegram bot minimal
@@ -0,0 +1,72 @@
+{
+  "evaluation_report_path": "E:\\web\\PolyWeather\\artifacts\\probability_calibration\\evaluation_report.json",
+  "shadow_report_path": "E:\\web\\PolyWeather\\artifacts\\probability_calibration\\shadow_report.json",
+  "evaluation_report_exists": true,
+  "shadow_report_exists": true,
+  "decision": {
+    "decision": "hold",
+    "ready_for_primary": false,
+    "summary": "当前指标不足以切换 emos_primary，应继续保持 shadow。",
+    "thresholds": {
+      "evaluation_min_samples": 80,
+      "shadow_min_samples": 50,
+      "max_delta_mae": 0.05,
+      "min_delta_crps": -0.02,
+      "min_delta_bucket_hit_rate": 0.0,
+      "max_delta_bucket_brier_promote": 0.02,
+      "max_delta_bucket_brier_observe": 0.15
+    },
+    "evaluation": {
+      "sample_count": 105,
+      "delta_crps": -0.093663,
+      "delta_mae": 0.0,
+      "delta_bucket_hit_rate": 0.0
+    },
+    "shadow": {
+      "sample_count": 103,
+      "delta_mae": 0.012708,
+      "delta_bucket_hit_rate": 0.009709,
+      "delta_bucket_brier": 0.293835
+    },
+    "blocking_reasons": [
+      "shadow bucket brier 退化超限：delta=0.293835"
+    ],
+    "worst_shadow_regressions": [
+      {
+        "city": "dallas",
+        "samples": 4,
+        "delta_mae": 0.114807,
+        "delta_bucket_hit_rate": 0.0,
+        "delta_bucket_brier": 0.778678
+      },
+      {
+        "city": "chicago",
+        "samples": 4,
+        "delta_mae": 0.075265,
+        "delta_bucket_hit_rate": 0.0,
+        "delta_bucket_brier": 0.746156
+      },
+      {
+        "city": "seattle",
+        "samples": 4,
+        "delta_mae": 0.11262,
+        "delta_bucket_hit_rate": 0.0,
+        "delta_bucket_brier": 0.692003
+      },
+      {
+        "city": "atlanta",
+        "samples": 4,
+        "delta_mae": 0.293028,
+        "delta_bucket_hit_rate": -0.25,
+        "delta_bucket_brier": 0.601425
+      },
+      {
+        "city": "miami",
+        "samples": 4,
+        "delta_mae": 0.241559,
+        "delta_bucket_hit_rate": -0.5,
+        "delta_bucket_brier": 0.478245
+      }
+    ]
+  }
+}
@@ -1 +1,37 @@
 {"city": "ankara", "timestamp": "2026-03-20T12:00:00+03:00", "date": "2026-03-20", "temp_symbol": "°C", "raw_mu": 15.2, "raw_sigma": 1.2, "deb_prediction": 15.4, "ensemble": {"p10": 14.8, "median": 15.8, "p90": 17.9}, "multi_model": {"ECMWF": 15.8, "GFS": 14.1, "ICON": 15.9}, "max_so_far": 15.0, "peak_status": "before", "prob_snapshot": [{"v": 15, "p": 0.552}, {"v": 16, "p": 0.377}], "shadow_prob_snapshot": [{"v": 15, "p": 0.324}, {"v": 16, "p": 0.238}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320130245", "calibration_source": "artifacts/probability_calibration/default.json", "calibrated_mu": 15.1, "calibrated_sigma": 1.25}
+{"city": "test_city", "timestamp": "2026-03-04 10:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.5625, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 26.0, "peak_status": "before", "prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "shadow_prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.5625}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 33.3, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 33.0, "peak_status": "in_window", "prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "shadow_prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 33.3, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 22:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 16:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.85, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.5, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 29.5, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "shadow_prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.85, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 14:30", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 10:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.5625, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 26.0, "peak_status": "before", "prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "shadow_prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.5625}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 33.3, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 33.0, "peak_status": "in_window", "prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "shadow_prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 33.3, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 22:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 16:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.85, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.5, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 29.5, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "shadow_prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.85, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 14:30", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 10:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.5625, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 26.0, "peak_status": "before", "prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "shadow_prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.5625}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 33.3, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 33.0, "peak_status": "in_window", "prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "shadow_prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 33.3, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 22:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 16:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.85, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.5, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 29.5, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "shadow_prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.85, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 14:30", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 10:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.5625, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 26.0, "peak_status": "before", "prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "shadow_prob_snapshot": [{"v": 30, "p": 0.254}, {"v": 29, "p": 0.234}, {"v": 31, "p": 0.185}, {"v": 28, "p": 0.146}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.5625}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 33.3, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 33.0, "peak_status": "in_window", "prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "shadow_prob_snapshot": [{"v": 33, "p": 0.456}, {"v": 34, "p": 0.391}, {"v": 35, "p": 0.153}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 33.3, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 17:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 22:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": null, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "past", "prob_snapshot": [{"v": 28, "p": 1.0}], "shadow_prob_snapshot": [], "probability_engine": "legacy", "probability_mode": "legacy", "calibration_version": null, "calibration_source": null, "calibrated_mu": null, "calibrated_sigma": null}
+{"city": "test_city", "timestamp": "2026-03-04 16:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 23.0, "raw_sigma": 0.46875, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 23.0, "peak_status": "past", "prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "shadow_prob_snapshot": [{"v": 23, "p": 0.834}, {"v": 24, "p": 0.166}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 23.0, "calibrated_sigma": 0.46875}
+{"city": "test_city", "timestamp": "2026-03-04 14:00", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.85, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.5, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 29.5, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "shadow_prob_snapshot": [{"v": 30, "p": 0.565}, {"v": 31, "p": 0.341}, {"v": 32, "p": 0.094}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.85, "calibrated_sigma": 1.09375}
+{"city": "test_city", "timestamp": "2026-03-04 14:30", "date": "2026-03-04", "temp_symbol": "°C", "raw_mu": 29.7, "raw_sigma": 1.09375, "deb_prediction": null, "ensemble": {"p10": 27.0, "median": 29.0, "p90": 31.0}, "multi_model": {"Open-Meteo": 30.0}, "max_so_far": 28.0, "peak_status": "in_window", "prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "shadow_prob_snapshot": [{"v": 30, "p": 0.35}, {"v": 29, "p": 0.299}, {"v": 31, "p": 0.187}, {"v": 28, "p": 0.117}], "probability_engine": "legacy", "probability_mode": "emos_shadow", "calibration_version": "emos-20260320132525", "calibration_source": "artifacts\\probability_calibration\\default.json", "calibrated_mu": 29.7, "calibrated_sigma": 1.09375}
@@ -0,0 +1,53 @@
+
+import argparse
+import json
+import os
+import sys
+
+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if PROJECT_ROOT not in sys.path:
+    sys.path.insert(0, PROJECT_ROOT)
+
+from src.database.runtime_state import (  # noqa: E402
+    DailyRecordRepository,
+    OpenMeteoCacheRepository,
+    ProbabilitySnapshotRepository,
+    TelegramAlertStateRepository,
+)
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Export runtime state from SQLite back to legacy JSON files.')
+    parser.add_argument('--daily-records', default=os.path.join(PROJECT_ROOT, 'data', 'daily_records.json'))
+    parser.add_argument('--telegram-state', default=os.path.join(PROJECT_ROOT, 'data', 'telegram_alert_state.json'))
+    parser.add_argument('--snapshots', default=os.path.join(PROJECT_ROOT, 'data', 'probability_training_snapshots.jsonl'))
+    parser.add_argument('--open-meteo-cache', default=os.path.join(PROJECT_ROOT, 'data', 'open_meteo_cache.json'))
+    parser.add_argument('--open-meteo-max-age', type=int, default=int(os.getenv('OPEN_METEO_DISK_CACHE_MAX_AGE_SEC', '86400')))
+    args = parser.parse_args()
+
+    daily = DailyRecordRepository().load_all()
+    telegram = TelegramAlertStateRepository().load_state()
+    snapshots = ProbabilitySnapshotRepository().load_all_rows()
+    open_meteo = OpenMeteoCacheRepository().load_payload(args.open_meteo_max_age)
+
+    os.makedirs(os.path.dirname(os.path.abspath(args.daily_records)), exist_ok=True)
+    with open(args.daily_records, 'w', encoding='utf-8') as fh:
+        json.dump(daily, fh, ensure_ascii=False, indent=2)
+    with open(args.telegram_state, 'w', encoding='utf-8') as fh:
+        json.dump(telegram, fh, ensure_ascii=False, indent=2)
+    with open(args.snapshots, 'w', encoding='utf-8') as fh:
+        for row in snapshots:
+            fh.write(json.dumps(row, ensure_ascii=False) + '\n')
+    with open(args.open_meteo_cache, 'w', encoding='utf-8') as fh:
+        json.dump(open_meteo, fh, ensure_ascii=False)
+
+    print(json.dumps({
+        'daily_records_exported': sum(len(v) for v in daily.values()),
+        'telegram_state_exported': len((telegram.get('last_by_city') or {})) + len((telegram.get('by_signature') or {})),
+        'snapshots_exported': len(snapshots),
+        'open_meteo_cache_exported': sum(len((open_meteo.get(k) or {})) for k in ('forecast', 'ensemble', 'multi_model')),
+    }, ensure_ascii=False, indent=2))
+
+
+if __name__ == '__main__':
+    main()
@@ -14,6 +14,11 @@ from src.analysis.probability_calibration import (  # noqa: E402
    fit_calibration,
 )
 from src.analysis.deb_algorithm import load_history  # noqa: E402
+from src.database.runtime_state import (  # noqa: E402
+    ProbabilitySnapshotRepository,
+    STATE_STORAGE_SQLITE,
+    get_state_storage_mode,
+)


 def _sf(value):
@@ -41,6 +46,8 @@ def _load_history_with_fallback(path):


 def _load_snapshot_rows(path):
+    if get_state_storage_mode() == STATE_STORAGE_SQLITE:
+        return ProbabilitySnapshotRepository().load_all_rows()
    rows = []
    if not path or not os.path.exists(path):
        return rows
@@ -0,0 +1,56 @@
+import argparse
+import json
+import os
+import sys
+
+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if PROJECT_ROOT not in sys.path:
+    sys.path.insert(0, PROJECT_ROOT)
+
+from src.analysis.probability_rollout import build_rollout_report  # noqa: E402
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Judge whether EMOS is ready for primary rollout.")
+    parser.add_argument(
+        "--evaluation-report",
+        default=os.path.join(
+            PROJECT_ROOT,
+            "artifacts",
+            "probability_calibration",
+            "evaluation_report.json",
+        ),
+    )
+    parser.add_argument(
+        "--shadow-report",
+        default=os.path.join(
+            PROJECT_ROOT,
+            "artifacts",
+            "probability_calibration",
+            "shadow_report.json",
+        ),
+    )
+    parser.add_argument(
+        "--output",
+        default=os.path.join(
+            PROJECT_ROOT,
+            "artifacts",
+            "probability_calibration",
+            "rollout_report.json",
+        ),
+    )
+    args = parser.parse_args()
+
+    payload = build_rollout_report(args.evaluation_report, args.shadow_report)
+    output_dir = os.path.dirname(os.path.abspath(args.output))
+    if output_dir:
+        os.makedirs(output_dir, exist_ok=True)
+    with open(args.output, "w", encoding="utf-8") as fh:
+        json.dump(payload, fh, ensure_ascii=False, indent=2)
+
+    print(json.dumps(payload["decision"], ensure_ascii=False, indent=2))
+    print(f"saved rollout report to {args.output}")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,73 @@
+
+import argparse
+import json
+import os
+import sys
+
+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if PROJECT_ROOT not in sys.path:
+    sys.path.insert(0, PROJECT_ROOT)
+
+from src.database.runtime_state import (  # noqa: E402
+    DailyRecordRepository,
+    OpenMeteoCacheRepository,
+    ProbabilitySnapshotRepository,
+    TelegramAlertStateRepository,
+)
+
+
+def _load_json(path, default):
+    if not path or not os.path.exists(path):
+        return default
+    with open(path, 'r', encoding='utf-8') as fh:
+        data = json.load(fh)
+    return data
+
+
+def _load_jsonl(path):
+    rows = []
+    if not path or not os.path.exists(path):
+        return rows
+    with open(path, 'r', encoding='utf-8') as fh:
+        for line in fh:
+            line = line.strip()
+            if not line:
+                continue
+            try:
+                row = json.loads(line)
+            except Exception:
+                continue
+            if isinstance(row, dict):
+                rows.append(row)
+    return rows
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Migrate runtime JSON state into SQLite.')
+    parser.add_argument('--daily-records', default=os.path.join(PROJECT_ROOT, 'data', 'daily_records.json'))
+    parser.add_argument('--telegram-state', default=os.path.join(PROJECT_ROOT, 'data', 'telegram_alert_state.json'))
+    parser.add_argument('--snapshots', default=os.path.join(PROJECT_ROOT, 'data', 'probability_training_snapshots.jsonl'))
+    parser.add_argument('--open-meteo-cache', default=os.path.join(PROJECT_ROOT, 'data', 'open_meteo_cache.json'))
+    parser.add_argument('--open-meteo-max-age', type=int, default=int(os.getenv('OPEN_METEO_DISK_CACHE_MAX_AGE_SEC', '86400')))
+    args = parser.parse_args()
+
+    daily = _load_json(args.daily_records, {})
+    telegram = _load_json(args.telegram_state, {'last_by_city': {}, 'by_signature': {}})
+    snapshots = _load_jsonl(args.snapshots)
+    open_meteo = _load_json(args.open_meteo_cache, {'forecast': {}, 'ensemble': {}, 'multi_model': {}, 'saved_at': 0})
+
+    daily_count = DailyRecordRepository().replace_all(daily if isinstance(daily, dict) else {})
+    telegram_count = TelegramAlertStateRepository().replace_from_state(telegram if isinstance(telegram, dict) else {})
+    snapshot_count = ProbabilitySnapshotRepository().replace_all(snapshots)
+    cache_count = OpenMeteoCacheRepository().replace_payload(open_meteo if isinstance(open_meteo, dict) else {}, args.open_meteo_max_age)
+
+    print(json.dumps({
+        'daily_records_imported': daily_count,
+        'telegram_state_imported': telegram_count,
+        'snapshots_imported': snapshot_count,
+        'open_meteo_cache_imported': cache_count,
+    }, ensure_ascii=False, indent=2))
+
+
+if __name__ == '__main__':
+    main()
@@ -0,0 +1,101 @@
+
+import argparse
+import json
+import os
+import sys
+
+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if PROJECT_ROOT not in sys.path:
+    sys.path.insert(0, PROJECT_ROOT)
+
+from src.database.runtime_state import (  # noqa: E402
+    DailyRecordRepository,
+    OpenMeteoCacheRepository,
+    ProbabilitySnapshotRepository,
+    TelegramAlertStateRepository,
+)
+
+
+def _load_json(path, default):
+    if not path or not os.path.exists(path):
+        return default
+    with open(path, 'r', encoding='utf-8') as fh:
+        return json.load(fh)
+
+
+def _load_jsonl(path):
+    rows = []
+    if not path or not os.path.exists(path):
+        return rows
+    with open(path, 'r', encoding='utf-8') as fh:
+        for line in fh:
+            line = line.strip()
+            if not line:
+                continue
+            try:
+                row = json.loads(line)
+            except Exception:
+                continue
+            if isinstance(row, dict):
+                rows.append(row)
+    return rows
+
+
+def _norm_json(obj):
+    return json.loads(json.dumps(obj, ensure_ascii=False, sort_keys=True))
+
+
+def _norm_open_meteo_payload(payload):
+    payload = dict(payload or {})
+    payload.pop('saved_at', None)
+    return _norm_json(payload)
+
+
+def main():
+    parser = argparse.ArgumentParser(description='Verify SQLite runtime state against legacy JSON files.')
+    parser.add_argument('--daily-records', default=os.path.join(PROJECT_ROOT, 'data', 'daily_records.json'))
+    parser.add_argument('--telegram-state', default=os.path.join(PROJECT_ROOT, 'data', 'telegram_alert_state.json'))
+    parser.add_argument('--snapshots', default=os.path.join(PROJECT_ROOT, 'data', 'probability_training_snapshots.jsonl'))
+    parser.add_argument('--open-meteo-cache', default=os.path.join(PROJECT_ROOT, 'data', 'open_meteo_cache.json'))
+    parser.add_argument('--open-meteo-max-age', type=int, default=int(os.getenv('OPEN_METEO_DISK_CACHE_MAX_AGE_SEC', '86400')))
+    args = parser.parse_args()
+
+    file_daily = _load_json(args.daily_records, {})
+    file_telegram = _load_json(args.telegram_state, {'last_by_city': {}, 'by_signature': {}})
+    file_snapshots = _load_jsonl(args.snapshots)
+    file_cache = _load_json(args.open_meteo_cache, {'forecast': {}, 'ensemble': {}, 'multi_model': {}, 'saved_at': 0})
+
+    db_daily = DailyRecordRepository().load_all()
+    db_telegram = TelegramAlertStateRepository().load_state()
+    db_snapshots = ProbabilitySnapshotRepository().load_all_rows()
+    db_cache = OpenMeteoCacheRepository().load_payload(args.open_meteo_max_age)
+
+    report = {
+        'daily_records': {
+            'file_cities': len(file_daily or {}),
+            'db_cities': len(db_daily or {}),
+            'equal': _norm_json(file_daily or {}) == _norm_json(db_daily or {}),
+        },
+        'telegram_state': {
+            'equal': _norm_json(file_telegram or {}) == _norm_json(db_telegram or {}),
+            'file_last_by_city': len((file_telegram or {}).get('last_by_city') or {}),
+            'db_last_by_city': len((db_telegram or {}).get('last_by_city') or {}),
+        },
+        'snapshots': {
+            'file_rows': len(file_snapshots),
+            'db_rows': len(db_snapshots),
+            'equal': _norm_json(file_snapshots) == _norm_json(db_snapshots),
+        },
+        'open_meteo_cache': {
+            'file_forecast': len((file_cache or {}).get('forecast') or {}),
+            'db_forecast': len((db_cache or {}).get('forecast') or {}),
+            'equal': _norm_open_meteo_payload(file_cache or {}) == _norm_open_meteo_payload(db_cache or {}),
+        },
+    }
+    report['ok'] = all(section.get('equal') for section in report.values() if isinstance(section, dict))
+    print(json.dumps(report, ensure_ascii=False, indent=2))
+    raise SystemExit(0 if report['ok'] else 1)
+
+
+if __name__ == '__main__':
+    main()
@@ -3,6 +3,13 @@ import json
 from datetime import datetime, timedelta
 import requests
 from src.analysis.settlement_rounding import apply_city_settlement
+from loguru import logger
+from src.database.runtime_state import (
+    DailyRecordRepository,
+    STATE_STORAGE_DUAL,
+    STATE_STORAGE_SQLITE,
+    get_state_storage_mode,
+)

 # Cross-platform file locking
 import sys
@@ -36,6 +43,7 @@ else:
 # Simple memory cache to avoid blasting the disk if queried 10 times a minute
 _history_cache = {}
 _history_mtime = 0
+_daily_record_repo = DailyRecordRepository()


 def _sf(value):
@@ -54,7 +62,24 @@ def _is_excluded_model_name(model_name: str) -> bool:

 def load_history(filepath):
    global _history_cache, _history_mtime
+    mode = get_state_storage_mode()
+
+    if mode == STATE_STORAGE_SQLITE:
+        try:
+            data = _daily_record_repo.load_all()
+            _history_cache = data
+            return data
+        except Exception as e:
+            logger.error(f"Error loading daily records from sqlite, fallback to file: {e}")
+
    if not os.path.exists(filepath):
+        if mode == STATE_STORAGE_DUAL:
+            try:
+                data = _daily_record_repo.load_all()
+                _history_cache = data
+                return data
+            except Exception:
+                return {}
        return {}

    try:
@@ -80,6 +105,18 @@ def load_history(filepath):
 def save_history(filepath, data):
    global _history_cache, _history_mtime
    _history_cache = data
+    mode = get_state_storage_mode()
+
+    if mode in {STATE_STORAGE_DUAL, STATE_STORAGE_SQLITE}:
+        try:
+            _daily_record_repo.replace_all(data)
+        except Exception as e:
+            logger.error(f"Error saving daily records to sqlite: {e}")
+            if mode == STATE_STORAGE_SQLITE:
+                return
+
+    if mode == STATE_STORAGE_SQLITE:
+        return
    try:
        with open(filepath, "w", encoding="utf-8") as f:
            _lock_ex(f)
@@ -251,6 +288,7 @@ def update_daily_record(
    )
    history_file = os.path.join(project_root, "data", "daily_records.json")

+    mode = get_state_storage_mode()
    data = load_history(history_file)
    if city_name not in data:
        data[city_name] = {}
@@ -359,7 +397,18 @@ def update_daily_record(
        for d in old_dates:
            del data[city][d]

-    save_history(history_file, data)
+    if mode in {STATE_STORAGE_DUAL, STATE_STORAGE_SQLITE}:
+        try:
+            _daily_record_repo.upsert_record(city_name, date_str, existing)
+            cutoff = (datetime.now() - timedelta(days=14)).strftime("%Y-%m-%d")
+            _daily_record_repo.delete_older_than(cutoff)
+        except Exception as e:
+            logger.error(f"Error upserting daily record to sqlite city={city_name} date={date_str}: {e}")
+            if mode == STATE_STORAGE_SQLITE:
+                raise
+
+    if mode != STATE_STORAGE_SQLITE:
+        save_history(history_file, data)


 def calculate_dynamic_weights(city_name, current_forecasts, lookback_days=7):
@@ -0,0 +1,217 @@
+from __future__ import annotations
+
+import json
+import os
+from typing import Any, Dict, List, Optional
+
+
+def _load_json_file(path: str) -> Dict[str, Any]:
+    try:
+        with open(path, "r", encoding="utf-8") as fh:
+            data = json.load(fh)
+        return data if isinstance(data, dict) else {}
+    except Exception:
+        return {}
+
+
+def _sf(value: Any) -> Optional[float]:
+    if value is None:
+        return None
+    try:
+        return float(value)
+    except Exception:
+        return None
+
+
+def _append_reason(reasons: List[str], condition: bool, message: str) -> None:
+    if condition:
+        reasons.append(message)
+
+
+def _top_shadow_regressions(by_city: Dict[str, Any], limit: int = 5) -> List[Dict[str, Any]]:
+    rows: List[Dict[str, Any]] = []
+    for city, metrics in (by_city or {}).items():
+        if not isinstance(metrics, dict):
+            continue
+        rows.append(
+            {
+                "city": city,
+                "samples": int(metrics.get("samples") or 0),
+                "delta_mae": _sf(metrics.get("delta_mae")),
+                "delta_bucket_hit_rate": _sf(metrics.get("delta_bucket_hit_rate")),
+                "delta_bucket_brier": _sf(metrics.get("delta_bucket_brier")),
+            }
+        )
+
+    rows.sort(
+        key=lambda row: (
+            -(row["delta_bucket_brier"] or 0.0),
+            row["delta_bucket_hit_rate"] or 0.0,
+            -(row["delta_mae"] or 0.0),
+        )
+    )
+    return rows[:limit]
+
+
+def judge_probability_rollout(
+    evaluation_report: Dict[str, Any],
+    shadow_report: Dict[str, Any],
+) -> Dict[str, Any]:
+    thresholds = {
+        "evaluation_min_samples": 80,
+        "shadow_min_samples": 50,
+        "max_delta_mae": 0.05,
+        "min_delta_crps": -0.02,
+        "min_delta_bucket_hit_rate": 0.0,
+        "max_delta_bucket_brier_promote": 0.02,
+        "max_delta_bucket_brier_observe": 0.15,
+    }
+
+    eval_summary = (evaluation_report or {}).get("summary") or {}
+    eval_delta = eval_summary.get("delta") or {}
+    shadow_summary = (shadow_report or {}).get("summary") or {}
+
+    eval_samples = int(eval_summary.get("sample_count") or 0)
+    shadow_samples = int(shadow_summary.get("samples") or 0)
+    delta_crps = _sf(eval_delta.get("crps"))
+    delta_mae = _sf(eval_delta.get("mae"))
+    delta_hit = _sf(eval_delta.get("bucket_hit_rate"))
+    shadow_delta_mae = _sf(shadow_summary.get("delta_mae"))
+    shadow_delta_hit = _sf(shadow_summary.get("delta_bucket_hit_rate"))
+    shadow_delta_brier = _sf(shadow_summary.get("delta_bucket_brier"))
+
+    promote_reasons: List[str] = []
+    _append_reason(
+        promote_reasons,
+        eval_samples < thresholds["evaluation_min_samples"],
+        f"离线评估样本不足：{eval_samples} < {thresholds['evaluation_min_samples']}",
+    )
+    _append_reason(
+        promote_reasons,
+        shadow_samples < thresholds["shadow_min_samples"],
+        f"shadow 样本不足：{shadow_samples} < {thresholds['shadow_min_samples']}",
+    )
+    _append_reason(
+        promote_reasons,
+        delta_crps is None or delta_crps > thresholds["min_delta_crps"],
+        f"离线 CRPS 改善不足：delta={delta_crps}",
+    )
+    _append_reason(
+        promote_reasons,
+        delta_mae is None or delta_mae > thresholds["max_delta_mae"],
+        f"离线 MAE 退化超限：delta={delta_mae}",
+    )
+    _append_reason(
+        promote_reasons,
+        delta_hit is None or delta_hit < thresholds["min_delta_bucket_hit_rate"],
+        f"离线 bucket 命中率下降：delta={delta_hit}",
+    )
+    _append_reason(
+        promote_reasons,
+        shadow_delta_mae is None or shadow_delta_mae > thresholds["max_delta_mae"],
+        f"shadow MAE 退化超限：delta={shadow_delta_mae}",
+    )
+    _append_reason(
+        promote_reasons,
+        shadow_delta_hit is None or shadow_delta_hit < thresholds["min_delta_bucket_hit_rate"],
+        f"shadow bucket 命中率下降：delta={shadow_delta_hit}",
+    )
+    _append_reason(
+        promote_reasons,
+        shadow_delta_brier is None
+        or shadow_delta_brier > thresholds["max_delta_bucket_brier_promote"],
+        f"shadow bucket brier 退化超限：delta={shadow_delta_brier}",
+    )
+
+    if not promote_reasons:
+        decision = "promote"
+        summary = "离线与 shadow 指标均达标，可以考虑切换 emos_primary。"
+    else:
+        observe_reasons: List[str] = []
+        _append_reason(
+            observe_reasons,
+            eval_samples < thresholds["evaluation_min_samples"],
+            f"离线评估样本不足：{eval_samples}",
+        )
+        _append_reason(
+            observe_reasons,
+            delta_crps is None or delta_crps > thresholds["min_delta_crps"],
+            f"离线 CRPS 改善不足：delta={delta_crps}",
+        )
+        _append_reason(
+            observe_reasons,
+            delta_mae is None or delta_mae > thresholds["max_delta_mae"],
+            f"离线 MAE 退化超限：delta={delta_mae}",
+        )
+        _append_reason(
+            observe_reasons,
+            delta_hit is None or delta_hit < thresholds["min_delta_bucket_hit_rate"],
+            f"离线 bucket 命中率下降：delta={delta_hit}",
+        )
+        _append_reason(
+            observe_reasons,
+            shadow_samples < thresholds["shadow_min_samples"],
+            f"shadow 样本不足：{shadow_samples}",
+        )
+        _append_reason(
+            observe_reasons,
+            shadow_delta_mae is None or shadow_delta_mae > thresholds["max_delta_mae"],
+            f"shadow MAE 退化超限：delta={shadow_delta_mae}",
+        )
+        _append_reason(
+            observe_reasons,
+            shadow_delta_hit is None or shadow_delta_hit < thresholds["min_delta_bucket_hit_rate"],
+            f"shadow bucket 命中率下降：delta={shadow_delta_hit}",
+        )
+        _append_reason(
+            observe_reasons,
+            shadow_delta_brier is None
+            or shadow_delta_brier > thresholds["max_delta_bucket_brier_observe"],
+            f"shadow bucket brier 退化偏大：delta={shadow_delta_brier}",
+        )
+
+        if not observe_reasons:
+            decision = "observe"
+            summary = "离线评估达标，但 shadow 仍需继续观察，暂不切主路径。"
+        else:
+            decision = "hold"
+            summary = "当前指标不足以切换 emos_primary，应继续保持 shadow。"
+
+    return {
+        "decision": decision,
+        "ready_for_primary": decision == "promote",
+        "summary": summary,
+        "thresholds": thresholds,
+        "evaluation": {
+            "sample_count": eval_samples,
+            "delta_crps": delta_crps,
+            "delta_mae": delta_mae,
+            "delta_bucket_hit_rate": delta_hit,
+        },
+        "shadow": {
+            "sample_count": shadow_samples,
+            "delta_mae": shadow_delta_mae,
+            "delta_bucket_hit_rate": shadow_delta_hit,
+            "delta_bucket_brier": shadow_delta_brier,
+        },
+        "blocking_reasons": promote_reasons,
+        "worst_shadow_regressions": _top_shadow_regressions(
+            (shadow_report or {}).get("by_city") or {}
+        ),
+    }
+
+
+def build_rollout_report(
+    evaluation_report_path: str,
+    shadow_report_path: str,
+) -> Dict[str, Any]:
+    evaluation_report = _load_json_file(evaluation_report_path)
+    shadow_report = _load_json_file(shadow_report_path)
+    decision = judge_probability_rollout(evaluation_report, shadow_report)
+    return {
+        "evaluation_report_path": evaluation_report_path,
+        "shadow_report_path": shadow_report_path,
+        "evaluation_report_exists": os.path.exists(evaluation_report_path),
+        "shadow_report_exists": os.path.exists(shadow_report_path),
+        "decision": decision,
+    }
@@ -5,10 +5,18 @@ import os
 from datetime import datetime
 from typing import Any, Dict, List, Optional

+from src.database.runtime_state import (
+    ProbabilitySnapshotRepository,
+    STATE_STORAGE_DUAL,
+    STATE_STORAGE_SQLITE,
+    get_state_storage_mode,
+)
+
 DEDUP_SCAN_LINES = 200
 MU_THRESHOLD = 0.2
 SIGMA_THRESHOLD = 0.15
 MAX_SO_FAR_THRESHOLD = 0.2
+_snapshot_repo = ProbabilitySnapshotRepository()


 def _sf(value: Any) -> Optional[float]:
@@ -79,7 +87,15 @@ def _load_recent_rows(path: str, max_lines: int = DEDUP_SCAN_LINES) -> List[Dict


 def _should_skip_append(path: str, payload: Dict[str, Any]) -> bool:
-    recent_rows = _load_recent_rows(path)
+    mode = get_state_storage_mode()
+    if mode == STATE_STORAGE_SQLITE:
+        recent_rows = _snapshot_repo.load_recent_rows(
+            str(payload.get("city") or ""),
+            str(payload.get("date") or ""),
+            DEDUP_SCAN_LINES,
+        )
+    else:
+        recent_rows = _load_recent_rows(path)
    city = payload.get("city")
    date_str = payload.get("date")
    if not city or not date_str:
@@ -183,5 +199,10 @@ def append_probability_snapshot(
    if _should_skip_append(path, payload):
        return

-    with open(path, "a", encoding="utf-8") as fh:
-        fh.write(json.dumps(payload, ensure_ascii=False) + "\n")
+    mode = get_state_storage_mode()
+    if mode in {STATE_STORAGE_DUAL, STATE_STORAGE_SQLITE}:
+        _snapshot_repo.append_snapshot(payload)
+
+    if mode != STATE_STORAGE_SQLITE:
+        with open(path, "a", encoding="utf-8") as fh:
+            fh.write(json.dumps(payload, ensure_ascii=False) + "\n")
@@ -8,6 +8,8 @@ from typing import Dict, List, Optional
 import requests
 from loguru import logger

+from src.utils.metrics import record_source_call
+

 class MetarSourceMixin:
    def get_icao_code(self, city: str) -> Optional[str]:
@@ -24,9 +26,11 @@ class MetarSourceMixin:
        self, city: str, use_fahrenheit: bool = False, utc_offset: int = 0
    ) -> Optional[Dict]:
        """从 NOAA Aviation Weather Center 获取 METAR 航空气象数据。"""
+        started = time.perf_counter()
        icao = self.get_icao_code(city)
        if not icao:
            logger.warning(f"未找到城市 {city} 对应的 ICAO 代码")
+            record_source_call("metar", "current", "missing_icao", (time.perf_counter() - started) * 1000.0)
            return None

        cache_key = f"{icao}:{utc_offset}:{use_fahrenheit}"
@@ -35,6 +39,7 @@ class MetarSourceMixin:
            cached = self._metar_cache.get(cache_key)
            if cached and now_ts - cached["t"] < self.metar_cache_ttl_sec:
                logger.debug(f"METAR cache hit {icao} age={int(now_ts - cached['t'])}s")
+                record_source_call("metar", "current", "cache_hit", (time.perf_counter() - started) * 1000.0)
                return cached["d"]

        try:
@@ -201,6 +206,7 @@ class MetarSourceMixin:
            )
            with self._metar_cache_lock:
                self._metar_cache[cache_key] = {"d": result, "t": now_ts}
+            record_source_call("metar", "current", "success", (time.perf_counter() - started) * 1000.0)
            return result

        except requests.exceptions.RequestException as exc:
@@ -209,10 +215,13 @@ class MetarSourceMixin:
                stale = self._metar_cache.get(cache_key)
                if stale:
                    logger.warning(f"METAR {icao} 请求失败，使用缓存回退")
+                    record_source_call("metar", "current", "stale_cache", (time.perf_counter() - started) * 1000.0)
                    return stale["d"]
+            record_source_call("metar", "current", "error", (time.perf_counter() - started) * 1000.0)
            return None
        except (KeyError, IndexError, TypeError) as exc:
            logger.error(f"METAR 数据解析失败 ({icao}): {exc}")
+            record_source_call("metar", "current", "parse_error", (time.perf_counter() - started) * 1000.0)
            return None

    def fetch_metar_nearby_cluster(self, icaos: List[str], use_fahrenheit: bool = False) -> list:
@@ -5,12 +5,15 @@ from typing import Dict, Optional

 from loguru import logger

+from src.utils.metrics import record_source_call
+

 class MgmSourceMixin:
    def fetch_from_mgm(self, istno: str) -> Optional[Dict]:
        """
        从土耳其气象局 (MGM) 获取实时数据和预测 (由用户提供其内部 API)
        """
+        started = datetime.now()
        base_url = "https://servis.mgm.gov.tr/web"
        # 必须带 Origin，否则会被反爬拦截
        headers = {
@@ -202,9 +205,21 @@ class MgmSourceMixin:
                        f"hourly points={len(hourly_rows)}, horizon={horizon_hours:.1f}h"
                    )

+            record_source_call(
+                "mgm",
+                "station",
+                "success" if "current" in results else "empty",
+                (datetime.now() - started).total_seconds() * 1000.0,
+            )
            return results if "current" in results else None
        except Exception as e:
            logger.error(f"MGM API 请求失败 ({istno}): {e}")
+            record_source_call(
+                "mgm",
+                "station",
+                "error",
+                (datetime.now() - started).total_seconds() * 1000.0,
+            )
            return None

    def fetch_mgm_nearby_stations(self, province: str, root_ist_no: str = None) -> list:
@@ -212,6 +227,7 @@ class MgmSourceMixin:
        获取一个土耳其省份内所有气象站的当前温度及经纬度
        使用多线程辅助抓取，因为直接通过 il={province} 往往只返回 1 个站。
        """
+        started = datetime.now()
        base_url = "https://servis.mgm.gov.tr/web"
        headers = {
            "Origin": "https://www.mgm.gov.tr",
@@ -243,6 +259,12 @@ class MgmSourceMixin:

            if not province_ist_nos:
                logger.warning(f"MGM 找不到省份 {province} 的站点元数据")
+                record_source_call(
+                    "mgm",
+                    "nearby",
+                    "empty",
+                    (datetime.now() - started).total_seconds() * 1000.0,
+                )
                return []

            # 同时确保我们关心的几个核心站一定在里面
@@ -318,8 +340,20 @@ class MgmSourceMixin:
                })

            logger.info(f"📍 MGM 周边测站: 成功并发抓取 {len(results)} 个 {province} 站点的实时气温")
+            record_source_call(
+                "mgm",
+                "nearby",
+                "success" if results else "empty",
+                (datetime.now() - started).total_seconds() * 1000.0,
+            )
            return results
        except Exception as e:
            logger.error(f"Failed to fetch MGM nearby stations for {province}: {e}")
+            record_source_call(
+                "mgm",
+                "nearby",
+                "error",
+                (datetime.now() - started).total_seconds() * 1000.0,
+            )
            return []

@@ -6,6 +6,8 @@ from typing import Dict, Optional

 from loguru import logger

+from src.utils.metrics import record_source_call
+

 class NwsOpenMeteoSourceMixin:
    def fetch_nws(self, lat: float, lon: float) -> Optional[Dict]:
@@ -13,6 +15,7 @@ class NwsOpenMeteoSourceMixin:
        从 NWS (美国国家气象局) 获取高精度预报
        仅适用于美国城市，全球 VPS 均可访问
        """
+        started = time.perf_counter()
        try:
            # 1. 获取网格点
            points_url = f"https://api.weather.gov/points/{lat},{lon}"
@@ -28,6 +31,7 @@ class NwsOpenMeteoSourceMixin:
            forecast_url = properties.get("forecast")
            hourly_url = properties.get("forecastHourly")
            if not forecast_url:
+                record_source_call("nws", "forecast", "empty", (time.perf_counter() - started) * 1000.0)
                return None

            # 2. 获取预报
@@ -39,6 +43,7 @@ class NwsOpenMeteoSourceMixin:

            periods = forecast_data.get("properties", {}).get("periods", [])
            if not periods:
+                record_source_call("nws", "forecast", "empty", (time.perf_counter() - started) * 1000.0)
                return None

            hourly_periods = []
@@ -89,7 +94,7 @@ class NwsOpenMeteoSourceMixin:
                        today_high = p.get("temperature")
                        break

-            return {
+            result = {
                "source": "nws",
                "today_high": today_high,
                "unit": "fahrenheit",
@@ -124,8 +129,11 @@ class NwsOpenMeteoSourceMixin:
                ],
                "active_alerts": active_alerts,
            }
+            record_source_call("nws", "forecast", "success", (time.perf_counter() - started) * 1000.0)
+            return result
        except Exception as e:
            logger.warning(f"NWS 请求失败: {e}")
+            record_source_call("nws", "forecast", "error", (time.perf_counter() - started) * 1000.0)
            return None

    def fetch_from_open_meteo(
@@ -144,6 +152,7 @@ class NwsOpenMeteoSourceMixin:
            forecast_days: Number of forecast days to fetch (default 14 to cover all market dates)
            use_fahrenheit: Whether to return temperatures in Fahrenheit (for US markets)
        """
+        started = time.perf_counter()
        cache_key = (
            f"{round(float(lat), 4)}:{round(float(lon), 4)}:"
            f"{forecast_days}:{'f' if use_fahrenheit else 'c'}"
@@ -156,9 +165,11 @@ class NwsOpenMeteoSourceMixin:
                remaining = int(self._open_meteo_rate_limit_until - now_ts)
                logger.debug(f"Open-Meteo 冷却期中，跳过请求，还需 {remaining}s")
                with self._open_meteo_cache_lock:
-                    stale = self._open_meteo_cache.get(cache_key)
-                    if stale and isinstance(stale.get("data"), dict):
-                        return dict(stale["data"])
+                        stale = self._open_meteo_cache.get(cache_key)
+                        if stale and isinstance(stale.get("data"), dict):
+                            record_source_call("open_meteo", "forecast", "stale_cache", (time.perf_counter() - started) * 1000.0)
+                            return dict(stale["data"])
+                record_source_call("open_meteo", "forecast", "cooldown_skip", (time.perf_counter() - started) * 1000.0)
                return None
        with self._open_meteo_cache_lock:
            cached = self._open_meteo_cache.get(cache_key)
@@ -168,6 +179,7 @@ class NwsOpenMeteoSourceMixin:
            ):
                cached_data = cached.get("data")
                if isinstance(cached_data, dict):
+                    record_source_call("open_meteo", "forecast", "cache_hit", (time.perf_counter() - started) * 1000.0)
                    return dict(cached_data)
        try:
            url = "https://api.open-meteo.com/v1/forecast"
@@ -259,6 +271,7 @@ class NwsOpenMeteoSourceMixin:
                    "data": dict(result),
                }
            self._flush_open_meteo_disk_cache()
+            record_source_call("open_meteo", "forecast", "success", (time.perf_counter() - started) * 1000.0)
            return result
        except Exception as e:
            status_code = getattr(getattr(e, "response", None), "status_code", None)
@@ -287,7 +300,9 @@ class NwsOpenMeteoSourceMixin:
                if stale and isinstance(stale.get("data"), dict):
                    fallback = dict(stale["data"])
                    fallback["stale_cache"] = True
+                    record_source_call("open_meteo", "forecast", "stale_cache", (time.perf_counter() - started) * 1000.0)
                    return fallback
+            record_source_call("open_meteo", "forecast", "error", (time.perf_counter() - started) * 1000.0)
            return None

    def fetch_ensemble(
@@ -300,6 +315,7 @@ class NwsOpenMeteoSourceMixin:
        从 Open-Meteo Ensemble API 获取 51 成员集合预报
        用于计算预报不确定性范围（散度）
        """
+        started = time.perf_counter()
        cache_key = (
            f"{round(float(lat), 4)}:{round(float(lon), 4)}:"
            f"{'f' if use_fahrenheit else 'c'}"
@@ -314,7 +330,9 @@ class NwsOpenMeteoSourceMixin:
                with self._ensemble_cache_lock:
                    stale = self._ensemble_cache.get(cache_key)
                    if stale and isinstance(stale.get("data"), dict):
+                        record_source_call("open_meteo", "ensemble", "stale_cache", (time.perf_counter() - started) * 1000.0)
                        return dict(stale["data"])
+                record_source_call("open_meteo", "ensemble", "cooldown_skip", (time.perf_counter() - started) * 1000.0)
                return None
                
        with self._ensemble_cache_lock:
@@ -326,6 +344,7 @@ class NwsOpenMeteoSourceMixin:
            ):
                cached_data = cached.get("data")
                if isinstance(cached_data, dict):
+                    record_source_call("open_meteo", "ensemble", "cache_hit", (time.perf_counter() - started) * 1000.0)
                    return dict(cached_data)
        try:
            url = "https://ensemble-api.open-meteo.com/v1/ensemble"
@@ -371,6 +390,7 @@ class NwsOpenMeteoSourceMixin:

            if len(today_highs) < 3:
                logger.warning(f"Ensemble 数据不足: 仅获取 {len(today_highs)} 个成员")
+                record_source_call("open_meteo", "ensemble", "empty", (time.perf_counter() - started) * 1000.0)
                return None

            today_highs.sort()
@@ -400,6 +420,7 @@ class NwsOpenMeteoSourceMixin:
                    "data": dict(result),
                }
            self._flush_open_meteo_disk_cache()
+            record_source_call("open_meteo", "ensemble", "success", (time.perf_counter() - started) * 1000.0)
            return result
        except Exception as e:
            status_code = getattr(getattr(e, "response", None), "status_code", None)
@@ -425,7 +446,9 @@ class NwsOpenMeteoSourceMixin:
                if stale and isinstance(stale.get("data"), dict):
                    fallback = dict(stale["data"])
                    fallback["stale_cache"] = True
+                    record_source_call("open_meteo", "ensemble", "stale_cache", (time.perf_counter() - started) * 1000.0)
                    return fallback
+            record_source_call("open_meteo", "ensemble", "error", (time.perf_counter() - started) * 1000.0)
            return None

    def fetch_multi_model(
@@ -448,6 +471,7 @@ class NwsOpenMeteoSourceMixin:

        返回 3 天的预报数据，支持今日+明日共识分析
        """
+        started = time.perf_counter()
        cache_city = str(city or "").strip().lower()
        cache_key = (
            f"{round(float(lat), 4)}:{round(float(lon), 4)}:{cache_city}:"
@@ -463,7 +487,9 @@ class NwsOpenMeteoSourceMixin:
                with self._multi_model_cache_lock:
                    stale = self._multi_model_cache.get(cache_key)
                    if stale and isinstance(stale.get("data"), dict):
+                        record_source_call("open_meteo", "multi_model", "stale_cache", (time.perf_counter() - started) * 1000.0)
                        return dict(stale["data"])
+                record_source_call("open_meteo", "multi_model", "cooldown_skip", (time.perf_counter() - started) * 1000.0)
                return None

        with self._multi_model_cache_lock:
@@ -475,6 +501,7 @@ class NwsOpenMeteoSourceMixin:
            ):
                cached_data = cached.get("data")
                if isinstance(cached_data, dict):
+                    record_source_call("open_meteo", "multi_model", "cache_hit", (time.perf_counter() - started) * 1000.0)
                    return dict(cached_data)
        try:
            url = "https://api.open-meteo.com/v1/forecast"
@@ -524,6 +551,7 @@ class NwsOpenMeteoSourceMixin:

            if not daily_forecasts:
                logger.warning("Multi-model: 无有效模型数据")
+                record_source_call("open_meteo", "multi_model", "empty", (time.perf_counter() - started) * 1000.0)
                return None

            # 今天的预报 (向后兼容)
@@ -548,6 +576,7 @@ class NwsOpenMeteoSourceMixin:
                    "data": dict(result),
                }
            self._flush_open_meteo_disk_cache()
+            record_source_call("open_meteo", "multi_model", "success", (time.perf_counter() - started) * 1000.0)
            return result
        except Exception as e:
            status_code = getattr(getattr(e, "response", None), "status_code", None)
@@ -573,6 +602,8 @@ class NwsOpenMeteoSourceMixin:
                if stale and isinstance(stale.get("data"), dict):
                    fallback = dict(stale["data"])
                    fallback["stale_cache"] = True
+                    record_source_call("open_meteo", "multi_model", "stale_cache", (time.perf_counter() - started) * 1000.0)
                    return fallback
+            record_source_call("open_meteo", "multi_model", "error", (time.perf_counter() - started) * 1000.0)
            return None

@@ -5,11 +5,30 @@ import os
 import time

 from loguru import logger
+from src.database.runtime_state import (
+    OpenMeteoCacheRepository,
+    STATE_STORAGE_DUAL,
+    STATE_STORAGE_SQLITE,
+    get_state_storage_mode,
+)
+
+_open_meteo_cache_repo = OpenMeteoCacheRepository()


 class OpenMeteoCacheMixin:
    def _load_open_meteo_disk_cache(self) -> None:
        """启动时从磁盘加载 Open-Meteo 三类缓存，避免重启后冷启动打爆 API"""
+        mode = get_state_storage_mode()
+        if mode == STATE_STORAGE_SQLITE:
+            try:
+                saved = _open_meteo_cache_repo.load_payload(self._disk_cache_max_age_sec)
+                loaded = self._merge_open_meteo_payload(saved)
+                self._disk_cache_last_mtime = _open_meteo_cache_repo.latest_updated_at()
+                if loaded:
+                    logger.info(f"✅ 从 SQLite 加载 Open-Meteo 缓存 {loaded} 条")
+                return
+            except Exception as exc:
+                logger.error(f"SQLite Open-Meteo 缓存加载失败，fallback file: {exc}")
        try:
            path = self._disk_cache_path
            if not os.path.exists(path):
@@ -58,11 +77,26 @@ class OpenMeteoCacheMixin:
            self._disk_cache_last_mtime = current_mtime
            if loaded:
                logger.info(f"✅ 从磁盘加载 Open-Meteo 缓存 {loaded} 条 ({self._disk_cache_path})")
+            if mode == STATE_STORAGE_DUAL:
+                try:
+                    _open_meteo_cache_repo.replace_payload(saved, self._disk_cache_max_age_sec)
+                except Exception as exc:
+                    logger.warning(f"SQLite Open-Meteo 缓存同步失败: {exc}")
        except Exception as exc:
            logger.warning(f"磁盘缓存加载失败（首次启动不影响运行）: {exc}")

    def _maybe_reload_open_meteo_disk_cache(self) -> None:
        """跨进程共享缓存：当缓存文件有更新时增量重载到当前进程内存"""
+        mode = get_state_storage_mode()
+        if mode == STATE_STORAGE_SQLITE:
+            try:
+                latest_updated_at = _open_meteo_cache_repo.latest_updated_at()
+                if latest_updated_at <= self._disk_cache_last_mtime:
+                    return
+                self._load_open_meteo_disk_cache()
+            except Exception:
+                pass
+            return
        try:
            path = self._disk_cache_path
            if not os.path.exists(path):
@@ -76,8 +110,8 @@ class OpenMeteoCacheMixin:

    def _flush_open_meteo_disk_cache(self) -> None:
        """将三类 Open-Meteo 内存缓存持久化到磁盘"""
+        mode = get_state_storage_mode()
        try:
-            os.makedirs(os.path.dirname(self._disk_cache_path), exist_ok=True)
            with self._open_meteo_cache_lock:
                forecast_snapshot = dict(self._open_meteo_cache)
            with self._ensemble_cache_lock:
@@ -90,15 +124,50 @@ class OpenMeteoCacheMixin:
                "multi_model": multi_model_snapshot,
                "saved_at": time.time(),
            }
-            with self._disk_cache_lock:
-                tmp_path = self._disk_cache_path + ".tmp"
-                with open(tmp_path, "w", encoding="utf-8") as f:
-                    json.dump(payload, f)
-                os.replace(tmp_path, self._disk_cache_path)
-            self._disk_cache_last_mtime = os.path.getmtime(self._disk_cache_path)
+            if mode in {STATE_STORAGE_DUAL, STATE_STORAGE_SQLITE}:
+                _open_meteo_cache_repo.replace_payload(payload, self._disk_cache_max_age_sec)
+                self._disk_cache_last_mtime = _open_meteo_cache_repo.latest_updated_at()
+            if mode != STATE_STORAGE_SQLITE:
+                os.makedirs(os.path.dirname(self._disk_cache_path), exist_ok=True)
+                with self._disk_cache_lock:
+                    tmp_path = self._disk_cache_path + ".tmp"
+                    with open(tmp_path, "w", encoding="utf-8") as f:
+                        json.dump(payload, f)
+                    os.replace(tmp_path, self._disk_cache_path)
+                self._disk_cache_last_mtime = max(
+                    self._disk_cache_last_mtime,
+                    os.path.getmtime(self._disk_cache_path),
+                )
        except Exception as exc:
            logger.warning(f"磁盘缓存写入失败: {exc}")

+    def _merge_open_meteo_payload(self, saved: dict) -> int:
+        now = time.time()
+        max_age = max(600, self._disk_cache_max_age_sec)
+        loaded = 0
+        with self._open_meteo_cache_lock:
+            for key, entry in (saved.get("forecast", {}) or {}).items():
+                if now - float(entry.get("t", 0)) < max_age:
+                    old = self._open_meteo_cache.get(key)
+                    if old is None or float(entry.get("t", 0)) >= float(old.get("t", 0)):
+                        self._open_meteo_cache[key] = entry
+                        loaded += 1
+        with self._ensemble_cache_lock:
+            for key, entry in (saved.get("ensemble", {}) or {}).items():
+                if now - float(entry.get("t", 0)) < max_age:
+                    old = self._ensemble_cache.get(key)
+                    if old is None or float(entry.get("t", 0)) >= float(old.get("t", 0)):
+                        self._ensemble_cache[key] = entry
+                        loaded += 1
+        with self._multi_model_cache_lock:
+            for key, entry in (saved.get("multi_model", {}) or {}).items():
+                if now - float(entry.get("t", 0)) < max_age:
+                    old = self._multi_model_cache.get(key)
+                    if old is None or float(entry.get("t", 0)) >= float(old.get("t", 0)):
+                        self._multi_model_cache[key] = entry
+                        loaded += 1
+        return loaded
+
    def _wait_open_meteo_slot(self, endpoint: str) -> None:
        """Simple per-process rate gate for Open-Meteo endpoints."""
        min_interval = self._open_meteo_min_interval_sec
@@ -0,0 +1,500 @@
+
+from __future__ import annotations
+
+import json
+import os
+import sqlite3
+import threading
+import time
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+from loguru import logger
+
+from src.database.db_manager import DBManager
+
+STATE_STORAGE_FILE = "file"
+STATE_STORAGE_DUAL = "dual"
+STATE_STORAGE_SQLITE = "sqlite"
+VALID_STATE_STORAGE_MODES = {
+    STATE_STORAGE_FILE,
+    STATE_STORAGE_DUAL,
+    STATE_STORAGE_SQLITE,
+}
+
+_LOGGED_MODES: set[str] = set()
+
+
+def get_state_storage_mode() -> str:
+    raw = str(os.getenv("POLYWEATHER_STATE_STORAGE_MODE") or STATE_STORAGE_DUAL).strip().lower()
+    if raw not in VALID_STATE_STORAGE_MODES:
+        logger.warning(
+            f"invalid POLYWEATHER_STATE_STORAGE_MODE={raw!r}, fallback to {STATE_STORAGE_DUAL}"
+        )
+        raw = STATE_STORAGE_DUAL
+    if raw not in _LOGGED_MODES:
+        logger.info(f"runtime state storage mode={raw}")
+        _LOGGED_MODES.add(raw)
+    return raw
+
+
+class RuntimeStateDB:
+    _instance: Optional["RuntimeStateDB"] = None
+    _instance_lock = threading.Lock()
+
+    def __init__(self, db_path: Optional[str] = None):
+        self.db_path = DBManager(db_path).db_path
+        self._init_tables()
+
+    @classmethod
+    def instance(cls) -> "RuntimeStateDB":
+        with cls._instance_lock:
+            if cls._instance is None:
+                cls._instance = cls()
+            return cls._instance
+
+    def connect(self) -> sqlite3.Connection:
+        conn = sqlite3.connect(self.db_path)
+        conn.row_factory = sqlite3.Row
+        return conn
+
+    def _init_tables(self) -> None:
+        with self.connect() as conn:
+            conn.execute(
+                """
+                CREATE TABLE IF NOT EXISTS daily_records_store (
+                    city TEXT NOT NULL,
+                    target_date TEXT NOT NULL,
+                    actual_high REAL,
+                    deb_prediction REAL,
+                    mu REAL,
+                    updated_at REAL NOT NULL,
+                    payload_json TEXT NOT NULL,
+                    PRIMARY KEY (city, target_date)
+                )
+                """
+            )
+            conn.execute(
+                """
+                CREATE TABLE IF NOT EXISTS telegram_alert_last_by_city (
+                    city TEXT PRIMARY KEY,
+                    signature TEXT,
+                    trigger_key TEXT,
+                    severity TEXT,
+                    ts INTEGER,
+                    active INTEGER DEFAULT 0,
+                    cleared_ts INTEGER,
+                    evidence_json TEXT
+                )
+                """
+            )
+            conn.execute(
+                """
+                CREATE TABLE IF NOT EXISTS telegram_alert_signature_state (
+                    signature TEXT PRIMARY KEY,
+                    ts INTEGER NOT NULL
+                )
+                """
+            )
+            conn.execute(
+                """
+                CREATE TABLE IF NOT EXISTS probability_training_snapshots_store (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    city TEXT NOT NULL,
+                    target_date TEXT NOT NULL,
+                    timestamp TEXT NOT NULL,
+                    raw_mu REAL,
+                    raw_sigma REAL,
+                    max_so_far REAL,
+                    peak_status TEXT,
+                    probability_mode TEXT,
+                    legacy_top_bucket INTEGER,
+                    shadow_top_bucket INTEGER,
+                    payload_json TEXT NOT NULL
+                )
+                """
+            )
+            conn.execute(
+                "CREATE INDEX IF NOT EXISTS idx_probability_snapshot_city_date ON probability_training_snapshots_store(city, target_date, id DESC)"
+            )
+            conn.execute(
+                """
+                CREATE TABLE IF NOT EXISTS open_meteo_cache_store (
+                    source_kind TEXT NOT NULL,
+                    cache_key TEXT NOT NULL,
+                    updated_at REAL NOT NULL,
+                    expires_at REAL,
+                    payload_json TEXT NOT NULL,
+                    PRIMARY KEY (source_kind, cache_key)
+                )
+                """
+            )
+            conn.execute(
+                "CREATE INDEX IF NOT EXISTS idx_open_meteo_cache_expires ON open_meteo_cache_store(source_kind, expires_at)"
+            )
+            conn.commit()
+
+
+class DailyRecordRepository:
+    def __init__(self, db: Optional[RuntimeStateDB] = None):
+        self.db = db or RuntimeStateDB.instance()
+
+    def load_all(self) -> Dict[str, Dict[str, Dict[str, Any]]]:
+        out: Dict[str, Dict[str, Dict[str, Any]]] = {}
+        with self.db.connect() as conn:
+            rows = conn.execute(
+                "SELECT city, target_date, payload_json FROM daily_records_store ORDER BY city, target_date"
+            ).fetchall()
+        for row in rows:
+            try:
+                payload = json.loads(row["payload_json"])
+            except Exception:
+                continue
+            city = str(row["city"])
+            date_str = str(row["target_date"])
+            out.setdefault(city, {})[date_str] = payload
+        return out
+
+    def upsert_record(self, city: str, target_date: str, record: Dict[str, Any]) -> None:
+        payload_json = json.dumps(record, ensure_ascii=False)
+        updated_at = time.time()
+        with self.db.connect() as conn:
+            conn.execute(
+                """
+                INSERT INTO daily_records_store (
+                    city, target_date, actual_high, deb_prediction, mu, updated_at, payload_json
+                ) VALUES (?, ?, ?, ?, ?, ?, ?)
+                ON CONFLICT(city, target_date) DO UPDATE SET
+                    actual_high = excluded.actual_high,
+                    deb_prediction = excluded.deb_prediction,
+                    mu = excluded.mu,
+                    updated_at = excluded.updated_at,
+                    payload_json = excluded.payload_json
+                """,
+                (
+                    city,
+                    target_date,
+                    record.get("actual_high"),
+                    record.get("deb_prediction"),
+                    record.get("mu"),
+                    updated_at,
+                    payload_json,
+                ),
+            )
+            conn.commit()
+
+    def replace_all(self, data: Dict[str, Dict[str, Dict[str, Any]]]) -> int:
+        count = 0
+        with self.db.connect() as conn:
+            conn.execute("DELETE FROM daily_records_store")
+            for city, city_rows in (data or {}).items():
+                if not isinstance(city_rows, dict):
+                    continue
+                for target_date, record in city_rows.items():
+                    payload_json = json.dumps(record, ensure_ascii=False)
+                    conn.execute(
+                        """
+                        INSERT INTO daily_records_store (
+                            city, target_date, actual_high, deb_prediction, mu, updated_at, payload_json
+                        ) VALUES (?, ?, ?, ?, ?, ?, ?)
+                        """,
+                        (
+                            city,
+                            target_date,
+                            record.get("actual_high"),
+                            record.get("deb_prediction"),
+                            record.get("mu"),
+                            time.time(),
+                            payload_json,
+                        ),
+                    )
+                    count += 1
+            conn.commit()
+        return count
+
+    def delete_older_than(self, cutoff_date: str) -> int:
+        with self.db.connect() as conn:
+            cur = conn.execute(
+                "DELETE FROM daily_records_store WHERE target_date < ?",
+                (cutoff_date,),
+            )
+            conn.commit()
+            return int(cur.rowcount or 0)
+
+
+class TelegramAlertStateRepository:
+    def __init__(self, db: Optional[RuntimeStateDB] = None):
+        self.db = db or RuntimeStateDB.instance()
+
+    def load_state(self) -> Dict[str, Any]:
+        state = {"last_by_city": {}, "by_signature": {}}
+        with self.db.connect() as conn:
+            city_rows = conn.execute(
+                "SELECT city, signature, trigger_key, severity, ts, active, cleared_ts, evidence_json FROM telegram_alert_last_by_city"
+            ).fetchall()
+            sig_rows = conn.execute(
+                "SELECT signature, ts FROM telegram_alert_signature_state"
+            ).fetchall()
+        for row in city_rows:
+            entry = {
+                "signature": row["signature"],
+                "trigger_key": row["trigger_key"],
+                "severity": row["severity"],
+                "ts": row["ts"],
+                "active": bool(row["active"]),
+            }
+            if row["cleared_ts"] is not None:
+                entry["cleared_ts"] = row["cleared_ts"]
+            if row["evidence_json"]:
+                try:
+                    entry["evidence"] = json.loads(row["evidence_json"])
+                except Exception:
+                    pass
+            state["last_by_city"][str(row["city"])] = entry
+        for row in sig_rows:
+            state["by_signature"][str(row["signature"])] = int(row["ts"] or 0)
+        return state
+
+    def save_state(self, state: Dict[str, Any]) -> None:
+        last_by_city = state.get("last_by_city") or {}
+        by_signature = state.get("by_signature") or {}
+        with self.db.connect() as conn:
+            conn.execute("DELETE FROM telegram_alert_last_by_city")
+            conn.execute("DELETE FROM telegram_alert_signature_state")
+            for city, row in last_by_city.items():
+                if not isinstance(row, dict):
+                    continue
+                conn.execute(
+                    """
+                    INSERT INTO telegram_alert_last_by_city (
+                        city, signature, trigger_key, severity, ts, active, cleared_ts, evidence_json
+                    ) VALUES (?, ?, ?, ?, ?, ?, ?, ?)
+                    """,
+                    (
+                        city,
+                        row.get("signature"),
+                        row.get("trigger_key"),
+                        row.get("severity"),
+                        int(row.get("ts") or 0),
+                        1 if row.get("active") else 0,
+                        row.get("cleared_ts"),
+                        json.dumps(row.get("evidence"), ensure_ascii=False)
+                        if row.get("evidence") is not None
+                        else None,
+                    ),
+                )
+            for signature, ts in by_signature.items():
+                conn.execute(
+                    "INSERT INTO telegram_alert_signature_state (signature, ts) VALUES (?, ?)",
+                    (signature, int(ts or 0)),
+                )
+            conn.commit()
+
+    def replace_from_state(self, state: Dict[str, Any]) -> int:
+        self.save_state(state)
+        return len((state.get("last_by_city") or {})) + len((state.get("by_signature") or {}))
+
+
+class ProbabilitySnapshotRepository:
+    def __init__(self, db: Optional[RuntimeStateDB] = None):
+        self.db = db or RuntimeStateDB.instance()
+
+    def append_snapshot(self, payload: Dict[str, Any]) -> None:
+        legacy_top = _top_bucket(payload.get("prob_snapshot"))
+        shadow_top = _top_bucket(payload.get("shadow_prob_snapshot"))
+        with self.db.connect() as conn:
+            conn.execute(
+                """
+                INSERT INTO probability_training_snapshots_store (
+                    city, target_date, timestamp, raw_mu, raw_sigma, max_so_far,
+                    peak_status, probability_mode, legacy_top_bucket, shadow_top_bucket, payload_json
+                ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+                """,
+                (
+                    payload.get("city"),
+                    payload.get("date"),
+                    payload.get("timestamp"),
+                    payload.get("raw_mu"),
+                    payload.get("raw_sigma"),
+                    payload.get("max_so_far"),
+                    payload.get("peak_status"),
+                    payload.get("probability_mode"),
+                    legacy_top,
+                    shadow_top,
+                    json.dumps(payload, ensure_ascii=False),
+                ),
+            )
+            conn.commit()
+
+    def load_recent_rows(self, city: str, target_date: str, limit: int) -> List[Dict[str, Any]]:
+        with self.db.connect() as conn:
+            rows = conn.execute(
+                """
+                SELECT payload_json
+                FROM probability_training_snapshots_store
+                WHERE city = ? AND target_date = ?
+                ORDER BY id DESC
+                LIMIT ?
+                """,
+                (city, target_date, int(limit)),
+            ).fetchall()
+        out = []
+        for row in rows:
+            try:
+                out.append(json.loads(row["payload_json"]))
+            except Exception:
+                continue
+        return out
+
+    def load_all_rows(self) -> List[Dict[str, Any]]:
+        with self.db.connect() as conn:
+            rows = conn.execute(
+                "SELECT payload_json FROM probability_training_snapshots_store ORDER BY id"
+            ).fetchall()
+        out = []
+        for row in rows:
+            try:
+                out.append(json.loads(row["payload_json"]))
+            except Exception:
+                continue
+        return out
+
+    def replace_all(self, rows: List[Dict[str, Any]]) -> int:
+        count = 0
+        with self.db.connect() as conn:
+            conn.execute("DELETE FROM probability_training_snapshots_store")
+            for payload in rows or []:
+                if not isinstance(payload, dict):
+                    continue
+                legacy_top = _top_bucket(payload.get("prob_snapshot"))
+                shadow_top = _top_bucket(payload.get("shadow_prob_snapshot"))
+                conn.execute(
+                    """
+                    INSERT INTO probability_training_snapshots_store (
+                        city, target_date, timestamp, raw_mu, raw_sigma, max_so_far,
+                        peak_status, probability_mode, legacy_top_bucket, shadow_top_bucket, payload_json
+                    ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+                    """,
+                    (
+                        payload.get("city"),
+                        payload.get("date"),
+                        payload.get("timestamp"),
+                        payload.get("raw_mu"),
+                        payload.get("raw_sigma"),
+                        payload.get("max_so_far"),
+                        payload.get("peak_status"),
+                        payload.get("probability_mode"),
+                        legacy_top,
+                        shadow_top,
+                        json.dumps(payload, ensure_ascii=False),
+                    ),
+                )
+                count += 1
+            conn.commit()
+        return count
+
+
+class OpenMeteoCacheRepository:
+    def __init__(self, db: Optional[RuntimeStateDB] = None):
+        self.db = db or RuntimeStateDB.instance()
+
+    def replace_payload(self, payload: Dict[str, Any], max_age: int) -> int:
+        count = 0
+        now = time.time()
+        with self.db.connect() as conn:
+            conn.execute("DELETE FROM open_meteo_cache_store")
+            for source_kind in ("forecast", "ensemble", "multi_model"):
+                bucket = payload.get(source_kind) or {}
+                if not isinstance(bucket, dict):
+                    continue
+                for cache_key, entry in bucket.items():
+                    if not isinstance(entry, dict):
+                        continue
+                    updated_at = float(entry.get("t") or now)
+                    expires_at = updated_at + max_age
+                    conn.execute(
+                        """
+                        INSERT INTO open_meteo_cache_store (
+                            source_kind, cache_key, updated_at, expires_at, payload_json
+                        ) VALUES (?, ?, ?, ?, ?)
+                        """,
+                        (
+                            source_kind,
+                            cache_key,
+                            updated_at,
+                            expires_at,
+                            json.dumps(entry, ensure_ascii=False),
+                        ),
+                    )
+                    count += 1
+            conn.commit()
+        return count
+
+    def load_payload(self, max_age: int) -> Dict[str, Any]:
+        now = time.time()
+        payload: Dict[str, Any] = {
+            "forecast": {},
+            "ensemble": {},
+            "multi_model": {},
+            "saved_at": now,
+        }
+        with self.db.connect() as conn:
+            rows = conn.execute(
+                "SELECT source_kind, cache_key, updated_at, payload_json FROM open_meteo_cache_store"
+            ).fetchall()
+        for row in rows:
+            updated_at = float(row["updated_at"] or 0)
+            if now - updated_at >= max(600, max_age):
+                continue
+            try:
+                entry = json.loads(row["payload_json"])
+            except Exception:
+                continue
+            payload.setdefault(str(row["source_kind"]), {})[str(row["cache_key"])] = entry
+        return payload
+
+    def latest_updated_at(self) -> float:
+        with self.db.connect() as conn:
+            row = conn.execute(
+                "SELECT MAX(updated_at) AS max_updated_at FROM open_meteo_cache_store"
+            ).fetchone()
+        if not row:
+            return 0.0
+        try:
+            return float(row["max_updated_at"] or 0.0)
+        except Exception:
+            return 0.0
+
+
+def _top_bucket(snapshot: Optional[List[Dict[str, Any]]]) -> Optional[int]:
+    best_value = None
+    best_prob = -1.0
+    for row in snapshot or []:
+        if not isinstance(row, dict):
+            continue
+        value = row.get("v")
+        if value is None:
+            value = row.get("value")
+        try:
+            ivalue = int(value)
+        except Exception:
+            continue
+        prob = row.get("p")
+        if prob is None:
+            prob = row.get("probability")
+        try:
+            fprob = float(prob)
+        except Exception:
+            continue
+        if fprob > best_prob:
+            best_prob = fprob
+            best_value = ivalue
+    return best_value
+
+
+def get_runtime_data_dir() -> str:
+    raw = str(os.getenv("POLYWEATHER_RUNTIME_DATA_DIR") or "").strip()
+    if raw:
+        return raw
+    project_root = Path(__file__).resolve().parents[2]
+    return str(project_root / "data")
@@ -0,0 +1,139 @@
+from __future__ import annotations
+
+import threading
+from typing import Dict, Iterable, List, Optional, Tuple
+
+
+LabelTuple = Tuple[Tuple[str, str], ...]
+
+
+class _MetricsRegistry:
+    def __init__(self) -> None:
+        self._lock = threading.Lock()
+        self._counters: Dict[Tuple[str, LabelTuple], float] = {}
+        self._gauges: Dict[Tuple[str, LabelTuple], float] = {}
+        self._histograms: Dict[Tuple[str, LabelTuple], Dict[str, float]] = {}
+
+    @staticmethod
+    def _normalize_labels(labels: Dict[str, object]) -> LabelTuple:
+        return tuple(
+            sorted((str(key), str(value)) for key, value in labels.items() if value is not None)
+        )
+
+    def inc_counter(self, name: str, amount: float = 1.0, **labels: object) -> None:
+        key = (name, self._normalize_labels(labels))
+        with self._lock:
+            self._counters[key] = self._counters.get(key, 0.0) + amount
+
+    def set_gauge(self, name: str, value: float, **labels: object) -> None:
+        key = (name, self._normalize_labels(labels))
+        with self._lock:
+            self._gauges[key] = value
+
+    def observe(self, name: str, value: float, **labels: object) -> None:
+        key = (name, self._normalize_labels(labels))
+        with self._lock:
+            bucket = self._histograms.setdefault(
+                key,
+                {"count": 0.0, "sum": 0.0, "max": 0.0},
+            )
+            bucket["count"] += 1.0
+            bucket["sum"] += value
+            bucket["max"] = max(bucket["max"], value)
+
+    def snapshot(self) -> Dict[str, object]:
+        with self._lock:
+            return {
+                "counters": dict(self._counters),
+                "gauges": dict(self._gauges),
+                "histograms": {
+                    key: dict(value) for key, value in self._histograms.items()
+                },
+            }
+
+    def export_prometheus(self) -> str:
+        snap = self.snapshot()
+        lines: List[str] = []
+        for name, labels, value in _iter_metrics(snap["counters"]):
+            lines.append(_prom_line(name, value, labels))
+        for name, labels, value in _iter_metrics(snap["gauges"]):
+            lines.append(_prom_line(name, value, labels))
+        for (name, labels), stats in sorted(snap["histograms"].items()):
+            lines.append(_prom_line(f"{name}_count", stats["count"], labels))
+            lines.append(_prom_line(f"{name}_sum", stats["sum"], labels))
+            lines.append(_prom_line(f"{name}_max", stats["max"], labels))
+        return "\n".join(lines) + ("\n" if lines else "")
+
+
+def _iter_metrics(entries: Dict[Tuple[str, LabelTuple], float]) -> Iterable[Tuple[str, LabelTuple, float]]:
+    for (name, labels), value in sorted(entries.items()):
+        yield name, labels, value
+
+
+def _prom_line(name: str, value: float, labels: LabelTuple) -> str:
+    if labels:
+        def _escape(label_value: str) -> str:
+            return label_value.replace("\\", "\\\\").replace('"', '\\"')
+
+        label_str = ",".join(
+            f'{key}="{_escape(str(val))}"'
+            for key, val in labels
+        )
+        return f"{name}{{{label_str}}} {value}"
+    return f"{name} {value}"
+
+
+METRICS = _MetricsRegistry()
+
+
+def counter_inc(name: str, amount: float = 1.0, **labels: object) -> None:
+    METRICS.inc_counter(name, amount=amount, **labels)
+
+
+def gauge_set(name: str, value: float, **labels: object) -> None:
+    METRICS.set_gauge(name, value=value, **labels)
+
+
+def histogram_observe(name: str, value: float, **labels: object) -> None:
+    METRICS.observe(name, value=value, **labels)
+
+
+def record_source_call(source: str, operation: str, outcome: str, duration_ms: Optional[float] = None) -> None:
+    counter_inc(
+        "polyweather_source_requests_total",
+        source=source,
+        operation=operation,
+        outcome=outcome,
+    )
+    if duration_ms is not None:
+        histogram_observe(
+            "polyweather_source_request_duration_ms",
+            duration_ms,
+            source=source,
+            operation=operation,
+            outcome=outcome,
+        )
+
+
+def build_metrics_summary() -> Dict[str, object]:
+    snapshot = METRICS.snapshot()
+    request_total = 0.0
+    source_total = 0.0
+    source_errors = 0.0
+    for (name, labels), value in snapshot["counters"].items():
+        if name == "polyweather_http_requests_total":
+            request_total += value
+        if name == "polyweather_source_requests_total":
+            source_total += value
+            label_map = dict(labels)
+            if label_map.get("outcome") not in {"success", "cache_hit"}:
+                source_errors += value
+    return {
+        "http_requests_total": int(request_total),
+        "source_requests_total": int(source_total),
+        "source_error_total": int(source_errors),
+    }
+
+
+def export_prometheus_metrics() -> str:
+    return METRICS.export_prometheus()
@@ -9,6 +9,12 @@ from typing import Any, Dict, List, Optional, Tuple

 from loguru import logger

+from src.database.runtime_state import (
+    STATE_STORAGE_DUAL,
+    STATE_STORAGE_SQLITE,
+    TelegramAlertStateRepository,
+    get_state_storage_mode,
+)
 from src.data_collection.city_registry import CITY_REGISTRY
 from src.utils.telegram_chat_ids import get_telegram_chat_ids_from_env

@@ -19,6 +25,7 @@ SEVERITY_RANK = {
    "medium": 2,
    "high": 3,
 }
+_telegram_state_repo = TelegramAlertStateRepository()


 def _env_bool(name: str, default: bool) -> bool:
@@ -184,7 +191,18 @@ def _state_file() -> str:


 def _load_state(path: str) -> Dict[str, Any]:
+    mode = get_state_storage_mode()
+    if mode == STATE_STORAGE_SQLITE:
+        try:
+            return _telegram_state_repo.load_state()
+        except Exception as exc:
+            logger.error(f"failed to load telegram push state from sqlite: {exc}")
    if not os.path.exists(path):
+        if mode == STATE_STORAGE_DUAL:
+            try:
+                return _telegram_state_repo.load_state()
+            except Exception:
+                return {"last_by_city": {}, "by_signature": {}}
        return {"last_by_city": {}, "by_signature": {}}
    try:
        with open(path, "r", encoding="utf-8") as fh:
@@ -199,6 +217,11 @@ def _load_state(path: str) -> Dict[str, Any]:


 def _save_state(path: str, state: Dict[str, Any]) -> None:
+    mode = get_state_storage_mode()
+    if mode in {STATE_STORAGE_DUAL, STATE_STORAGE_SQLITE}:
+        _telegram_state_repo.save_state(state)
+    if mode == STATE_STORAGE_SQLITE:
+        return
    os.makedirs(os.path.dirname(path), exist_ok=True)
    tmp_path = f"{path}.tmp"
    with open(tmp_path, "w", encoding="utf-8") as fh:
@@ -0,0 +1,65 @@
+from src.analysis.probability_rollout import judge_probability_rollout
+
+
+def test_judge_probability_rollout_holds_on_shadow_brier_regression():
+    evaluation_report = {
+        "summary": {
+            "sample_count": 105,
+            "delta": {
+                "crps": -0.09,
+                "mae": 0.0,
+                "bucket_hit_rate": 0.0,
+            },
+        }
+    }
+    shadow_report = {
+        "summary": {
+            "samples": 103,
+            "delta_mae": 0.01,
+            "delta_bucket_hit_rate": 0.01,
+            "delta_bucket_brier": 0.29,
+        },
+        "by_city": {
+            "miami": {
+                "samples": 4,
+                "delta_mae": 0.24,
+                "delta_bucket_hit_rate": -0.5,
+                "delta_bucket_brier": 0.47,
+            }
+        },
+    }
+
+    payload = judge_probability_rollout(evaluation_report, shadow_report)
+
+    assert payload["decision"] == "hold"
+    assert payload["ready_for_primary"] is False
+    assert payload["blocking_reasons"]
+    assert payload["worst_shadow_regressions"][0]["city"] == "miami"
+
+
+def test_judge_probability_rollout_promotes_on_clean_metrics():
+    evaluation_report = {
+        "summary": {
+            "sample_count": 120,
+            "delta": {
+                "crps": -0.08,
+                "mae": 0.0,
+                "bucket_hit_rate": 0.02,
+            },
+        }
+    }
+    shadow_report = {
+        "summary": {
+            "samples": 110,
+            "delta_mae": 0.0,
+            "delta_bucket_hit_rate": 0.01,
+            "delta_bucket_brier": 0.01,
+        },
+        "by_city": {},
+    }
+
+    payload = judge_probability_rollout(evaluation_report, shadow_report)
+
+    assert payload["decision"] == "promote"
+    assert payload["ready_for_primary"] is True
+    assert payload["blocking_reasons"] == []
@@ -1,9 +1,15 @@
 import json
 from pathlib import Path
+import pytest

 from src.analysis.probability_snapshot_archive import append_probability_snapshot


+@pytest.fixture(autouse=True)
+def _force_file_mode(monkeypatch):
+    monkeypatch.setenv("POLYWEATHER_STATE_STORAGE_MODE", "file")
+
+
 def test_append_probability_snapshot_writes_jsonl(tmp_path: Path):
    archive_path = tmp_path / "probability_training_snapshots.jsonl"

@@ -0,0 +1,74 @@
+
+import time
+
+from src.database.runtime_state import (
+    DailyRecordRepository,
+    OpenMeteoCacheRepository,
+    ProbabilitySnapshotRepository,
+    RuntimeStateDB,
+    TelegramAlertStateRepository,
+)
+
+
+def test_daily_record_repository_roundtrip(tmp_path, monkeypatch):
+    monkeypatch.setenv('POLYWEATHER_DB_PATH', str(tmp_path / 'polyweather.db'))
+    repo = DailyRecordRepository(RuntimeStateDB(str(tmp_path / 'polyweather.db')))
+    repo.upsert_record('ankara', '2026-03-20', {'actual_high': 15.2, 'deb_prediction': 14.8, 'mu': 15.0})
+    data = repo.load_all()
+    assert data['ankara']['2026-03-20']['actual_high'] == 15.2
+
+
+def test_telegram_alert_state_repository_roundtrip(tmp_path, monkeypatch):
+    monkeypatch.setenv('POLYWEATHER_DB_PATH', str(tmp_path / 'polyweather.db'))
+    repo = TelegramAlertStateRepository(RuntimeStateDB(str(tmp_path / 'polyweather.db')))
+    state = {
+        'last_by_city': {
+            'ankara': {
+                'signature': 'sig-1',
+                'trigger_key': 'mkt:test',
+                'severity': 'medium',
+                'ts': 123,
+                'active': True,
+                'evidence': {'x': 1},
+            }
+        },
+        'by_signature': {'sig-1': 123},
+    }
+    repo.save_state(state)
+    loaded = repo.load_state()
+    assert loaded == state
+
+
+def test_probability_snapshot_repository_recent_rows(tmp_path, monkeypatch):
+    monkeypatch.setenv('POLYWEATHER_DB_PATH', str(tmp_path / 'polyweather.db'))
+    repo = ProbabilitySnapshotRepository(RuntimeStateDB(str(tmp_path / 'polyweather.db')))
+    repo.append_snapshot({
+        'city': 'ankara',
+        'date': '2026-03-20',
+        'timestamp': '2026-03-20T12:00:00Z',
+        'raw_mu': 15.2,
+        'raw_sigma': 1.1,
+        'max_so_far': 14.9,
+        'peak_status': 'before',
+        'probability_mode': 'emos_shadow',
+        'prob_snapshot': [{'v': 15, 'p': 0.6}],
+        'shadow_prob_snapshot': [{'v': 15, 'p': 0.4}],
+    })
+    rows = repo.load_recent_rows('ankara', '2026-03-20', 5)
+    assert len(rows) == 1
+    assert rows[0]['raw_mu'] == 15.2
+
+
+def test_open_meteo_cache_repository_roundtrip(tmp_path, monkeypatch):
+    monkeypatch.setenv('POLYWEATHER_DB_PATH', str(tmp_path / 'polyweather.db'))
+    repo = OpenMeteoCacheRepository(RuntimeStateDB(str(tmp_path / 'polyweather.db')))
+    payload = {
+        'forecast': {'ankara': {'t': time.time(), 'temp': 15}},
+        'ensemble': {'ankara': {'t': time.time(), 'spread': 1.5}},
+        'multi_model': {},
+        'saved_at': 1000,
+    }
+    repo.replace_payload(payload, 86400)
+    loaded = repo.load_payload(86400)
+    assert loaded['forecast']['ankara']['temp'] == 15
+    assert loaded['ensemble']['ankara']['spread'] == 1.5
@@ -0,0 +1,37 @@
+
+from fastapi.testclient import TestClient
+
+from web.app import app
+
+
+client = TestClient(app)
+
+
+def test_healthz_returns_ok_shape():
+    response = client.get('/healthz')
+    assert response.status_code == 200
+    payload = response.json()
+    assert payload['status'] in {'ok', 'degraded'}
+    assert 'db' in payload
+    assert 'state_storage_mode' in payload
+    assert 'cities_count' in payload
+
+
+def test_system_status_returns_summary_shape():
+    response = client.get('/api/system/status')
+    assert response.status_code == 200
+    payload = response.json()
+    assert 'db' in payload
+    assert 'features' in payload
+    assert 'integrations' in payload
+    assert 'cache' in payload
+    assert 'probability' in payload
+    assert 'rollout' in payload['probability']
+    assert payload['probability']['rollout']['decision']['decision'] in {'hold', 'observe', 'promote'}
+    assert 'cities_count' in payload
+
+
+def test_metrics_endpoint_returns_prometheus_payload():
+    response = client.get('/metrics')
+    assert response.status_code == 200
+    assert 'polyweather_http_requests_total' in response.text
@@ -3,6 +3,9 @@ PolyWeather Web Core Context
 """

 import os
+import sqlite3
+import time
+from datetime import datetime, timezone
 from typing import Dict, Any, Optional

 from fastapi import FastAPI, HTTPException, Request
@@ -16,7 +19,19 @@ from src.data_collection.weather_sources import WeatherDataCollector
 from src.data_collection.city_risk_profiles import CITY_RISK_PROFILES  # noqa: F401
 from src.data_collection.polymarket_readonly import PolymarketReadOnlyLayer
 from src.auth.supabase_entitlement import SUPABASE_ENTITLEMENT, extract_bearer_token
+from src.utils.metrics import (
+    build_metrics_summary,
+    counter_inc,
+    gauge_set,
+    histogram_observe,
+)
+from src.analysis.probability_calibration import (
+    DEFAULT_CALIBRATION_FILE,
+    resolve_probability_engine_mode,
+)
+from src.analysis.probability_rollout import build_rollout_report
 from src.database.db_manager import DBManager
+from src.database.runtime_state import get_state_storage_mode
 from src.payments import PAYMENT_CHECKOUT, PaymentCheckoutError  # noqa: F401
 from src.data_collection.city_registry import CITY_REGISTRY

@@ -65,6 +80,59 @@ SETTLEMENT_SOURCE_LABELS: Dict[str, str] = {
 _cache: Dict[str, Dict] = {}
 CACHE_TTL = 300
 CACHE_TTL_ANKARA = 60
+_PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+_PROBABILITY_EVALUATION_REPORT = os.path.join(
+    _PROJECT_ROOT,
+    "artifacts",
+    "probability_calibration",
+    "evaluation_report.json",
+)
+
+
+@app.middleware("http")
+async def _metrics_middleware(request: Request, call_next):
+    started = time.perf_counter()
+    try:
+        response = await call_next(request)
+    except Exception:
+        duration_ms = (time.perf_counter() - started) * 1000.0
+        counter_inc(
+            "polyweather_http_requests_total",
+            method=request.method,
+            path=request.url.path,
+            status="500",
+        )
+        histogram_observe(
+            "polyweather_http_request_duration_ms",
+            duration_ms,
+            method=request.method,
+            path=request.url.path,
+            status="500",
+        )
+        raise
+
+    duration_ms = (time.perf_counter() - started) * 1000.0
+    status_code = str(response.status_code)
+    counter_inc(
+        "polyweather_http_requests_total",
+        method=request.method,
+        path=request.url.path,
+        status=status_code,
+    )
+    histogram_observe(
+        "polyweather_http_request_duration_ms",
+        duration_ms,
+        method=request.method,
+        path=request.url.path,
+        status=status_code,
+    )
+    return response
+_PROBABILITY_SHADOW_REPORT = os.path.join(
+    _PROJECT_ROOT,
+    "artifacts",
+    "probability_calibration",
+    "shadow_report.json",
+)


 def _env_bool(name: str, default: bool = False) -> bool:
@@ -264,3 +332,97 @@ def _sf(v) -> Optional[float]:
 def _is_excluded_model_name(model_name: str) -> bool:
    normalized = str(model_name or "").strip().lower().replace(" ", "").replace("_", "").replace("-", "")
    return "meteoblue" in normalized
+
+
+def _sqlite_health() -> Dict[str, Any]:
+    try:
+        with sqlite3.connect(_account_db.db_path) as conn:
+            conn.execute("SELECT 1").fetchone()
+        return {"ok": True, "db_path": _account_db.db_path}
+    except Exception as exc:
+        return {"ok": False, "db_path": _account_db.db_path, "error": str(exc)}
+
+
+def _cache_summary() -> Dict[str, Any]:
+    gauge_set("polyweather_api_cache_entries", len(_cache))
+    gauge_set(
+        "polyweather_open_meteo_forecast_cache_entries",
+        len(getattr(_weather, "_open_meteo_cache", {}) or {}),
+    )
+    gauge_set(
+        "polyweather_open_meteo_ensemble_cache_entries",
+        len(getattr(_weather, "_ensemble_cache", {}) or {}),
+    )
+    gauge_set(
+        "polyweather_open_meteo_multi_model_cache_entries",
+        len(getattr(_weather, "_multi_model_cache", {}) or {}),
+    )
+    return {
+        "api_cache_entries": len(_cache),
+        "open_meteo_forecast_entries": len(getattr(_weather, "_open_meteo_cache", {}) or {}),
+        "open_meteo_ensemble_entries": len(getattr(_weather, "_ensemble_cache", {}) or {}),
+        "open_meteo_multi_model_entries": len(getattr(_weather, "_multi_model_cache", {}) or {}),
+    }
+
+
+def _feature_flags_summary() -> Dict[str, Any]:
+    return {
+        "auth_enabled": bool(SUPABASE_ENTITLEMENT.enabled),
+        "auth_required": bool(_SUPABASE_AUTH_REQUIRED),
+        "entitlement_guard_enabled": bool(_ENTITLEMENT_GUARD_ENABLED),
+        "payment_enabled": bool(getattr(PAYMENT_CHECKOUT, "enabled", False)),
+        "state_storage_mode": get_state_storage_mode(),
+    }
+
+
+def _integration_summary() -> Dict[str, Any]:
+    weather_cfg = _config.get("weather", {}) if isinstance(_config, dict) else {}
+    return {
+        "supabase_configured": bool(SUPABASE_ENTITLEMENT.configured),
+        "meteoblue_configured": bool(os.getenv("METEOBLUE_API_KEY")),
+        "telegram_bot_configured": bool((_config.get("telegram", {}) or {}).get("bot_token")),
+        "walletconnect_configured": bool(os.getenv("NEXT_PUBLIC_WALLETCONNECT_PROJECT_ID")),
+        "weather_sources": {
+            "openweather": bool(weather_cfg.get("openweather_api_key")),
+            "wunderground": bool(weather_cfg.get("wunderground_api_key")),
+            "visualcrossing": bool(weather_cfg.get("visualcrossing_api_key")),
+        },
+    }
+
+
+def _probability_summary() -> Dict[str, Any]:
+    rollout = build_rollout_report(
+        _PROBABILITY_EVALUATION_REPORT,
+        _PROBABILITY_SHADOW_REPORT,
+    )
+    return {
+        "engine_mode": resolve_probability_engine_mode(),
+        "calibration_file": os.getenv("POLYWEATHER_PROBABILITY_CALIBRATION_FILE")
+        or DEFAULT_CALIBRATION_FILE,
+        "rollout": rollout,
+    }
+
+
+def build_health_payload() -> Dict[str, Any]:
+    db = _sqlite_health()
+    return {
+        "status": "ok" if db.get("ok") else "degraded",
+        "time_utc": datetime.now(timezone.utc).isoformat(),
+        "db": db,
+        "state_storage_mode": get_state_storage_mode(),
+        "cities_count": len(CITIES),
+    }
+
+
+def build_system_status_payload() -> Dict[str, Any]:
+    return {
+        "status": build_health_payload()["status"],
+        "time_utc": datetime.now(timezone.utc).isoformat(),
+        "db": _sqlite_health(),
+        "features": _feature_flags_summary(),
+        "integrations": _integration_summary(),
+        "cache": _cache_summary(),
+        "metrics": build_metrics_summary(),
+        "probability": _probability_summary(),
+        "cities_count": len(CITIES),
+    }
@@ -4,10 +4,12 @@ import os
 from typing import Optional

 from fastapi import APIRouter, HTTPException, Request
+from fastapi.responses import PlainTextResponse
 from loguru import logger

 from src.analysis.deb_algorithm import load_history
 from src.data_collection.city_registry import ALIASES
+from src.utils.metrics import export_prometheus_metrics
 from web.analysis_service import (
    _analyze,
    _build_city_detail_payload,
@@ -31,6 +33,8 @@ from web.core import (
    _SUPABASE_AUTH_REQUIRED,
    _assert_entitlement,
    _bind_optional_supabase_identity,
+    build_health_payload,
+    build_system_status_payload,
    _require_supabase_identity,
    _resolve_auth_points,
    _resolve_weekly_profile,
@@ -49,6 +53,27 @@ def _normalize_city_or_404(name: str) -> str:
    return city


+@router.get("/healthz")
+async def healthz():
+    payload = build_health_payload()
+    if payload.get("status") != "ok":
+        raise HTTPException(status_code=503, detail=payload)
+    return payload
+
+
+@router.get("/api/system/status")
+async def system_status():
+    return build_system_status_payload()
+
+
+@router.get("/metrics", response_class=PlainTextResponse)
+async def metrics():
+    return PlainTextResponse(
+        export_prometheus_metrics(),
+        media_type="text/plain; version=0.0.4; charset=utf-8",
+    )
+
+
@router.get("/api/cities")
 async def list_cities(request: Request):
    _assert_entitlement(request)