Commit Graph

106 Commits

Author SHA1 Message Date
2569718930@qq.com a32a49f3c7 接入 FMI 赫尔辛基-万塔机场 10 分钟实时数据源
新增 fmi_sources.py,通过 FMI Open Data WFS API 获取 Helsinki-Vantaa
机场观测数据(FMISID 100968, WMO 2974)。每 10 分钟更新,参数含温度/
风速/气压。免费无需 API key。集成至高频机场推送体系。

Constraint: FMI WFS 无 rate limit,无需注册
Scope-risk: 中,新增数据源
Tested: pytest 176 passed, FMI 实测 temp=13.1°C wind=9.5kt pressure=1000.9hPa
2026-05-12 19:51:32 +08:00
2569718930@qq.com 574f007607 @
主循环移除动量突变规则,温度急变检测由机场高频循环独立承担

30 分钟主循环不再做 momentum_spike 检测,避免对机场城市产生
冗余告警(含市场分布/AI 建议的格式)。其余 47 城原本就没有
高频机场数据,动量检测也无实际意义。

Constraint: 其余 3 条规则(Ankara DEB、预报突破、暖平流)保持不变
Scope-risk: 低,仅影响主循环告警规则集
Tested: ruff check 通过
@
2026-05-12 18:05:11 +08:00
2569718930@qq.com 12b0c76caf @
砍掉机场快照,改为 per-city 独立告警;新增安卡拉 MGM 17128 高频监控

快照定时推送无实际价值(没变化也报),改为仅温度急变时触发告警。
各城市独立检测、独立冷却,不再捆绑推送。
同时接入安卡拉 Esenboğa 机场 MGM 站点 17128 的实时温度数据。

Constraint: 快照已运行验证格式无误,砍掉不影响现有急变告警功能
Scope-risk: 中低,仅影响高频通道逻辑,主循环不变
Tested: ruff check 通过
@
2026-05-12 17:51:40 +08:00
2569718930@qq.com 64f8ff21ec @
新增机场高频推送:10分钟级温度监控 + DEB预测快报

为首尔/釜山/东京三大机场城市实现高频温度监控通道:
- 新增 airport_obs_log 表积累观测数据,支持趋势检测
- AMOS/JMA 成功后自动写入观测日志
- 新增 airport_rapid_temp_change 告警规则(20min窗口、0.5°C/10min阈值)
- 10分钟间隔高频子循环 + 30分钟机场快照(含 DEB 预测最高温)
- 最高温锁定后自动跳过,快照仅当地 08:00-20:00 发送

Tested: ruff check 通过
@
2026-05-12 17:04:17 +08:00
2569718930@qq.com b3ea8dcfa7 数据链路 P1 修复:ETag 缓存 + 校准漂移检测
P1-5 ETag 支持:
- 后端新增 _etag_middleware:GET /api/* 自动返回 ETag (MD5)
- 支持 If-None-Match 请求头,匹配时返回 304 + 30s Cache-Control
- 前端 cache: no-store → default,浏览器自动处理 ETag/304 节省带宽

P1-6 校准漂移检测:
- probability_calibration.py 新增 check_calibration_drift()
- 对比最近 200 条 daily_records 的 CRPS 与校准基线
- 漂移 >15% 时返回 warning 提示重新训练
- 集成到 /api/system/status 的 probability.drift 字段

Tested: python -m ruff check ., npx tsc --noEmit
2026-05-10 16:54:48 +08:00
2569718930@qq.com 2183c0399d 移除错误定价信号和市场聚焦摘要功能
保留关键市场警报推送(4 条触发规则不变),
移除 mispricing 信号分类/推送、focus digest 评分/聚合/定时推送、
/markets 手动查询等全套逻辑,净删除 1049 行。
2026-05-07 21:42:06 +08:00
2569718930@qq.com 974b55e34f Add full probability distributions to dashboard 2026-04-22 01:43:13 +08:00
2569718930@qq.com 22060efae6 Clamp EMOS calibrated mu to observed max and add tests 2026-04-19 03:56:35 +08:00
2569718930@qq.com 1892d638fa Promote EMOS to the primary probability engine 2026-04-19 03:36:26 +08:00
2569718930@qq.com 9ec86e0504 Group model stack and dedupe DEB families 2026-04-17 00:35:48 +08:00
2569718930@qq.com 611e2d9009 Expand calibration samples and extend training retention 2026-04-16 01:05:43 +08:00
2569718930@qq.com 53d2ae4c10 Retire dual storage mode and default runtime state to sqlite 2026-04-16 00:52:55 +08:00
2569718930@qq.com a5b5711863 Remove AI analysis and rank cities by recent DEB performance 2026-04-05 07:13:00 +08:00
2569718930@qq.com 781c247952 Add ops dashboards for training data and model coverage 2026-04-03 00:57:19 +08:00
2569718930@qq.com 603546f2f9 Align Taipei and Shenzhen settlement sources with Wunderground 2026-04-02 21:46:41 +08:00
2569718930@qq.com 4f017d501b Refine market alert wording and focus digest scheduling 2026-04-01 18:56:39 +08:00
2569718930@qq.com 342ed77283 Document market monitor bot settings and retire wallet activity 2026-04-01 17:53:31 +08:00
2569718930@qq.com 0cc4f4f0d3 Add LightGBM daily high forecasting pipeline 2026-03-29 23:29:17 +08:00
2569718930@qq.com 7237278f5a Generalize NOAA settlement handling across dashboard and docs 2026-03-27 20:58:38 +08:00
2569718930@qq.com 2007ded86d Add HKO-backed Shek Kong settlement and history support 2026-03-27 20:11:08 +08:00
2569718930@qq.com 7de31ec429 Parse Wunderground history observations for settlement highs 2026-03-25 18:38:34 +08:00
2569718930@qq.com 2b0e6a6f67 Add Wunderground as Shenzhen settlement source 2026-03-25 17:00:36 +08:00
2569718930@qq.com 4ac690228e Bootstrap recent METAR history for new cities 2026-03-24 01:02:18 +08:00
2569718930@qq.com 5ac5caf3ef Preserve model forecasts and add gap backfill script 2026-03-23 23:41:56 +08:00
2569718930@qq.com 7adec4e17c Add peak-12h DEB history reference and bump extension to 0.1.5 2026-03-23 23:17:48 +08:00
2569718930@qq.com ace9c71743 Align Taipei settlement to NOAA RCTP across web and extension 2026-03-23 14:39:41 +08:00
2569718930@qq.com c795d90993 Sync dynamic commentary across bot and dashboard 2026-03-21 18:54:23 +08:00
2569718930@qq.com 43749fff7c Unify runtime state in SQLite and add rollout observability 2026-03-20 23:00:07 +08:00
2569718930@qq.com 3196552c78 Archive probability snapshots and wire them into training 2026-03-20 21:30:52 +08:00
2569718930@qq.com 03dcb4329b Refine EMOS calibration and add Chinese training report 2026-03-20 21:17:37 +08:00
2569718930@qq.com 1c84893bed Add CI and shadow calibration reporting 2026-03-20 20:59:30 +08:00
2569718930@qq.com e4cc5b4263 feat: Add initial web dashboard and centralize weather trend analysis logic into a shared engine. 2026-03-20 18:16:21 +08:00
2569718930@qq.com dbf10253a8 feat: Implement multi-source weather data collection with caching, rate limiting, and disk persistence. 2026-03-20 13:35:34 +08:00
2569718930@qq.com f049a69ee3 Fix mispricing alerts for bearish low-Yes signals 2026-03-19 17:38:04 +08:00
2569718930@qq.com b33a741385 Switch Ankara MGM station selection to 17128 airport 2026-03-18 21:01:31 +08:00
2569718930@qq.com 1ae9b55509 Add RP5 forecast scraping support 2026-03-17 23:15:13 +08:00
2569718930@qq.com 54544e40b9 feat: Implement core data models, backend services for weather and market data, and initial dashboard components. 2026-03-17 13:21:14 +08:00
2569718930@qq.com af2561d019 feat: Introduce a runtime coordinator for bot loops and add new payment checkout and confirmation modules. 2026-03-13 22:11:55 +08:00
2569718930@qq.com f4fea03f35 feat: Implement new bot architecture including handlers, services, analysis modules, and comprehensive tests. 2026-03-12 11:44:52 +08:00
2569718930@qq.com 0877efe0c3 feat: Introduce new ai_analyzer and trend_engine modules for comprehensive weather trend and AI analysis. 2026-03-12 10:26:28 +08:00
2569718930@qq.com 20d3e940e0 feat: Add Polymarket read-only data layer, Telegram push utility, and market alert engine. 2026-03-12 10:06:09 +08:00
2569718930@qq.com de6cf68ee7 feat: Implement reconciliation of recent actual high temperatures using METAR data and integrate it into the DEB command. 2026-03-12 02:43:19 +08:00
2569718930@qq.com 1060945d08 feat: Implement Dynamic Ensemble Blending (DEB) algorithm with historical data management, dynamic weight calculation, and accuracy tracking. 2026-03-12 02:32:10 +08:00
2569718930@qq.com db3b1c995a feat: Implement PolyWeather web map API with integrated weather data analysis, market alert engine, and Telegram push utility. 2026-03-11 12:15:09 +08:00
2569718930@qq.com 1958b2764b feat: Implement core PolyWeather application with city weather query service, trend analysis, data collection, and dashboard UI. 2026-03-11 10:49:35 +08:00
2569718930@qq.com d2a40462c5 feat: Add city weather query service with city resolution, forecast processing, and weather summary generation. 2026-03-11 10:30:46 +08:00
2569718930@qq.com b3f46430ad feat: Introduce comprehensive weather data querying, analysis, and display services, integrate Polymarket data collection, and add Telegram notification utilities. 2026-03-11 10:23:33 +08:00
2569718930@qq.com af4bee12f5 feat: Introduce a web frontend with new city API routes and refactor bot city query logic into a dedicated service. 2026-03-11 08:46:32 +08:00
2569718930@qq.com 2b681513ab fix(trend-engine): fallback to non-ensemble probability computation when OM drops to keep UI alive 2026-03-11 05:07:41 +08:00
2569718930@qq.com c5010a8b97 fix: safely fallback to metric unit per city instead of overwriting F to C randomly; allow trend engine to run even if Open-Meteo drops 2026-03-11 04:57:58 +08:00