Add `python app.py snapshot-seed [--date]` (offline clone, Debian /
docker compose exec friendly). When seed fires while full-site snapshot
is running, or last_completed is missing (no_source), record skip as
non-success and retry in 10 minutes until a real seed succeeds; fans
and quiet-window rules unchanged.
Incident: evening full snapshot could be recorded as ok with
already_completed/elapsed=0 after mutual skip with hourly sync, then
next_run jumped to the next day.
- Full-site snapshot never skips because another task is running
- 23:00-01:00 Asia/Shanghai: defer discovery / snapshot_recent /
candidate_cleanup to after 01:00; fans never defers
- Skip records ok=false + skipped=true (not success / not completed)
- Skip/fail reschedule soon (~5m); daily success still next snapshot_at
- TaskControl return value drives next_run; system UI shows skip fields
Show the real percentage on the card, but sort with delta / (baseline + prior) so a small baseline cannot dominate. Same-day publishes stay off the rate board.
Load business config as JSONC (comments + trailing commas) via an
in-repo stripper then stdlib json — no new deps. Default path prefers
config.jsonc, falls back to legacy config.json. Auth write-back targets
only sibling secrets.json so JSONC comments are never wiped; secrets
overlay merges on load. Examples, compose, Docker, docs, favsort and
tools all share the same loader. Phase A only — no DB settings migration.
Move daily snapshot/fans to 23:15 and add 00:01 seed that clones
last_completed ok=1 rows into today without advancing completed.
Evening runs force through seed markers; growth previous is per-aid
adjacent. Home shows short-update clock + 全站数据 date, and 收录视频
counts.
Reuse a pooled httpx client for favsort GET and POST requests, share the raw-request-derived browser headers, and add folder-specific referers for ids, infos and sort. Add transport tests for header parity and anonymous-to-authenticated fallback.
Add httpx and route Bilibili and Telegram requests through pooled clients with keepalive and explicit timeouts. Align Bilibili headers with data/raw-request.txt: Chrome 155 UA, sec-ch-ua, space.bilibili.com origin, folder-specific referer, and priority header. Close replaced clients when credentials change. Add transport contract tests and live read-only validation for nav, folder sync, WBI search, and Telegram getUpdates.
Implement WBI video search with legacy fallback, classification scoring, candidate pool, permanent seen-aid dedupe, blacklist filtering, and batched Telegram notifications. Add write client for fav deal/sort, target-folder selection, insertion position calculation, read-back verification, fav_action locking, retry on failure, ignore and blacklist actions. Wire discovery, candidate cleanup, Telegram long polling, and new discover/candidates CLI commands into the scheduler. Add migration v3 for candidate duration and index stats.
Backfill same-day snapshots for newly added or restored videos, preserve successful fan snapshots on resume, persist incomplete state before forced re-runs, and split collection completion from backup completion so failed backups can retry without API calls. Reject missing data payloads instead of treating them as empty folders, guard against mass-unfav on suspicious empty remote folders, rebuild Bilibili clients when credentials change, recompute primary folder ownership from active relations with per-folder fav_time, preserve original titles for dropped invalid videos, poll config every 60 seconds, and fix favsort verify interval and mixed resource types.
Implement P0 of the reaction dashboard: config loading with hot reload, SQLite migrations with backups, single-writer queue, read-only Bilibili fav client, folder import/diff sync, invalid-resource handling, blacklist seed, scheduler, and JSON redacted logging. Includes offline unit and transport-level contract tests; all existing and new tests pass.
bot-ci
kairee