Support a random daily run window (random_window_start/end) where the
next auto-run time is derived from a stable per-job hash, avoiding
thundering-herd starts while keeping one run per day. Honors the last
auto-run on initial scheduling so restarts don't double-fire.
Manual triggers can opt into dedupe (dedupe_manual_triggers): an
advisory-locked path collapses concurrent requests onto an existing
pending/running job instead of queueing duplicates. Scheduler error
messages are localized to Chinese.
Register the media_supply_cache_sync job (dry-run aware) and add the
ops migration that seeds it.
The desktop publish queue could stall for a long time and end users assumed
the software was broken. Root cause: a hung adapter held its execution slot
with no wall-clock timeout while auto-renewing its lease forever, so the
server never reclaimed it and every queued task behind it stayed 等待发布.
Auto-recovery also risked silently re-posting a non-idempotent article.
Client (Electron):
- per-task wall-clock deadline + abort; progress-gated lease renewal that
stops and aborts a stalled task instead of renewing it forever
- decouple the concurrency cap from CDP-induced CPU/memory pressure
(admission gate instead of self-throttling collapse); 15s watchdog pump
- all adapter network I/O now has fetch timeouts and honors context.signal;
bounded image-upload concurrency with per-image timeout
- surface live adapter progress, elapsed time, queue position and a
working-vs-queued distinction in the publish view
Server (tenant-api):
- publish lease-recovery worker (every 3m) + supporting index: reclaim
expired in_progress publish leases, requeue (<3 attempts) or terminal-fail
- 3-minute lease TTL with client-presence-gated extension; max 3 attempts
Idempotency (production-grade core):
- durable publish_submit_started_at marker, set before the irreversible
platform submit POST; recovery and abort route a maybe-submitted task to
unknown (manual reconcile, kept in the dedup set) instead of re-posting
- desktop UI requires explicit confirmation before retrying a possibly-
already-published task
Verified: go build/vet/test (incl. resolvePublishRecoveryOutcome), vue-tsc,
vitest 141/141, gofmt all clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extend schedule tasks to dispatch KOL subscription prompts in addition
to prompt rules, add daily/weekly scheduling with fixed or random time
windows, and track dispatch outcomes (last run, status, consecutive
failures).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the scheduler-polled KnowledgeDeletedCleanupWorker with an
event-driven KnowledgeDeletedCleanupEventWorker in tenant-api; cleanup
events are now enqueued transactionally at delete time instead of swept
periodically. Also fixes statement_timeout parameterization in
MonitoringRetentionWorker and propagates errors from cleanupDeletedKnowledgeItem.
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
The state-consistency worker now detects articles still flagged as
generating whose latest generation_task is already failed or completed
(with a version) and aligns the article status to the terminal task,
invalidating article caches on the way out. Adds supporting partial
indexes and switches UpdateGenerationTaskStatus to COALESCE started_at
and completed_at so partial status updates don't wipe the timestamps.
- Replace single Load() with a watching Store that re-reads config files
(and config.local.yaml) and fans out a ReloadEvent with a per-field diff
so consumers can decide whether the change is hot-applicable or requires
a process restart.
- Wrap llm, retrieval, vector store, and object storage clients in
Reloadable* shells so the bootstrap can swap their underlying impls when
config changes without re-instantiating handlers.
- Make jwt.Manager and ops TokenIssuer mutable under a lock so secrets and
TTLs can be rotated live; thread default plan code through a setter on
the ops AdminUserService.
- Wire ConfigStore through bootstrap and every cmd/main.go, scheduler /
worker / tenant-api / ops-api start the watcher; services and handlers
take a config.Provider so they always read current values for things
like generation.stream_enabled, scheduler dispatch, retrieval, etc.
- Switch shared/config decoding off viper to a Kratos-derived runtime
package so env placeholders (\${VAR:default}) resolve consistently and
the same source machinery powers both the loader and the watcher.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replace the legacy target_platform string on schedule tasks with a
workspace-aware auto-publish payload (auto_publish + publish_account_ids
JSONB + cover_asset_url + cover_image_asset_id) and migrate the schema,
sqlc queries, generated models, domain struct, ScheduleTaskService DTOs,
and dispatch worker to round-trip the new fields. PromptRuleGeneration
gains a WithPublishJobService hook so executeGeneration can enqueue an
auto-publish job once the article is ready, and worker-generate wires
the publish-job service in. On the admin-web side, extract
PublishArticleModal's account-card builders into a shared
publish-account-cards module, rebuild GenerateTaskDrawer's schedule
mode around account selection plus a CoverPickerModal, and surface the
new "auto publish" column on ScheduleTaskTab. Shared types and i18n
strings cover the new fields.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Daily monitoring tasks were pinned to a single primary client. If that
client went offline the task wedged even when another desktop client of
the same workspace was online and bound to the same account. Drop the
primary-client constraint on the materialized collect rows, and instead
pick a target per-platform from live account/client presence in Redis,
falling back to the DB client_id when the desktop client is still flagged
online. Desktop task lease for kind=monitor now matches by account
ownership in addition to target_client_id, so any client owning the
account can drain the queue.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Scheduler now runs a token-protected HTTP server on scheduler.http_port
(default 8081, -1 disables) that exposes Prometheus /metrics and the
daily-task JSON snapshot endpoint. Workers gain a blocking Run(ctx) and
are supervised by a WaitGroup: on SIGINT/SIGTERM the metrics server
shuts down first, worker ticks stop scheduling but ongoing runOnce
calls keep their own context and finish, and the process waits up to
60s before exiting. Adds scheduler.http_host/http_port/internal_metrics_token
to config with SCHEDULER_* env overrides and exposes port 8081 in the
Docker image.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduce a generic read-through caching infrastructure and wire it into
all major tenant app services to reduce database load on hot read paths.
Key changes:
- Add `DeletePrefix` to Cache interface with memory (prefix scan) and
Redis (SCAN + DEL) implementations
- New `readthrough.go`: generic `LoadJSON` / `LoadJSONWithEmpty` helpers
backed by singleflight to prevent cache stampedes; supports jittered TTL
- New `cache_support.go`: centralized cache key builders and invalidation
helpers for all entities (workspace, brand, prompt rules, schedule tasks,
articles)
- Wire optional cache into ArticleService, BrandService, WorkspaceService,
PromptRuleService, ScheduleTaskService, TemplateService, MediaService,
PromptGenerateService via `WithCache()` builder pattern
- ScheduleDispatchWorker invalidates schedule task cache after dispatching
- ArticleService gains a new `Detail` endpoint with empty-result caching
- Update cmd entrypoints and transport handlers to propagate cache
Extract monitoring recovery/inspection workers and schedule dispatch
into a dedicated scheduler process. Add worker-generate process for
article generation and template-assist queue consumption. Introduce
shared/schedule runtime for cron-based worker lifecycle management.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
root