You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Trigger multi-agent fan-out (reverted from v1.108.0) — BaseTrigger.agentIds?: string[], TriggerFireOptions.dedupeSourceType / dedupeSourceId, the per-agent deliveryDedupMap in trigger-service, and trigger-service.test.ts are gone. The fan-out approach was abandoned in favor of two per-instance Slack triggers with per-agent filters (e.g. Soporte excludes david, Bolba includes own DMs) — they already deliver the same physical message to each subscribed agent without a fan-out primitive, and v1.109.0's new excludeChannelIds / channelIdAllowlist (Slack) and excludeChatIds / chatIdAllowlist (WhatsApp) make this pattern ergonomic. No back-compat shim: agentIds was only a few hours old and not used in any persisted config.
Changed
Slack polling-interval lower bound lowered to 5s (was 10s) — slack-config.ts and slack-polling-client.ts clamp pollingIntervalSec to [5, 600]. The schema description now notes that in search mode (pollingUseSearch), Slack's ~10–30s search-index lag is the real latency floor, so sub-10s values mostly add API calls without delivering faster — they're useful only for the per-channel polling path.
Technical Details
Net diff vs v1.109.0: −307 / +88 lines across 5 source files (3 reverts + 2 polling-clamp tweaks). The size is mostly removed code.
agentId is the single delivery target again; one trigger_events row per fire.
Multi-agent delivery is now expressed at the trigger layer, not the fire layer: define one trigger per instance, scope it with (exclude|allow)ChannelIds (Slack) or (exclude|allow)ChatIds (WhatsApp), and point each at the agent you want.