Trigger.dev

Background tasks for long-running AI workloads got two scheduling additions in v4.5.12.

Scheduled tasks now support stable execution windows. Declarative scheduled tasks can define fixed execution windows; the Schedule API now exposes both the nominal CRON time and its assigned execution time. The dashboard surfaces configured windows alongside upcoming assignments, giving operators a clear picture of when tasks will actually run.

Deployments can carry external IDs. trigger.dev deploy --external-id tags a deployment with any identifier — a commit SHA, a CI run ID, or a release tag — so runs triggered by a given version of your app route to that deployment. Per-release tracing and rollbacks become straightforward without additional bookkeeping.

Durable chat agents persist across restarts. The chat.agent primitive maintains conversations in a Session that survives restarts and redeploys. chat.headStart runs the first LLM turn server-side while the agent boots, cutting cold start from 2801ms to 1218ms. AI SDK v7 (ESM-only) is supported alongside v5 and v6.

AI Prompts version templates without redeploying. Define with prompts.define({ id, model, config, variables, content }); every deploy snapshots the version. Dashboard overrides swap text or model per environment, and resolved.toAISDKTelemetry() links each span to its prompt version.

Tasks can reach private VPC resources directly. AWS PrivateLink support lets tasks connect to RDS, ElastiCache, and internal APIs without public endpoints. The platform holds HIPAA Business Associate compliance for workloads that process protected health information. Helm chart releases track each platform version for self-hosted deployments.