BetaWeekly digests are a beta — we're trying something new. Feedback welcome.
n8n refines runtime stability while Trigger.dev ships 4.5 and Inngest expands AI tooling
August 17–23, 2026
n8n released a wave of patch updates across all supported lines to harden task runner management and expression evaluation, while Trigger.dev's v4.5.12 brings a mature Helm chart after a rapid release train, and Inngest added AI experiment pages to its dev server.
n8n: steady hardening across all versions
The week's biggest story is n8n's unusually broad maintenance push, touching every active release line. On the 2.36 track, n8n 2.36.5 raised the AI Assistant model verification token limit, a small but welcome fix for users hitting ceilings with long prompts. The 2.36.3 and 2.36.2 releases (linked via the 2.36.2 entry) are more notable: they avoid restarting task runners that are merely slow (a stability fix for busy instances), only initialize the expression engine for commands that actually evaluate expressions, and resolve end-user credentials when loading node parameters — closing a real gap for multi-tenant setups. The 1.123 line received equivalent fixes at 1.123.75, while 2.35 got the same treatment at 2.35.7 and 2.35.4 — the latter also adding a Google Ads fix for the renamed video views metric on API v25.
Two deeper items stand out. The 1.123 series made the VM expression engine the default in 1.123.73, a significant performance change for self-hosters still on the legacy line. And 2.36.0 (linked from its release entry) arrived as a feature release with a batch of AI-builder improvements — aligning HITL approval resume schemas, fixing the Save button in the agent tool modal, and other UX polish for AI workflows. The steady cadence of 1.123 maintenance (also see 1.123.72 and 1.123.74) suggests the team is committed to keeping the older line viable for a long tail of installs.
Trigger.dev: a release train culminating in 4.5.12
Trigger.dev had a characteristically busy week, publishing a dozen Helm chart versions. The meaningful work is in the application release: v4.5.12 is the current stable, and the chart that goes with it — Helm Chart 4.5.12 — is the one you'll want for production. The rapid series of 4.5.0-rc.x charts (from rc.0 through rc.7) and the incremental 4.5.0→4.5.11 chart tags (4.5.0, 4.5.5, 4.5.11) show a team comfortable shipping fixes as soon as they're ready. Self-hosters running on older 4.4.x have chart updates too — 4.4.4 and 4.4.6 are the notable ones. The takeaway: if you've been holding off on the 4.5 upgrade, the current chart is mature enough to adopt.
Inngest and Temporal: AI features and SDK polish
Inngest continued its AI-themed expansion with v1.43.0, adding app filtering to the Functions table in the dashboard — a quality-of-life improvement for larger projects. More interesting is v1.42.0, which introduces Scores and Experiments pages to the dev server (AI section, beta). These pages show how to get started with step.score() and group.experiment(), signaling where Inngest is heading with AI evaluation features even if data only renders in Cloud for now. The dev server will also help you register interest in local support.
Temporal's Go SDK reached v1.48.0 with a few experimental but potentially useful additions: workflow.GetRandomStream for named deterministic pseudorandom values, workflow.IsReadOnly for context path detection, and client options to override the SDK name/version in heartbeats. The latter is handy for fleet operators who want clearer worker attribution. It also exposes GetFirstExecutionRunID for workflow runs — a small but welcome convenience.