n8n ships v2.27 while Inngest and Temporal add rerun, sessions, and compression

n8n v2.27 lands with a substantial execution index migration

The headline for n8n this week is n8n v2.27.0, which shipped on Tuesday with a database migration that adds an index to the execution_entity table. n8n warns that on large instances the migration may take several minutes, though the database stays online during the process — a worthwhile tradeoff for faster execution lookups at scale. The release also includes a fix for conditional credential fields in the API, making them optional rather than forbidden, which tightens credential validation without breaking existing workflows.

The rest of the week on the n8n side was a steady stream of patch releases across both the v2.27 and v2.26 lines, as well as the LTS v1.123 branch. A notable fix in n8n v2.27.1 and its backport v2.26.6 suspends query acquisition during database connection recovery — a defensive measure that should prevent cascading failures when the database is under load or briefly unavailable. The Form Trigger Node also got a critical fix in both v2.26.8 and v2.27.3: a default value was added for the authentication parameter to prevent crashes on old workflows that lack it. And on the security front, n8n v1.123.57 and v1.123.56 patched five vulnerabilities in @grpc/grpc-js, hono, and vue-i18n.

Inngest rolls out sessions, rerun APIs, and deprecation guidance

Inngest had a busy week across its server and JS SDK, with three server releases and two SDK bumps. The Inngest Server v1.31.0 is the standout: it introduces session server parsing and emission, plus a new sessions dashboard UI. This is a meaningful step toward giving developers visibility into groups of related runs, which is especially useful for multi-step user journeys. On the operational side, v1.30.0 adds a v2 API rerun endpoint — POST /v2/runs/{run_id}/rerun — along with a CLI shorthand via npx inngest-cli@latest api rerun. This unblocks programmatic retries of failed agentic workflows, a frequent request as Inngest's v2 agent mode matures.

The JS SDK v4.7.0 adds OpenTelemetry metadata to steps containing LLM calls instrumented by Langfuse, which is a nice bridge for teams already using Langfuse for observability. And v4.6.0 deprecates optimizeParallelism: false in favor of the newer group.parallel({ mode: "race" }) syntax, signalling that the SDK is moving toward explicit parallelism semantics. The parallel race mode was already available, but this change makes the deprecation official and pushes users toward the clearer API.

Temporal Go SDK tightens transport and workflow behavior

Temporal Go SDK v1.45.0 enables gRPC gzip compression by default for all client connections, which reduces wire size for requests and allows gzip-compressed responses. For teams operating at scale, this can meaningfully cut network overhead. The SDK also flips two workflow behavior fixes — SDKFlagBlockedSelectorSignalReceive and SDKFlagWorkflowNewChannelLostMessages — on by default for new workflow executions. These flags previously gated fixes for selector/signal receive behavior and workflow channel message handling, so new executions will now get the corrected behavior without manual flagging.

Trigger.dev agent skills and n8n AI Assistant polish

Trigger.dev v4.5.0-rc.7 bundles agent skills and a curated snapshot of the docs those skills reference directly into the @trigger.dev/sdk package. This is a forward-looking move: as AI-assisted development becomes more common, shipping the knowledge base alongside the SDK means agents can answer questions about the platform without needing to fetch documentation at runtime. Over on n8n, v2.26.4 reduced the delay in displaying AI Assistant workflow preview examples, and v2.26.5 added telemetry for the import/export CLI commands — small quality-of-life improvements that round out the week.