LoginSign Up
← Back to all sparks
N

n8n

MKT AUTO
Velocity8.8

Workflow automation platform with fair-code license

n8n.io ↗

n8n's latest patches teach queue-mode workers to suspend running executions on shutdown.

ai agentsworkflow automationgovernancequeue modeself-hosted
◆Current state
n8n runs two release lines in parallel (2.43 and 2.44) with a patch landing almost every day. The October 16 pair splits: 2.43.7 makes workers suspend in-flight production executions at shutdown, while 2.44.3 stops Observer detail lines from being discarded. The substantive recent work sits on AI agents: 2.44 extended node-type and credential policies to agent tools, and earlier patches fixed tool-call resumption, Chat Hub in queue mode and Gemini model listing.
◆Where it's heading
The platform is turning admin governance into agent governance, with policies reaching agent tools, while the patch stream increasingly targets self-hosted operators running queue mode: graceful worker shutdown, per-request parsers, Chat Hub fixes. Same-day releases on both supported lines point to a maturing release process for operators who can't jump versions quickly.
◆Prediction
Expect the worker-shutdown behaviour to appear on the 2.44 line soon and 2.45 to push agent governance further; the entries don't show which policy controls come next.

◆Recent moves

  1. 1d ago

    2.43.7 suspends in-flight executions at worker shutdown

    Workers now suspend in-flight production executions when they shut down instead of cutting them off, which matters for anyone scaling or redeploying queue-mode workers. It is a small feature in a patch release, but it fits the line's steady focus on self-hosted operators.

    View source ↗
  2. 1d ago

    2.44.3 keeps Observer detail lines in responses

    A single fix that keeps Observer detail lines rather than discarding the response. Routine maintenance on the 2.44 line.

    View source ↗
  3. 2d ago

    2.43.6 backports per-request XML parser for webhooks

    The 2.43-line backport of the per-request XML parser fix, shipped minutes after 2.44.2. Same-day backports across both supported lines are now routine for n8n.

    View source ↗
  4. 2d ago

    2.44.2 gives each webhook request its own XML parser

    A single core fix: each webhook request body now gets a fresh XML parser instead of sharing one. It is invisible unless you hit the bug, but webhook ingestion is n8n's front door, so the fix lands on both supported lines at once.

    View source ↗
  5. 3d ago

    2.43.5 backports Gemini model picker and chunk() input validation

    A backport to the 2.43 line that adds input validation to the chunk() expression helper and the reordered Gemini model picker. It's maintenance for users who haven't moved to 2.44.

    View source ↗
  6. 3d ago

    n8n 2.44.1 lists Gemini models by name, prioritized first

    Model pickers now show Gemini models by name and list prioritized models first. It's a small UX fix, but it reflects how long the model list has grown as n8n keeps adding providers.

    View source ↗