AI-based release note check before automatic updates #11
ChrisBGL
started this conversation in
Show and tell
Replies: 1 comment
|
For AI-based release note checks before auto-updates, a simple approach: summarize changelog with an LLM, score for breaking changes (config/API), and require explicit approval if risk is high. Keep it deterministic with rules + LLM fallback. Cache results per version to avoid repeat calls and log decisions. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Following up on the discussion in #9, here is the Home Assistant automation I’m currently using.
It listens for update_manager_announced, fetches the release notes and lets a local LLM analyze them via ai_task.generate_data with structured output.
If the AI finds a concrete breaking change, required migration/manual step or another serious reason not to auto-update, the scheduled installation is cancelled via update_manager.cancel_scheduled_install and a notification is sent.
If the AI cannot make a reliable assessment, the update is not blocked (fail-open), but I get a warning instead.
I’ve tested the normal path, the unreliable-assessment path and an actual veto/cancel successfully. I’m now running it over time with real updates to see how well the approach performs in practice.
Here is the automation I’m using:
All reactions