Hey, this usually happens when another script, webhook, or dashboard change is overwriting your updates after they go through. If tools and config are updated in separate calls, a race condition can also cause the older version to “win” later. I’d recommend logging all update requests with timestamps and checking if anything else is touching those assistant IDs. Also try updating tools and config in one flow instead of multiple steps, and avoid reusing cached configs.
Do you have any background jobs, CI pipelines, or manual edits that might be redeploying these assistants?
@kiryhanchik