Running and past issues on Kagi Assistant are not reported to status page.

o_be_one

Hello Kagi <3,

yesterday, Kagi Assistant was not working properly, leading to an answer like "Your request can't be proceed". So I wen to https://status.kagi.com/ and nothing was reported here. Then I came to Discord and seen other people talking about the issue.

As I'm curious, techie and I value Kagi a lot, I came later that same day on https://status.kagi.com/ to know more about what happened with the Assistant. This incident has never been reported to the status page.

I report this as a bug as I feel this kind of issue must be reported to status even if it's partially automated. But if you feel it's more a feature request please let me know or, if you can, move this thread to the relevant queue.

Thanks a lot for the value you will give to this report.

I wish to see https://status.kagi.com/ up to date with at least incidents we, as users, have seen or catch. I hope the report to contain some explanation, even short and simple one, like for example "We had to raise up a rate limit from our AI API orchestrator" (most likely not what happened here, it's as an example).

DanielBarney

Hey o_be_one!

I'm Daniel, the SRE at Kagi.
The assistant outage on the 23rd was caused by a broken piece of code that was pushed to stage and production without adequate testing. We have alerts setup to specifically catch services like the assistant being offline, and as soon as we were alerted as to what had happened, we rolled back the deploy to the previous working version.
As a result of this, I have made some changes in our CI/CD pipeline to ensure that this specific problem can not happen again.

These alerts come from a monitoring service that we use to ensure that different components of Kagi are online and working as expected. This service also automatically creates and resolves an incident so at a minimum you all can see that something is having a problem. After we rolled back to the previous working deploy we noticed that an incident was not created. It appears that the assistant monitoring was not setup to create these automated incidents, so we made some changes to ensure thay they are created.

As far as filling out details about the incidents, I will make sure that they get prioritized correctly and appropriate details are reported. I acknowledge that we have been a little lax about adding details to the incidents recently, we will be better about it. It's important to me to be open about problems we are facing and what steps we are taking to fix those problems.

I appreciate the feed back, it makes it so much easier to make a better product when we know what bothers you and where we need to improve.

Thank you,
Daniel

o_be_one

DanielBarney Thank you so much for your perfect explanation! I really appreciate all the effort you did to write this complete answer and your amazing reactivity to it. I was not expecting an answer that quick ˆˆ.
Have a good week!