Debugging bandwidth usage
i was hoping i could do it with the local version but i guess not. does self hosted actually show the bandwidth usage?
7 Replies
self-hosted logs should have bandwidth usage in them, but unfortunately the logs page doesn't currently surface that. that's on our list of improvements
I'd recommend setting up the Axiom integration if you haven't already, and building some queries with alerts on bandwidth usage (the integration comes with some pre-built queries you can use as examples)
We'd like to make Axiom integrations available on self-hosted as well, but there's no time line for that right now
beauty, i'll look into it. just really need some sort of realtime feedback so i can fix things quickly before blowing through like 5GB in a day while testing.
Noted! We have some designs for the logs page improvements that would help here. Your feedback helps us move that up the prioritization list 🙂
I'm sure you know this already but I'd recommend using indexes instead of filters wherever possible!
so part of this is that i'm almost entirely using llms to generate the code. so a ton of the optimizations don't seem to happen. i'm a dev who can fix it myself, but a lot of my "audience" is non-technical, so trying to get in that headspace if convex is marketing towards that crowd. the optimizations and bandwidth seem to be the first major issue i've had compared to using supabase as a baas. likely because the usage is default more transactional on supabase, vs streaming/realtime stuff.
but i don't think filters are the issue since there aren't any in the codebase
these rules seem to avoid that approch nicely
Thanks for the feedback! Are those the official cursor rules? Yeah they should help a lot
That's good that the LLM is avoiding
filter -- the next thing I would check for is your cache hit rate on your queries. If the cache hit rate is low it could mean that you're requesting data too often
If your app is doing LLM response streaming to the client it may be trickier to send responses to the client efficiently, but I think we have some docs / advice about this. I'll ask some of my colleagues and let you knowAxiom integration just dropped for self-hosted earlier today