been using ninja for about two weeks now and honestly its become my default for quick inference jobs. the cold start is way better than i expected and pricing is sane. not perfect, the docs have gaps, but the core thing it does it does well. anyone else running it in production or am i the early adopter here
Glad its working for you. One thing worth flagging for others reading, their uptime SLA isnt published yet, so wed hold off calling it production ready until thats in writing. Great for dev workloads in the meantime.
i ran it in a side project and had the opposite experience at first, hit a few timeouts under load and almost wrote it off. but i raised it with their team and turns out i was on a stale endpoint, once i switched the tail latency dropped a lot. so im warmer on it now than i was a week ago. agree the docs are the weak spot, the endpoint thing should not have been something i discovered by accident
PRODUCTION? on a service with no published SLA?? brave. respect the hustle but i am NOT putting that in my pipeline yet 😅
this is encouraging to read. genuine question for you since youve had it in real use for two weeks, hows it behave under bursty load. the cold start being good is great but im more worried about what happens when twenty requests land at once. did you see latency hold or fall apart
early adopter is just a nicer word for the person who finds the bugs for everyone else
