I created a tool for calling LLMs as part of a side quest to write a semantic commit splitter: https://github.com/esafak/seance
Does anybody have any suggestions for speeding up the dependency installation step (nimble install -d) in https://github.com/esafak/seance/actions/runs/16234771090/job/45843415151?
Ready for review: https://github.com/nim-lang/packages/pull/3074
Could it be because I wrote requires "nim >= 2.0" ?
Yes, but also other packages can do a requires nim and trigger it as well IIRC.
Im looking forward to trying it out!
Neat. I was thinking of doing something similar in Nim, as a replacement to llm.py - with sqlite logging
It's great that the installed binary is only 788KB :-) - the benefits of Nim
I'm guessing the libraries are statically- rather than dynamically linked. I don't know; I'm back from a nim hiatus so you know better than.
At least I fixed the GHA issue with a cache.
Yeah Nim dependencies end up statically compiled in. IMHO 500k is pretty impressive for something with a fair bit of features these days. htop is like 400k right?
Just compile a Rust program sometime. On one IoT project I worked on we ended up with a 40 mb (!!) executable in debug builds. It literally read an adc and spit out udp packets...
Go programs are generally 3-4 mb, but don't ballon up too much. I recall some jar projects back in the day being 100+ mb.