Clojurians Log v2

I experimented a bit with using LLMs to add annotations themselves as a way to try GitHub Models. Might be of interest https://github.com/typedclojure/typedclojure/blob/4edaa0d63bb8ed9b93fa38e97bbd77f14256b39d/typed-clojure-port.prompt.yml

btw there should already be clj-kondo configs for typed clojure on the classpath at https://github.com/typedclojure/typedclojure/tree/main/typed/clj.runtime/resources/clj-kondo.exports/org.typedclojure/typed.clj.runtime, you might need to call --copy-configs https://github.com/clj-kondo/clj-kondo/blob/master/doc/config.md#importing

my biggest friction so far with LLM's + clojure has been paren matching.

I looked into parinfer (matching parens based on indentation, the latter of which LLM's are good at) but it's quite difficult to get a standalone impl that a coding agent can use. Most require editor integration.

I am using clojure-lsp with opencode to prevent bad parens from getting into the code base.

https://github.com/iwillig/csb/blob/main/opencode.json#L30-L44

But sometimes Claude gets mad at the linters and uses python to write the bad clojure to a file. Its a little frustrating.

Thanks for the clj-kondo advice

I was thinking a more positive directive like "if you have a syntax error, try running parinfer on the code and study the output to learn where the syntax error might be"

which tools consume opencode.json ?

looks like you mention clj-paren-repair-claude-hook in the project, maybe that's what I want

clj-paren-repair works pretty well.