(LLMs are already very adept at using `jq` so I would think it was preferable to be able to prompt a system that implements querying inside of source code as "this command uses the same format as `jq`")
Just use jq. None of the other ones are as flexible or widespread and you just end up with frustrated users.
I then switched to JavaScript / TypeScript, which I found much better overall: it's understandable to basically every developer, and LLMs are very good at it. So now in my app I have a button wherever a TypeScript snippet is required that asks the LLM for its implementation, and even "weak" models one-shot it correctly 99% of the times.
It's definitely more difficult to set up, though, as it requires a sandbox where you can run the code without fears. In my app I use QuickJS, which works very well for my use case, but might not be performant enough in other contexts.
it might just be a very limited subset?
jawns•1h ago
Alifatisk•1h ago
OrderlyTiamat•1h ago