search

simonpcouch / pal

LLM assistants for R
https://simonpcouch.github.io/pal/
Other
5 stars 3 forks source link

don't regenerate output for every `document()` #5

Closed simonpcouch closed 1 week ago

simonpcouch commented 2 weeks ago

Right now, every time the one calls document(), a bunch of pal requests are sent off—as the package includes more examples, I can imagine this becoming quite costly quite quickly.

For now, I've just been using gpt-4o-mini to keep costs super low (and then intermittently unsetting the option() to document with a better model), but that's not a good workflow long-term.

simonpcouch commented 1 week ago

"Hadley style fake output" may be the answer here😝 https://github.com/hadley/elmer/pull/89