logoalt Hacker News

jakswa • today at 3:53 PM • 2 replies • view on HN

did they stick with the image-based compaction (when vision is supported i guess)? It boggled my mind: "snapcompact, which archives older history into dense images for a vision-capable model"


Replies

jszymborski • today at 4:22 PM

it's in the docs! never have used it though https://omp.sh/docs/compaction

lejalv • today at 4:33 PM

I use it. It works flawlessly.