Try out `rtk-ai/rtk` to reduce LLM token consumption

[jonathanlab]

https://github.com/rtk-ai/rtk

[charlesvien]

It's not bad, I submitted a PR to add graphite support this weekend for some fun in Rust rtk-ai/rtk#290