carbon

This burns a ton of tokens, but the basic idea is to create a cache in ./llm/cache and then spawn a bunch of subagents to look in the cache first, and then write to the cache if learn anything new

This has repo has a great prebuilt llm docs sections. This seems like a killer feature for clients who want to roll their own customizations. https://github.com/crbnos/carbon/blob/main/CLAUDE.md https://github.com/crbnos/carbon/tree/main/llm

thanks, here's the code for it: https://github.com/crbnos/carbon/tree/main/apps/academy

these are all excellent questions! i think, interestingly, the answer to all of them is basically -- there's an off the shelf version that's opinionated and good. and if you want to take it a step further, you'd want to just modify the source code. i think open-source gives consultatnts and integrators super powers that they may not be used to -- everything currently has to be done through abstractions on abstractions. if you can just modify the source code, it becomes a little simpler i think. take the PDFs for example. just make them how you want them in your instance of the software using the most popular tools on the planet (i don't mean crystal reports): https://github.com/crbnos/carbon/tree/main/packages/document...

we give people access to the same supabase API as us, but it's scoped to their company with RLS. the docs are autogenerated: https://x.com/barbinbrad/status/1873043714454811100 you can use the API from inside the codebase, or outside of it: https://github.com/crbnos/carbon?tab=readme-ov-file#api