[Havoc]

Markdown files pushed into git.

Tried a couple other options and decided flat text is the only thing I’ve got long term confidence in

Briefly ran an obsidian setup on top Of that but leaning more towards simple folder structure.

Expecting that to play well with future LLM tech too. Flat markdown text files should ingest well

[eddd-ddde]

What ideas do you have for LLM tech in this?

Data aggregation? Summarisation? Retrieval? Or maybe some ingestion post-processing?

I'm getting ideas of making a service that works as a git origin that you can just push to update and then access in a web UI to get nice features, maybe a project for a weekend.

[Havoc]

Likely just as a glorified smart search. Maybe load the headings into context and RAG the rest. Not sure - not high on my todo list.

I just want the data to be in right format once o eventually get round to it