Adding an ingestion mechanism

[userbox020]

Sup bro, how can we input repos to the llm context so we can do something like chat with a repo feature?

[turboderp]

RAG is a very big feature that requires a whole framework with an embedding database on the back end along with a retrieval and prompt stuffing mechanism. It needs to be worked into the conversation format, and to work well it really calls for models that are specifically finetuned for retrieval. So it's a little outside the scope of what I can currently manage on my own. PRs are welcome, though, and I can help with integration into the UI and so on.

[userbox020]

you the boss bro, say where to start lets do small steps but solid ones