bestmate rag - Bestmate

Local mode builds a FAISS index on your machine from your twin’s chunks, then answers questions purely from it. No retrieval call leaves the device.

Sync the index

bestmate rag sync

Pulls your twin’s chunks from the cloud and builds the local index. Re-run it when you’ve ingested a meaningful amount of new material — the index is a snapshot, not a live view.

Ask locally

bestmate rag ask "what did we decide about pricing?"

Answered entirely from the on-device index.

When to use it

On a plane, or anywhere without a connection — the index is already local.
When latency matters more than freshness — no network round-trip.
When you want to see retrieval in isolation — local mode strips away the cloud ranking layers, so it’s a clean read on whether the chunk you expected is actually in the corpus.

Local mode answers from the index as it was at the last rag sync. Anything ingested since then is invisible to it until you sync again. When freshness matters, use bestmate ask.