A twin is the umbrella; context stores are the named compartments under it (“Botwick CLI”, “Trademark Filing”, “Q3 Rebrand”). Questions route to the relevant stores instead of scanning the whole knowledge base, which is what keeps retrieval sharp as the corpus grows.

​
List

Shows each store’s slug and visibility. Run this before ingesting — the slugs are what you pass to --store.

​
Create

The slug must be kebab-case and unique under the twin. Onboarding creates stores automatically — this is the explicit “I want a new bucket” path.

​
Rename

​
Merge

Combine stores that turned out to be the same thing. The target keeps everything; sources get re-tagged.

​
Route — build a pipeline

Auto-route incoming content into a store by source metadata, so you stop sorting by hand. Anything that doesn’t match a rule goes to the suggested pile, which you review with bestmate inbox.

​
Inspect and move sources

move-source is additive unless you pass --from naming the store to remove it from. delete-source cascades to the item’s chunks and memories — there’s no undo.

​
Reorg

Entity-anchored reorganization: stores become clients and companies, and theme stores fold in as labels.
Always read the preview before applying. reorg restructures every store under the twin at once, and --apply without --yes is deliberately still a dry run so a half-typed command can’t reshape your memory.