docs/loop/remember.md

Remember — the second stage of the convergence loop

Remember is what makes Unisona yours. A normal chatbot forgets you the moment the tab closes. Unisona keeps a durable, private memory of what you've told it and what it has learned, and brings the relevant pieces back when they matter.

What it means in Unisona

Memory is append-only: new facts, notes, and results are added to a log rather than overwriting the old — nothing is silently lost, and confidence in a fact can shift over time instead of vanishing. On top of that log sits the CSF archive, a compact, searchable store of everything worth keeping.

When you ask something, Unisona doesn't just read your latest message — it retrieves the notes, past conversations, documents, and profile details that bear on it, and feeds those into the answer. That's why a returning user doesn't have to re-explain themselves. You can browse this memory yourself in Explore and the Knowledge Center.

Crucially, Unisona improves by remembering more and retrieving better — not by retraining a model. Your experience is stored as data you own, not baked into weights.

Where it sits in the loop

Remember takes a fresh observation from Observe and surrounds it with relevant context. That enriched picture is what Reason thinks with.

The convergence loop

Observe → Remember → Reason → Act → Verify → Converge — then back to Observe.

  • Observe — take in the world
  • Remember — recall what matters (you are here)
  • Reason — decide what to do
  • Act — do it with real tools
  • Verify — check it against reality
  • Converge — record the outcome and improve

See the North Star briefing for the whole picture.