Long-term memory
How the agent remembers across sessions — quick notes, a searchable knowledge base, and self-learning.
The agent doesn't start from scratch every conversation. It has a few complementary ways to carry knowledge from one session to the next.
Quick memory notes
The agent has memory tools to store, search, update, and delete short notes it should remember — a fact, a preference, ongoing context. These persist between sessions and let it recall things you told it earlier without you repeating yourself.
Example: "remember that I prefer metric units" — it saves a note and applies it next time.
Knowledge base (GBrain)
For a full, browsable "second brain" — pages, dated timelines, and cited recall across everything you've captured — NevoFlux uses GBrain, a local, private knowledge base stored on your machine. The agent captures into it and recalls from it with the brain skills.
GBrain is off until you enable it in Settings → Knowledge Base. For the full picture — how it works, where your data lives, and the capture/recall/synthesize skills — see the Knowledge Base section.
Self-learning
A self-learning pipeline observes your sessions and, over time, proposes durable updates to the agent's persona ("soul") files — but only after enough confirming evidence. You stay in control of those files in Settings → General.
Personas — and their private memory — can also differ per browser Space. See Space Souls.