Memory

Users expect AI to remember what they’ve talked about before. Anuma provides two complementary memory systems that work as client-side tools the model can call during conversation.

The memory engine searches past conversations using semantic similarity. Your messages are the memory — no separate extraction step needed. When the model needs to recall something, it searches stored messages and returns the closest matches, even if they use different words than the original conversation.

The memory vault stores curated facts the model saves on behalf of the user — things like names, preferences, and project requirements. Unlike the engine, which treats every message as potential memory, the vault contains only information that was deliberately captured.

Both systems are built into useChatStorage and use the same embedding infrastructure for semantic search.