Engine layer
Persistent memory that decides what's worth keeping.
Feed the engine raw text or typed observations and it distils them into a knowledge graph with hybrid semantic + keyword search. Context carries from one session to the next — with provenance on every claim, and none of the prompt bloat.
memory.observe · memory.search · memory.reflect
What you get
Recall that compounds
Context carries from one session to the next instead of resetting each time.
No prompt bloat
The engine keeps what matters and lets the rest go, so prompts stay lean.
Provenance on every claim
Trace any recalled fact back to the observation it came from.
How it works
The loop, three calls.
- STEP 01
Observe
Feed raw text or typed observations — one call, memory.observe.
- STEP 02
Distil
The engine turns observations into a knowledge graph with hybrid semantic + keyword search.
- STEP 03
Recall
Retrieve relevant context, with provenance, in any later session.
memory.observe→memory.search→memory.reflectFeatures
Hybrid search
Semantic + keyword recall over a knowledge graph, not a flat log of chats.
Engine-decided retention
It keeps what matters and lets the rest go, so prompts stay lean.
Provenance on every claim
Trace any recalled fact back to the observation it came from.
FAQ
Questions, answered.
How is this different from RAG?+
RAG retrieves chunks from a static corpus. Memory learns about a subject over time, tracks changing facts, and carries provenance — a lifecycle, not a lookup.
What gets stored?+
The engine decides what's worth keeping from what you observe, distilling it into a queryable graph rather than a flat log.
More engine layers
Predictions
Foresight, not just recall.
Declare any target and the engine predicts it from a subject's history — calibrated, provenanced, and abstaining when the signal is thin. This is the layer mem0 and MemoryLake have no answer for.
lattice.predict (event · numeric · time · anomaly)
Behaviors
From patterns to next-best-action.
The engine mines behavior into emergent patterns and closes the loop: predict, let the agent act, record the outcome, and watch the patterns recalibrate. Decisions and outcomes are first-class.
lattice.mine · learning.recordDecision / recordOutcome
Recall that compounds.
Cross-session persistent memory with hybrid search and a knowledge graph. The engine decides what's worth keeping, so context carries between conversations without bloating your prompts.