Memory Graph
Entities, relationships, embeddings — one unified graph.
Read more →Feature
Time Travel
Two clocks. Four predicates. Yesterday's reasoning, today.
TimeTravelHow it works
Time Travel
The bi-temporal model separates `asserted_at` (when we learned) from `valid_from / valid_to` (when it was true). Every Recall accepts an `as_of` anchor; four edge predicates decide what survives.
- Bi-temporal edges — knowledge time independent of validity time
- Past-anchor recall without rebuilding parquet snapshots
Primitive
TimeTravel
Two clocks. Four predicates. Yesterday's reasoning, today.
In practice
Built for replay, audit, and scale.
Retracted facts stay visible to past-anchor recalls — exactly the behaviour legal defensibility demands.
- Supersession via `valid_to` — no destructive overwrite
- Audit ledger is a query surface, not a separate forensic log
# Declare a Reflex once, fire on every match.
reflex:
name: supplier-risk-spike
watch:
cx: entity(supplier).risk_score > 0.75
fire:
webhook: https://hooks.example.com/supplier-risk
cooldown: 1h
Related
Pair it with these primitives.
Context Packets
The only output shape — provenance assembled in stages.
Read more →Reflexes
Declarative triggers. CX in, webhooks out.
Read more →Try Time Travel on your stack.
Book a 30-minute walkthrough or wire it up via the API.