# persistent intelligence

Memory Engine — why sites improve over time

A local, persistent memory layer. Scenes, decisions, outcomes — summarised, decayed, and surfaced as priors into the next reasoning pass. No cloud round-trip.

# what's persisted

Eight properties of VAOS memory

💾

Local-first

Storage lives on the edge node. No data exits the site unless explicitly opted-in.

Temporal persistence

Scenes, events, and outcomes indexed by time and by site.

📝

Summaries, not raw

SceneLM-distilled temporal summaries instead of unbounded frame history.

🌊

Decay model

Tunable retention — events fade by importance, recency, and policy.

📚

Configurable retention

Hours, days, months — per event class, per site. GDPR-compatible.

📊

Expectation modeling

Builds priors — what does Monday morning at this loading bay normally look like?

📜

State history

Replayable state. Audit-ready full envelopes for any past decision.

🌐

Federated memory

Opt-in cross-site federation. Share priors without moving raw data.

# why this matters

Sites learn themselves

A camera that ran for six months on a real site is more accurate than the same camera on day one — not because the model retrained, but because the runtime accumulated context. That context lives in the Memory Engine.

No cloud required. Memory persists locally. Federation is opt-in. Sites retain control of their own historical context.

Continuous local intelligence.

The Memory Engine feeds priors into every reasoning pass.

See it feed reasoning OSTVAL loop