Skip to main content

​
Two-layer system

Membria CE is split into:
  1. Client layer: Decision Black Box (DBB) + Decision Surface (DS)
  2. Backend layer: Knowledge Cache Graph (KCG) + DoD + Peaq + Arweave + Gateway consensus
The client captures decision truth locally; the backend handles knowledge escalations and immutable storage.

​
Client layer (DBB + DS)

  • Observes interaction streams (chat, files, tools).
  • Detects decision moments and freezes reasoning.
  • Writes immutable decision records (append-only outcomes).
  • Renders signal cards: open loops, drift, pending outcomes, track record.

​
Backend layer (KCG + gateways)

  • Handles DoD escalations and LLM council reasoning.
  • Stores knowledge payloads on Arweave, with Peaq as the logic/consensus layer.
  • Uses gateway validators to verify knowledge, batch consensus events, and maintain indexes.
  • Exposes a read path for clients via SCR bundles and cache-first retrieval.

​
Data models

  • DecisionEvent (client): immutable decision record, no raw chat.
  • KnowledgeEvent (backend): validated knowledge payload with ArTxID + Peaq event linkage.
  • Outcome (client): append-only results; does not mutate decision record.

​
End-to-end flow

  1. Client captures a decision (DBB) and renders signals (DS).
  2. If confidence is low or external knowledge is needed, the client triggers DoD.
  3. Backend council produces verified knowledge, writes to KCG (Peaq + Arweave).
  4. Client stores the verified reference without altering the original decision.