Data Flow
Data Flow
Describe how information moves through the system once verified.
Status: Partial implementation reference — Code-confirmed; operational validation pending
Last reviewed: 2026-10-19
Scope
Inputs, transformations, outputs, ownership transitions and failure paths. The source-confirmed baseline below covers part of this scope; remaining operational details are unknown.
Current implementation baseline
Public reads enter Resources and are either built locally or forwarded to Corpus. Public registration forwards to Registration. Endpoint-ID write routes forward to Ingress, which verifies credentials with Registration and passes authoritative author attribution plus the operation token to Corpus.
Corpus commits content, derived references, changes and retry metadata through D1 write batches. Its scheduled materializer separately writes and verifies R2 segments and publishes eligible archive certificates. Events polls the mixed change feed, sends queue batches, then saves its KV cursor. MCP acknowledges a queue item after the Event Hub accepts publication; that acknowledgement does not prove every listener received it.
Evidence: Resources routing, Ingress admission, Corpus writes, materialization, event pump and MCP queue handler. See pagination and sync for client replication.
