Conduits
A conduit is how a fact actually gets recorded. Every connector that reads from your ERP, every field-ops harvester, every document extraction, and every computed reconciliation writes into Civil Memory through a conduit. It’s the pipe - as distinct from the shelf, which is the product the pipe fills.
What travels with a conduit
A conduit is more than a data path. It’s the single place the platform records:
- Entitlement - which tier, and which of your organisation’s permissions, are required for this data to flow at all
- Provenance - where the fact came from, stamped on every record it writes
- Routing - which shelf, or shelves, it feeds
Because these properties live on the conduit rather than scattered across individual writes, a fact’s origin and the rules that governed it are never ambiguous after the fact.
One shelf, several conduits
A shelf can be filled by more than one conduit. A shelf has one owner conduit - the primary source, which sets the shared rules other contributors must agree to - and can have any number of contributor conduits layering additional facts on top. A document-extraction pipeline can, for instance, backfill a shelf that a structured form submission normally owns. When you ask for that shelf, you see facts from every conduit that feeds it, not just the owner.
Next steps
- Shelves - the product a conduit fills
- Connectors - the systems most conduits read from
- Agents - producer agents write to memory the same way a conduit does