What it does
The asset registry stores four kinds of memory asset: chat memory, skills, wikis and code graphs. Each is owned by a user in a project, has immutable versions that reference content by address and digest, is shared through visibility and explicit grants, and moves from draft through review to published. Every change writes to a bounded audit trail. The registry stores references and governance, not the content itself.
How it works
POST /v1/memory/assets creates an asset with a name, a summary, a type and a visibility of private (the default), team or restricted. The caller must hold a project credential and must not be an anonymous key; the creator becomes the owner. A team id must name the project organisation, since that is the sole team boundary today. A new asset starts as a draft with no published version.
A version records a content_ref (a content-addressed storage key), a content_digest that must be a canonical 64-character lowercase hexadecimal SHA-256, a schema version, an optional source digest and a change summary. Versions are immutable and numbered by the server, and every mutation carries an expected_revision, so a stale writer gets a conflict instead of overwriting. Publishing moves the current-version pointer; a published asset may advance to a newer version.
Access follows ownership, visibility and grants. The owner has full access. Team visibility lets members of the organisation read and use the asset, and lets organisation owners and admins manage it. Restricted visibility needs an explicit, unexpired grant to a user or a role (owner, admin, member, editor, viewer or authenticated) with a permission of read, use, manage or review; review lets a person approve publication without being able to edit. An asset can hold at most 100 grants.
Assets restored from a backup are quarantined under an unverified owner until a project administrator adopts them. Adoption names a verified organisation member as the new owner, resets visibility to private and leaves restored grants inert, so access has to be granted again deliberately. Each mutation also appends to a durable audit ring that keeps the latest 1,000 entries per project.
Where you reach it
- HTTP
POST and GET /v1/memory/assets; GET, PATCH and DELETE /v1/memory/assets/{id}; POST and GET .../versions; POST .../publish; POST .../adopt; POST and GET .../grants; DELETE .../grants/{grant_id}
Example
curl -s "$KYNETRA_URL/v1/memory/assets" \
-H "Authorization: Bearer $PROJECT_SERVICE_KEY" -H "Content-Type: application/json" \
-d '{ "name": "Refund playbook", "asset_type": "skill", "visibility": "team", "summary": "Steps for handling refunds" }'
curl -s "$KYNETRA_URL/v1/memory/assets/$ASSET_ID/versions" \
-H "Authorization: Bearer $PROJECT_SERVICE_KEY" -H "Content-Type: application/json" \
-d '{ "content_ref": "objects/refund-playbook-v1", "content_digest": "<64 lowercase hex characters>", "expected_revision": 0 }'Limits
- The registry holds references and digests. The bytes behind content_ref live wherever you put them, and the registry does not fetch or verify them.
- Agents do not consume these assets today. Agent bindings and agent visibility exist in the store, but the HTTP routes reject them until verified agent credentials are available.
- The team boundary is the project organisation; finer teams are not modelled.
- The audit ring keeps the latest 1,000 entries per project, so older entries fall off.
- Listing is paged with a cursor and bounded work per page.
Evidence
- On this siteHTTP API reference
- Tests39 store tests and 23 route tests cover ownership and visibility, grants and expiry, the grant cap, revision conflicts, publish and review permissions, restore quarantine and adoption, deletion and audit retention
Test coverage is listed as present in the engine repository at this edition. This page does not report test results.
KynetraDB is in private preview
Request a preview to talk through the limits above against your own workload.