Skip to content
nara Documentation

Memory Objects

Records Validated

What are Memory Objects?

Section titled “What are Memory Objects?”

Memory Objects are the records built from your types. Every object ties to a specific type version, so validation and meaning stay in sync.

When you save an object, the platform:

  1. validates content against the type’s JSON Schema, and
  2. generates an embedding from the full content for semantic search.

Anatomy of an object

Section titled “Anatomy of an object”
{
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "key": "TICKET-4221",
  "typeId": "ticket-type-v1-id",
  "content": {
    "ticketId": "TICKET-4221",
    "status": "open",
    "priority": "critical",
    "description": "User cannot access VPN from the London office."
  }
}
  • content – your data; must satisfy the linked type’s schema.
  • key (optional) – a human‑readable identifier (for example an invoice number) for direct lookup. Keys are unique per type version (typeId) within your organization.
  • typeId – the specific type version this object uses.
Section titled “Semantic search”

Objects are indexed for meaning, not just keywords. A query about “internet down in the UK” can match an object that mentions “VPN” and “London” because the embedding captures related concepts.

Identity and scope

Section titled “Identity and scope”
  • Tenant‑scoped – data is isolated to your organization.
  • Keys – use them to “upsert” from external systems; uniqueness is enforced per type version, so you can reuse a key when you migrate to a new type version.