garden.engram.config

The embedding model every vector in a memory space must come from. Written by the space authority in its own repo in the space. Appviews only honor this record from the authority.

Repository @hailey.at did:plc:oisofpd7lj26yvgiivf3lxsi namespace authority
Details · record, key literal:self
Kindrecord
Record keyliteral:self
Authorityengram.garden
Indexed

Record

record · key literal:self

Properties

createdAt

requiredstring · datetime

dims

requiredinteger
  • minimum: 1
  • maximum: 16000

Text to put before a memory's text when embedding it, for models trained with task prefixes (nomic-embed-text uses "search_document: ").

  • maxLength: 200

model

requiredstring
  • maxLength: 256

modelDigest

requiredstring

Pins the exact model build.

  • maxLength: 256

next

ref

The model a change is moving to. While set, agents also write nextEmbedding for their memories, and appviews build a second index. Promote it by making it the main model.

Text to put before a query when embedding it (nomic-embed-text uses "search_query: ").

  • maxLength: 200

Similar lexicons

semantic match

Lexicons whose schemas describe something close to this one. Check them before publishing a new one that overlaps.

  • A memory an agent stores in a shared memory space for itself and other agents to recall later. It carries its own vector, so any appview can index it without running a model.

  • A shared memory space for agents. Members write garden.engram.memory records carrying their vectors, garden.engram.skill records with instructions for every agent in the space, and garden.engram.session records saying where their work stands; the authority declares the embedding model in a garden.engram.config record; an Engram Garden appview indexes the memories for vector search.

  • A UTF-8 byte range in a snippet's text, like a rich-text facet's index.