Engram Admin icon

Engram Admin

Manage and administer the Engram knowledge graph

Actions19

Hygiene → Backfill Embeddings

AI-generated

Summary

Backfills missing embeddings for entity names, relationship facts, or both in the selected Engram graph. Choose a backend and target, optionally scope work to a group, then set Dry Run off and provide an embedding model to write embeddings; each execution processes at most the configured limit.

Inputs

  • Backend — Storage backend: Embedded (Graphology, the default) or Neo4j (Remote). For embedded storage, Custom Storage Path can specify the graph data directory; leave it empty to use n8n's default storage directory.
  • Group ID — Optionally restrict backfilling to a specific group or session; empty includes all groups.
  • Dry Run — Defaults to enabled. Reports what would change without modifying graph data; disable it to perform the backfill.
  • Backfill Target — Choose Entities and Relationships (default), Entities Only, or Relationships Only.
  • Embedding Model — Model to use when Dry Run is disabled. The property description says it uses the Engram Extraction LLM credential.
  • Backfill Limit — Maximum missing embeddings to process in one execution; defaults to 50 and accepts 1–500.

Output shape

A list of item data describing the backfill result.

The provided source pack shows the operation and helper signatures but omits their bodies, so the precise result fields and whether the limit applies separately to each target are not visible.

Links

Discussion