consolidate

Consolidate a cluster of redundant memories into a single summary. Original memories get importance set to 0 and are linked to the summary via summarized_by edges.

Parameters

Parameter	Type	Required	Description
ids	number[]	Yes	Memory IDs to consolidate
content	string	Yes	Your written summary of the consolidated knowledge

Return Value

Returns the new summary memory’s ID.

Examples

Consolidate a cluster

{
  "ids": [12, 15, 23, 31],
  "content": "Auth system architecture: JWT (RS256) with 15-min expiry, refresh tokens in httpOnly cookies rotated on each use, keys managed via AWS KMS."
}

What Happens

1. A new summary memory is created with your content and a fresh embedding
2. Original memories get importance set to 0 (they still exist but rank lower in recall)
3. summarized_by edges link originals to the summary for graph traversal

When to Use

• When recall returns many similar memories on the same topic
• During periodic maintenance of a project’s memory space
• When find_clusters suggests consolidation

When NOT to Use

• When memories cover different subsystems — even if semantically similar, separate memories give better recall precision
• When memories are long, detailed references for distinct topics

Related Tools

• find_clusters — Discover groups to consolidate
• merge — Simpler option for 2-3 memories
• organize — Re-topic without consolidating content