remember
Store a new memory in Engram. The content is embedded semantically, enabling recall by meaning rather than exact keywords.
Parameters
Section titled “Parameters”| Parameter | Type | Required | Description |
|---|---|---|---|
content | string | Yes | The memory content to store |
project | string | No | Project scope (defaults to current project) |
topic | string | No | Topic category for organization |
importance | number | No | Boost recall ranking (1-5, default: 3) |
parent_id | number | No | ID of parent memory (creates part_of edge) |
is_private | boolean | No | If true, never synced to team graph |
source | string | No | Where the memory came from (e.g., “conversation”, “code-review”) |
expires_in_days | number | No | Auto-expire from recall after N days |
force | boolean | No | Store even if near-duplicate detected |
Return Value
Section titled “Return Value”Returns the new memory’s ID and any duplicate detection warnings.
Examples
Section titled “Examples”Basic memory
Section titled “Basic memory”{ "content": "API uses JWT tokens with 15-minute expiry", "project": "my-project", "topic": "architecture"}Private memory (never synced to team)
Section titled “Private memory (never synced to team)”{ "content": "My personal debugging shortcut: breakpoint at auth.ts:42", "project": "my-project", "is_private": true}Child of a hub memory
Section titled “Child of a hub memory”{ "content": "JWT signing uses RS256 with keys in AWS KMS", "parent_id": 42, "project": "my-project"}Duplicate Detection
Section titled “Duplicate Detection”Engram checks for near-duplicates using embedding distance and term overlap. If a duplicate is found, you’ll see the existing memory and suggestions to update instead. Use force: true to store anyway.