Memory API

Structured user memory management with categories, search, statistics, and batch operations. Routes are mounted at /api/memory (src/services/routes/memoryRoutes.js).

Get User List

GET /api/memory/users

Response

{
  "success": true,
  "data": [
    { "userId": "123456", "memoryCount": 15 }
  ]
}

Get Memory Statistics

GET /api/memory/stats

Response

{
  "success": true,
  "data": {
    "totalMemories": 120,
    "totalUsers": 8,
    "categories": {
      "profile": "Basic Information",
      "preference": "Preferences and Habits",
      "event": "Important Events",
      "relation": "Relationships",
      "topic": "Topic Interests",
      "custom": "Custom"
    }
  }
}

Get Category Definitions

GET /api/memory/categories

Get a User's Memories

GET /api/memory/user/:userId

Query Parameters

Parameter	Type	Description
format	string	tree (tree structure, default) or list
groupId	string	Restrict to a group
category	string	Filter by category
limit	number	Number of entries to return

Search Memories

POST /api/memory/search

Request Body

Parameter	Type	Description
query	string	Search keyword (required)
userId	string	Restrict to a user
groupId	string	Restrict to a group
category	string	Restrict to a category
limit	number	Number of entries (default 20)

Add a Memory

POST /api/memory/user/:userId

Request Body

{
  "content": "The user is a software engineer",
  "category": "profile"
}

Update a Memory

PUT /api/memory/:id

Delete a Memory

DELETE /api/memory/:id

The hard query parameter accepts 'true' / '1' (query parameters are always strings; the bug where the old === 'true' check was always false has been fixed).

Clear a User's Memories

DELETE /api/memory/user/:userId

The hard query parameter has the same semantics as above.

Summarize Memories

POST /api/memory/user/:userId/summarize

Runs an AI summary to merge the user's memories and reduce redundant entries. Summary results use structured lines in the form [category] content (category whitelist: profile / preference / event / relation / topic / custom); reasoning text output by the model is not stored.

Request Body

{
  "useLLM": true,
  "cleanup": true
}

Parameter	Type	Default	Description
useLLM	boolean	true	Whether to use an LLM for the summary
cleanup	boolean	true	Whether to run low-quality memory cleanup after summarizing (skipped when cleanup === false / 'false')
groupId	string	-	Restrict to a group
model	string	-	Specify the summary model

Clean Up Low-Quality Memories

POST /api/memory/user/:userId/cleanup

Cleans up the user's low-quality memories (low confidence / expired / too old / too short) without an LLM call.