Memory System

Linea's memory system gives your AI agents persistent, structured memory across workflow runs. Agents can remember facts, user preferences, and past events — and retrieve them later with natural-language search.

How it works

When you send text to the memory ingest endpoint, Linea:

1. Breaks the text into the smallest self-contained facts
2. Checks whether any of those facts conflict with existing memories
3. If a new fact supersedes an old one, the old memory is automatically retired
4. Stores the remaining facts and makes them searchable

This means your agent's memory stays accurate over time — stale facts are replaced, not duplicated.

Memory scopes

Fact types

Memories can be categorized for easier retrieval:

API Endpoints

Ingesting memories

POST /workspaces/:wId/memories/ingest
 
{
  "content": "Alex prefers concise responses and is based in London.",
  "scope": "user",
  "userId": "user_abc"
}

Response:

{
  "memoriesCreated": 2,
  "memoriesUpdated": 0,
  "sessionId": "uuid"
}

Searching memories

POST /workspaces/:wId/memories/search
 
{
  "query": "what does the user prefer?",
  "scope": "user",
  "limit": 10
}

Search uses a combination of semantic similarity and keyword matching to find the most relevant memories.

User profile

Get all memories about a user, grouped by fact type:

GET /workspaces/:wId/memories/profile?userId=user_abc

{
  "facts": [...],
  "preferences": [...],
  "events": [...],
  "profile": [...]
}

Using memory in workflows

The Memory node inside a workflow can read from or write to memory directly — no API calls needed. Facts stored during one run are available in all future runs.