Entities

Memorer automatically extracts structured, queryable entities from unstructured text. Instead of treating everything as text blobs for vector search, every person, place, organization, and preference becomes a typed node in the knowledge graph — with relationships, importance scores, and metadata you can filter on.

How entities work

When you store a memory, Memorer’s extraction pipeline identifies entities in the text:


result = user.remember("Alice is a software engineer at Acme Corp in Seattle")
# Entities extracted: Alice (person), Acme Corp (organization), Seattle (location)
# Relationships: Alice -> works_at -> Acme Corp, Alice -> lives_in -> Seattle

Entity properties

Each entity has:

Property	Description
`id`	Unique identifier
`content`	Entity text content
`type`	Category (person, organization, location, etc.)
`category`	Category grouping
`importance`	Score from 0.0 to 1.0
`community_id`	ID of the detected community cluster
`metadata`	Key-value metadata dict
`created_at`	When the entity was first created
`updated_at`	When the entity was last updated

Querying entities


# List all entities for a user
entities = user.entities.list()
 
for entity in entities:
    print(f"{entity.content} ({entity.type})")
 
# Filter by type
people = user.entities.list(type="person")
 
# Get a specific entity
entity = user.entities.get("entity-uuid")

Entity deduplication

Memorer automatically detects and merges duplicate entities. For example, “Alice”, “alice”, and “Alice Smith” may be merged into a single entity if the context indicates they refer to the same person.