hm_ingest

Decompose a block of dense text into individual graph entities (nodes and relationships) in a single LLM pass. This is the most efficient way to import unstructured content into the knowledge graph.

Parameters

Parameter	Type	Default	Description
text	str	required	The dense text to decompose
context	str | None	None	Optional label or context for the ingestion

How it works

1. The text is sent to an LLM that identifies distinct entities, facts, and relationships
2. Each entity is created as a node with an appropriate type and description
3. Relationships between entities are created as edges
4. The user_profile is updated if relevant information is detected

This is fundamentally different from hm_store which creates a single node — hm_ingest can create many nodes and edges from one input.

Rate limiting

hm_ingest has its own rate limit category, separate from read and write operations:

Plan	Ingest RPM
Free	3
Basic	5
Pro	20
Business	40
Enterprise	60

Example

{
  "text": "Sarah Chen joined as CTO in January 2026. She leads the platform team which is building the new checkout API using Redis for caching and PostgreSQL for persistence. The project deadline is Q3 2026.",
  "context": "Team update notes"
}

This single call might create nodes for person_sarah_chen, component_checkout_api, tech_redis, tech_postgresql, and edges connecting them.

CLI equivalent

hm ingest "Sarah Chen joined as CTO in January 2026..." --context "Team update notes"

REST endpoint

POST /api/v1/memory/ingest