Skip to main content
Version: 0.0.5 (stable)

Use a preset

Add entity types to DEFAULT_SCHEMA with a preset. A preset adds new types. It does not add relations or change descriptions. The default relations do not accept the new types.

Before you start​

  • Install Agentic GraphRAG. These steps do not need a database.
  • Choose the types you need. The default schema reference lists them.

Add one preset​

from agrag.common.data_models import DEFAULT_SCHEMA, OPERATIONS

schema = OPERATIONS.apply(DEFAULT_SCHEMA)

print(schema.name)
print([entity.label for entity in schema.entities])

apply returns a new schema. It does not change DEFAULT_SCHEMA.

Add two presets​

from agrag.common.data_models import DEFAULT_SCHEMA, LEGAL, OPERATIONS

schema = LEGAL.apply(OPERATIONS.apply(DEFAULT_SCHEMA))

print(len(schema.entities), len(schema.relations))

The result has 14 entity types and 8 relation types.

Give the schema to a graph​

Pass the new schema to Graph.open in place of DEFAULT_SCHEMA. Search with the same schema, so the engine searches the labels in the graph.

See Ingest documents for a complete Graph.open call.

Limits​

  • Each label adds extraction time. Add only the presets you need.
  • Apply each preset once. Applying the same preset again raises ValueError.
  • A preset cannot reuse a label that the schema already declares.