agrag.schema.loader.load_schema

load_schema(source:SchemaSource, *, extends:SchemaSource | None = None) -> GraphSchema

Load a schema from any accepted source.

source may be a GraphSchema, a Preset, a dict, a list of entity labels and relations, JSON text, a path to a .json, .yaml, .yml or .ttl file, a Pydantic class or a list of Pydantic classes, or the name of a shipped schema. Turtle (.ttl) import needs the optional schema-formats extra. Shorthand is accepted at every level. The result is always the canonical GraphSchema.

The shipped names are "default", "operations" and "legal". A preset name or a Preset value means that preset applied to the default schema, so "operations" is the schema named default+operations.

A schema with no extends stands alone and replaces the default schema. With extends, the schema builds on a base. A type or relation that the base already defines is an error unless it is marked override: true, and an override may not loosen the base. exclude removes base types or relations, with the patterns that use them. add_patterns adds pairs to a base relation.

Extraction settings such as thresholds are not part of a schema. A key that names one is an error.

Parameters:

• source (SchemaSource) – The schema to load.
• extends (SchemaSource | None) – The base schema, as a shipped name, a Preset, a path, a dict or a GraphSchema. A preset is applied to the default schema. Do not also set extends in source.

Returns:

• GraphSchema – The loaded schema. name defaults to custom (the file stem for a file) and version defaults to "1".

Raises:

• SchemaLoadError – The source cannot be read or is invalid. The message lists every problem with the path to the offending key.