ie_schema

Classes

ClassificationTask	Classification task definition with labels and threshold metadata.
EntityExtractionTask	Entity extraction task definition.
IESchema	Information-extraction schema loaded from JSON (or from a dataclass / Pydantic model type).
JSONStructureTask	Structured JSON extraction task with named children.
RelationExtractionTask	Relation extraction task between head and tail entity types.
StructureChild	Child field in a JSONStructureTask.
Task	Base class for all extraction tasks yielded by IESchema.

Module Contents

class ie_schema.ClassificationTask

Bases: Task

Classification task definition with labels and threshold metadata.

property labels: list[str]

Ordered list of class labels.

property multi_label: bool

Whether multiple labels may be assigned.

property task: str

Classification task name.

property threshold: float | None

Optional confidence threshold for the classification.

class ie_schema.EntityExtractionTask

Bases: Task

Entity extraction task definition.

property entities: list[str]

Entity labels that should be extracted.

class ie_schema.IESchema

Information-extraction schema loaded from JSON (or from a dataclass / Pydantic model type).

Build an IESchema from a JSON string with loads() (IE ingest JSON or a root JSON Schema object), from a path with load(), or from a stdlib dataclass / Pydantic v2 BaseModel by passing the class (or an instance) to loads(). Iterating over the object yields task instances in schema order.

Example: >>> import ie_schema >>> _j = ‘{“json_structures”:[{“name”:”Business”,”business_name”:{“dtype”:”str”}}]}’ >>> schema = ie_schema.IESchema.loads(_j) >>> isinstance(schema, ie_schema.IESchema) True >>> len(list(schema)) 1

classmethod load(path: str) → IESchema

Parse an IESchema from a JSON file path.

classmethod loads(input: Any) → IESchema

Parse an IESchema from a JSON string or from a dataclass / Pydantic v2 BaseModel type.

String input must be either IE ingest JSON (top-level keys such as json_structures, entities, …) or a root JSON Schema object (type, properties, …). Unknown top-level keys are rejected for the IE shape so JSON Schema is not misread as an empty ingest.

For a dataclass or BaseModel type (or instance), Pydantic v2 builds JSON Schema (TypeAdapter for dataclasses, model_json_schema() for BaseModel subclasses), which is then parsed like JSON Schema string input.

Example: >>> import ie_schema >>> schema = ie_schema.IESchema.loads(‘{“json_structures”:[{“name”:”Business”,”business_name”:{“dtype”:”str”}}]}’) >>> len(list(schema)) 1

prompt() → str

Render the generated extraction prompt as a debug string.

Example: >>> import ie_schema >>> schema = ie_schema.IESchema.loads(‘{“json_structures”:[{“name”:”Business”,”business_name”:{“dtype”:”str”}}]}’) >>> s = schema.prompt() >>> (“[P]” in s) and (“business_name” in s) True

class ie_schema.JSONStructureTask

Bases: Task

Structured JSON extraction task with named children.

property children: list[StructureChild]

Child fields that belong to this structure.

property name: str

Structure name.

class ie_schema.RelationExtractionTask

Bases: Task

Relation extraction task between head and tail entity types.

property description: str | None

Optional human-readable relation description.

property head: str

Head entity type.

property name: str

Relation name.

property tail: str

Tail entity type.

class ie_schema.StructureChild

Child field in a JSONStructureTask.

property choices: list[str]

Allowed string choices for this property.

property description: str | None

Optional child-field description.

property property: str

Property name for this child field.

class ie_schema.Task

Base class for all extraction tasks yielded by IESchema.