Usage Guide

Load a schema from JSON text

from ie_schema import IESchema

schema = IESchema.loads(
    """
    {
      "entities": ["gene", "disease"],
      "relations": [
        {"associated_with": {"head": "gene", "tail": "disease"}}
      ]
    }
    """
)

Load from a JSON Schema string

The same IESchema.loads accepts a root JSON Schema object as a string (for example from another tool or hand-written draft). Top-level keys must not match unknown IE ingest fields: native IE JSON rejects unknown top-level keys, so a document with type / properties is treated as JSON Schema, not as an empty IE schema.

from ie_schema import IESchema

schema = IESchema.loads(
    '{"type":"object","properties":{"title":{"type":"string"},"count":{"type":"integer"}}}'
)

Load from dataclass or Pydantic model

IESchema.loads also accepts a stdlib dataclass / Pydantic v2 BaseModel class (or instance). Install the optional dependency first:

uv add 'ie_schema[model]'

from dataclasses import dataclass
from ie_schema import IESchema

@dataclass
class BusinessRecord:
    business_name: str
    score: float

schema = IESchema.loads(BusinessRecord)

Load a schema from disk

schema = IESchema.load("schema.json")

Iterate planned tasks

IESchema is iterable and yields task objects for classification, entity extraction, relation extraction, and structure extraction.

for task in schema:
    print(type(task).__name__)

Render prompt debug string

print(schema.prompt())