JSON Schema
ConceptsJSON Schema is a standard (json-schema.org) for describing the structure and validation rules of JSON data. In DocLD, extraction schemas define the expected output shape: field names, types (string, number, boolean, date, array, object), required/optional, and optional descriptions or instructions. The LLM is prompted to produce JSON that conforms to this schema, yielding structured data.
In DocLD
- Schema definition — Fields, types, and instructions are stored as a schema (often represented or validated with JSON Schema concepts).
- Extraction — The model returns JSON that fits the schema; confidence score and citations are attached per field.
- Validation — Output can be validated against the schema to catch format errors.
Prebuilt schemas are ready-made schemas for common document types; you can customize or create schemas via the API or dashboard.
Related Concepts
JSON Schema describes the shape of structured data from extraction. Schema in DocLD aligns with this idea; field mapping and instructions complete the extraction config.