Docs / Schemas
Schemas 0.8.0
The 12 schemas in the latest release, released 17 October 2026. What changed
| Schema | Holds | Since | JSON |
|---|---|---|---|
Audio Transcription
audio-transcription
|
Store text transcribed from an audio source. Supports timed segments, speaker identification, and non-verbal events. | Since 0.1.2 | JSON |
Classification
classification
|
Represent the result of a classification. Supports predicted labels, confidence scores, and vocabulary. | Since 0.1.2 | JSON |
Document Extraction
document-extraction
|
Represent the layout and content of a document, including text blocks, geometric coordinates, and page structure. | Since 0.1.2 | JSON |
Embedding
embedding
|
Store a vector and (optionally) the algorithm or embedding model which generated it. Supports both array (dense) and object (sparse) representations. | Since 0.1.2 | JSON |
Entity Extraction
entity-extraction
|
Represent named entities, structural slots, or visual concepts extracted from unstructured data. Links raw evidence (character-offset text spans, document blocks, or geometric regions) to normalized values and business concepts, and can record directed relations between entities. | Since 0.1.2 | JSON |
Generic
generic
|
A flexible schema for storing arbitrary key-value data. | Since 0.1.2 | JSON |
Geolocation
geolocation
|
Represent a geographic feature. Defines a subset of GeoJSON (RFC 7946) optimized for predictable parsing and indexing. It restricts the root to a single 'Feature', supports all fixed-depth geometries (Point, Polygon, Multi-types), and enforces flat properties to prevent recursion and nesting attacks. | Since 0.1.2 | JSON |
LLM Output
llm-output
|
Store a single interaction with an LLM (a Prompt/Response pair). Captures the input prompt, the resulting output, and the configuration parameters used at that moment. | Since 0.1.2 | JSON |
Object Detection
object-detection
|
Represent objects detected within a source. Supports bounding boxes, polygons, hierarchical relationships, and confidence scores. | Since 0.1.2 | JSON |
Ranking
ranking
|
Represent a ranked list of items related to a query, such as search results, retrieved passages, recommendations or near-duplicate matches. | Since 0.8.0 | JSON |
Regression
regression
|
Represent continuous numerical predictions. Supports point estimates, confidence intervals, and time-series forecasting. | Since 0.1.2 | JSON |
Temporal Segmentation
temporal-segmentation
|
Represent labelled time intervals within a media source, such as chapters, scenes, shots, speaker turns or detected events. Segments may overlap, may nest, and need not cover the whole source. | Since 0.7.0 | JSON |
Other releases: 0.7.0, 0.6.0, 0.5.0, 0.4.0, 0.3.0, 0.2.0, 0.1.2. How versions work