Home | Specification | Spec v1.0 | Context | Primer | Examples | Governance | Limitations | Releases
This document provides an overview of the Recursive-LD standard, its purpose, the underlying data model, and links to the canonical v1.0 specification and context files.
The Recursive-LD specification defines the core vocabulary, structure, and interpretability primitives used to encode reasoning steps, recursive inference, epistemic lineage, and cognitive transparency. Recursive-LD is designed as a foundational substrate for transparent AI architectures, multi-agent systems, and introspective cognitive frameworks.
Recursive-LD builds upon JSON-LD and Linked Data principles. At its core, it introduces a standardized way to represent reasoning layers, recursive steps, interpretability metadata, and epistemic lineage in a machine-readable format.
The model encodes cognitive layers, their inputs, outputs, intent, transformations, and confidence metrics, enabling reasoning visibility across systems.
Each reasoning step may reference a parent step, forming explicit, traceable recursive chains. This allows reconstruction of cognitive sequences and detection of drift.
Recursive-LD makes it possible to encode the origin of reasoning content, including evidence, citations, training data references, and conceptual dependencies.
These components are exhaustively defined in Recursive-LD v1.0.
Below is the simplest valid Recursive-LD object:
{
"@context": "https://recursive-ld.org/context.json",
"@type": "CognitiveTrace",
"layerId": "layer-1",
"intent": "Determine user request",
"input": "User asked for the specification overview",
"output": "Provide summary of specification",
"confidence": 0.93
}
Recursive-LD is designed as an evolving standard. Future versions will expand support for:
Version tracking is maintained on the Releases page.
© Recursive-LD Standard — Maintained by the Recursive Architecture Intelligence Institute (RAI)