How to use knowledge-atoms
Read the catalog over HTTPS
Every artifact is served under stable URLs with correct content-types:
curl https://knowledge-atoms.com/exports/catalog.json
curl https://knowledge-atoms.com/atoms/entity-type/person.json
curl https://knowledge-atoms.com/knowledge-bases/olympus-core.json
curl https://knowledge-atoms.com/schemas/atom-v1.json Build a knowledge graph from atoms
A knowledge base composition assembles entity types, relationship types, and provenance requirements into a complete knowledge graph schema. To instantiate:
// pseudo-code
const kb = await fetch("/knowledge-bases/olympus-core.json").then(r => r.json());
const entityTypes = await Promise.all(
kb.references.entity_types.map(r => fetch(uriToUrl(r.ref)).then(r => r.json()))
);
const relationshipTypes = await Promise.all(
kb.references.relationship_types.map(r => fetch(uriToUrl(r.ref)).then(r => r.json()))
);
const provenance = await fetch(uriToUrl(kb.references.provenance.ref)).then(r => r.json());
const graph = buildKnowledgeGraph({ entityTypes, relationshipTypes, provenance }); Compatibility rules
Rules in /exports/catalog.json (under the rules
key) declare predicates over atoms. Example: a relationship-constraint rule may require
that a given relationship type only connects specific entity types. A composition that violates a
require-effect rule is malformed.
RAG and semantic memory integration
v0.1 ships the catalog and its JSON Schema. Adapters to specific vector databases (Pinecone,
Weaviate, pgvector) and RAG frameworks (LangChain, LlamaIndex) are v0.2 work — they translate
entity-type atoms into database-native schema definitions and confidence-primitive
atoms into scoring models.