Public reference
RAG Context
Definition
RAG context, in the CoreStory model, is explicitly not the end goal — "brittle RAG over raw files" is named as an anti-pattern to move away from. The target is governed, structured retrieval through MCP: ontology-aware, lifecycle-aware, trust-aware, environment-scoped, contradiction-aware, and provenance-aware — a semantic routing and grounding system rather than a flat retrieval index.
Key properties
- Contrast: classic RAG = embeddings + nearest-neighbor retrieval over raw text
- CoreStory direction = retrieval scoped by classification (audience, trustLevel, lifecycle, authority, ingestionSafety) and served through MCP, not a bare vector index
- MCP integration direction: ontology-aware retrieval, lifecycle-aware filtering, trust-aware retrieval, environment-scoped context, contradiction-aware inference, provenance-aware responses