Skip to main content

AI Context Core Concepts

What this page is for​

Define how RAGler structures and governs context used by LLM-driven applications.

Concepts​

  • Context is curated, not raw dump.
  • Published data is immutable per operation but replaceable by workflow.
  • Human approval gates changes before publication.
  • Retrieval quality depends on chunk structure and metadata quality.

Operational rules​

  1. Ingest creates draft sessions.
  2. Draft edits happen before publish.
  3. Publish updates collection knowledge state.
  4. Search only targets published content.

Verify​

  • No direct draft access from search endpoints.
  • Session workflow is required before publish.