AI Context Strategy
Markdown is the canonical source for documentation. Starlight and Docusaurus are publication surfaces, not the source of truth for AI ingestion.
Ingestion Principles
- AI ingestion should read Markdown source files, not scrape rendered HTML.
- Frontmatter must be preserved and used for filtering.
- Public, member, and internal content must be indexed separately.
- Internal ops docs must never be mixed into a public-facing assistant.
Future Indexes
Planned indexes:
cognimesh_public_docsdevman_agents_public_docsdevman_ops_internal
Required Filters
Future ingestion should filter by visibility, project, systems, ai_context, and status. Internal material should require an explicit internal index target and access-controlled runtime.