Skip to main content
Importing data is how you put books onto the shelf so the AI can read them later.

​
What this means in practice

Knowledge can be imported through:
  • MCP knowledge-base tools
  • HTTP endpoints
  • larger migration/import workflows in the repo

​
How it actually works

Key repo surfaces:
  • convex/crystal/knowledgeHttp.ts
  • convex/crystal/knowledgeBases.ts
  • packages/mcp-server/README.md
These files show how imports are accepted, stored, and enriched.

​
Commands / examples

Examples:
  • crystal_import_knowledge
  • POST /api/knowledge-bases/:knowledgeBaseId/import

​
Common mistakes

  • assuming imports are only an MCP concern and forgetting the HTTP path
  • documenting import as if it immediately means “fully enriched and searchable” without mentioning follow-up processing
  • mixing conversational memory ingestion and KB import semantics

​
Source of truth

Primary files behind this page:
  • convex/crystal/knowledgeHttp.ts
  • convex/crystal/knowledgeBases.ts
  • packages/mcp-server/README.md