Skip to main content
Want to help your users filter search results more effectively? You can add or update structured metadata on any existing knowledge base — no need to recreate it.

​
Opening the metadata editor

Click Configure Metadata in the top-right area of the knowledge base detail page. Knowledge base detail page with Configure Metadata button highlighted This opens the same metadata configuration dialog used during knowledge base creation.

​
What you can configure

​
Schema properties

Define what metadata fields the AI should extract from each document. Each property has:
FieldPurpose
Property NameThe name of the metadata field (e.g., category, author, document_type)
TypeThe data type — String, Number, Boolean, Object, Array, or Enum
DescriptionA description to guide the AI on what to extract from each document
Click + Add Property to add fields, or remove existing ones you no longer need. Switch between Builder mode (visual editor) and JSON mode (read-only view of the generated schema — useful for copying into API calls or external tools).

​
Advanced configuration

Fine-tune how the AI extracts metadata to get more accurate, consistent results:
  • LLM provider and Model: Choose which AI model performs the extraction (e.g., OpenAI / gpt-4.1-mini).
  • Schema description: Give the AI context about what kinds of documents it will process and what the metadata is for.
  • Extraction instructions: Set specific rules for how values should be extracted (e.g., “Always extract dates in ISO 8601 format” or “If the author is not explicitly stated, leave the field empty”).
  • Query instructions: Tell the search system how to apply metadata filters when users search — this ensures filters behave the way your users expect.

​
Context configuration

These options require Advanced Document Analysis to be enabled in the knowledge base’s permanent settings.
Give the AI more surrounding context to improve extraction accuracy:
  • Document item context: Choose how much of each document the AI sees — Short Summary (fast), Long Summary (more accurate), or Full Document (most accurate, uses more tokens). You can select one or more.
  • Sibling context: Enable when related documents in the same folder share context (e.g., chapters of the same report).
  • Parent context: Enable when your folder structure carries meaning (e.g., documents inside a “Legal” folder should be recognized as legal documents).

​
Saving changes

Click Save to apply your changes, or Cancel to discard.
Updated schemas only apply to newly indexed documents. Existing documents keep their original metadata — reindex them if you want the new schema applied.