Knowledge base API

Endpoint: /api/v1/workspaces/WORKSPACE_ID/knowledge-bases. Use knowledge:read for GET and knowledge:write for mutations.

Create a knowledge base

curl -X POST 'https://dashboard.verbatik.com/api/v1/workspaces/WORKSPACE_ID/knowledge-bases' \
  -H "Authorization: Bearer $VERBATIK_API_KEY" \
  -H 'Content-Type: application/json' \
  --data '{"name":"Clinic information","description":"Reviewed public clinic information","sources":[{"id":"clinic-hours","type":"text","name":"Opening hours","content":"Monday to Friday, 09:00–17:00. Weekend visits are by prior arrangement.","status":"saved"}]}'

This is illustrative content. Replace it with information reviewed by the clinic. Success returns 201 and { "data": ... }.

Source fields

A source has id, type (text, url or file), name, optional content, optional url and a saved status. A knowledge base has up to 100 source definitions; text content is bounded.

The resource route manages records and source definitions; it is not an unrestricted file-upload or storage-connector API. Use the dashboard’s supported upload, website import and cloud document selection flows for those sources. Do not put access tokens or private document URLs into source content.

Read, update and archive

GET returns the collection in data. PATCH uses an id plus changed fields. DELETE uses ?id=KNOWLEDGE_BASE_ID and returns its archive result.

After changing sources, review indexing/sync status in the dashboard, attach the intended base to the agent and test a known question. A saved record alone does not establish that a specific answer has been retrieved correctly.