Release overview — September 2025

The Curiosity Team
•

September was about moving and explaining a workspace: take its whole setup from one environment to another, and see why a search returned what it did. A busy month with five releases.

Back up and move a workspace's setup

You can now export a complete snapshot of a Curiosity Workspace's configuration (indexes, endpoints, code tasks, NLP models and patterns, AI providers and schema) to a single zip file, and import it into the same or another workspace.

  • Available to administrators from the workspace settings and from the CLI, where it replaces the previous custom-code-only export.
  • Imports report itemized errors and warnings for anything that can't be applied.
  • Custom API endpoints can also be imported and exported on their own, including their access group assignments.
"Definitions only"

The export deliberately contains definitions and no document data, which makes it suitable for backing up a workspace's setup and moving it between environments.

See how a search worked

  • Query explanation: show how a query was interpreted in a compact or detailed form, or have it always displayed rather than only when something was rewritten.
  • Search statistics for admins: a step-by-step breakdown of how a search was executed and how long each stage took.
  • Exclude filters: saved searches and facet queries can define an exclude clause, removing results that match it.
  • Pinned and reordered filters keep their order per search context across reloads, and a "Show all filters" button clears a filter search.

Connectors

  • OneNote on demand: OneNote content from OneDrive, SharePoint and Teams is extracted and rendered when you open it, then ingested in the background.
  • Confluence as a shared source: connect it as a multi-user data source so its pages and files are shared across the workspace.
  • Single-user syncs (Box, Confluence, Jira, Notion, Salesforce, HubSpot, Evernote, pCloud, Outlook Calendar) pause for an owner whose sign-in expired and prompt to re-authenticate, rather than retrying repeatedly.

Tighter access

  • Deleting a user, or a schema and its data, requires typing "DELETE" first.
  • Admin-only routes are blocked for non-admins in the browser, and metrics endpoints require system-admin access.
  • When uploads or account-preference editing are turned off, the server rejects them for regular users.
  • Related-node facets respect each user's access rights, and viewers can no longer edit saved searches.

For administrators

  • AI provider settings have their own page, connector run logs load in pages with timestamps, and usage analytics break down searches per user over time.
  • Rename a schema's key field, with validation against conflicts.
  • Custom sign-out URLs, names kept up to date from SAML, and login on read-only replicas through a custom authentication flow.
  • The CLI's git sync supports mutual TLS and a verbose mode.

The little things

  • A "More actions" button on result cards, and action menus that show only commands that apply.
  • Lists of related links can expand inline.
  • Summaries can be asked for in a specific language.
  • Pages opened with only filters, no typed query, now show results.
  • The model's reasoning steps show as a brief "Reasoning..." status.

This overview covers builds v25.9.60778 → v25.10.61879. For the full, build-by-build detail see the changelog.

Read next

Articles on context graphs, enterprise search and industrial AI

Connected knowledge for AI systems