Release overview — October 2026

The Curiosity Team

• 2026-10-18

October has been about running Curiosity Studio yourself: a hosted model provider for self-hosted workspaces, server tuning without a restart, and a round of indexing fixes. Here's the month so far.

Curiosity AI for self-hosted workspaces

Self-hosted workspaces can now pick Curiosity AI as their model provider, during onboarding or from the AI providers list. The administrator receives an activation code by email, and the resulting account credentials stay on the server and are never sent to the browser.

Tune the server without a restart

• A new Runtime Tuning page (Manage > Configure) adjusts cache intervals, query cache size, memory trimming, simultaneous indexes and linking queue limits live. It also sets how the server responds to high memory use: pause indexing, clear caches, both, or ignore. The settings travel with the workspace configuration as config/runtime-tuning.json.
• Memory diagnostics on the Profiling page track memory-management pauses over a window from 1 minute to 24 hours, with an explanation of each long pause.
• The NLP pipeline dialog gains a live Status tab, and a new metric shows whether the index queue is growing or shrinking.

Agents and chat

• Agents can carry skills, picked in the agent editor and included in configuration export and import. An agent can also opt out of background runs, so its result is always returned directly.
• When an assistant runs a shell command, the chat draws it as a terminal window with the working directory, command, output and exit status instead of raw JSON.
• Chats compacted more than once now update the previous summary rather than re-summarising it, so file paths, IDs and error messages survive long conversations.
• JSON parsing of model output repairs common malformations, and a model that calls a missing tool is told which tools it can call.
• The admin assistant dock and its conversation stay open across navigation and refresh, and the "Sudo Provider" setting is now honoured.

Search and indexing

• EmbeddingGemma 2 (768 dimensions, multilingual) can be selected for vector indexing.
• Custom code indexes can be marked high priority, and spotters built from a node type pick up new nodes within moments instead of waiting for a full rebuild. Spotters can also ignore dates.
• Filling access-restricted copies of a text index only processes the nodes each copy should hold, which makes adding groups much cheaper.

For developers

• The Catalyst.Rampart library, which detects personal information and redacts it reversibly on the server, is available to endpoints, AI tools and other workspace code.
• The new upload-front-end-source CLI command (and Graph.UploadFrontEndSourceAsync) commits a local front-end project and queues its build. An upload to the publishing branch goes live only after the build succeeds.
• Custom front-ends can show an environment badge such as "DEV" next to the application name.

The little things

• Change History filters by user and by code text.
• Guest access and self sign-up settings apply immediately, without a restart.
• Invite, password-reset and SSO links keep the sub path behind a reverse proxy.
• Search results loaded on scroll keep their highlights and pins.
• A migration job's status can be set from its editor.

This overview covers builds v26.10.72149 (October 1–18). For the full, build-by-build detail see the changelog.