Back to the changelog

• Feature

  Package the way you work

  Solutions gather an entire workspace's process maps, flows, connectors and — if you choose — AI assistant rules into a single versioned package. From Settings → Solutions you compose them by choosing what to include, export them with their own version (every export stays downloadable from its history, unchanged even after later edits), and import them into another workspace by choosing whether to link, create or skip each component, with a report of what was saved. Imported connectors stay switched off and free of credentials until you reconfigure them. Assistants connected through MCP can already list available Solutions and preview their package. This is a premium feature: your administrator turns it on when ready.

• Integration

  Toward GitHub

  We're preparing automatic saving of Solutions, maps and flows to a GitHub repository for your organization, on every save and publish. The feature is announced in Settings but not yet available.

• Design

  A clearer AI task panel

  "Expected output" now has its own (?) help like the other top fields of a task. Under the "System 2 · LLM" and "System 1 · Jev" buttons a line says when to use each: reasoning and content generation with the LLM, fast low-cost decisions with Jev. In the Jev branch the provider sits at the bottom, as in the LLM branch: a detail to touch only when needed.

• Feature

  System One presets

  In Settings you set up the context, questions and provider of a System One evaluation once and save them as a preset; in the task you pick it from a list, always seeing where each answer goes, without rewriting the questions. Changing the preset updates every task that uses it; "Unlink and edit" customises a single task. A catalog of four ready-made models (CV screening, lead scoring, triage, documents) gets you started right away.