v1.61

v1.61: Fabric CLI

|2 min read

What's New

  • Fabric CLI: Run Fabric from the command line — fabric -p "prompt" prints answers directly to the terminal, bare fabric opens an interactive mode, and --continue/--resume resume past sessions. Available from your terminal on launch.

  • Mid-Turn Steering: Messages sent while the assistant is working are delivered at the next natural break without interrupting the current action, with tiered Escape options to retract, reclaim, or cancel.

  • Four New Languages: The UI is now available in German, Spanish, French, and Italian, bringing the total to six languages. Switch languages live, and your OS language sets the default on first run.

  • Live Terminal Output: Terminal commands now show output in real time as it runs, with the exact command displayed above the results.

  • Word-by-Word Text: AI responses now reveal text word-by-word instead of in chunks. Conversation history renders instantly, and the animation respects reduced-motion preferences.

  • Thinking Blocks Controls: Choose how AI reasoning is displayed — expand all, collapse after responses complete, or collapse all by default with a summary. Available in Settings → General → Chat Management.

  • Inline Mention Pills: @file, @folder, and MCP mentions now appear as inline pills in the chat composer and history instead of plain text.

  • Beta Update Channel: Enterprise users can opt into nightly beta builds via Settings → General, without affecting stable users.


Bug Fixes

  • New projects now auto-select a default model instead of starting with none selected and blocking the chat input.

  • Fabric no longer scans for common local development servers at startup. Only Ollama and LM Studio are detected automatically; other local AI engines can be connected via Quick Connect in Settings.