v1.13.0

🚀 Version 1.13.0 is now available!

==================================

This release is all about the Workflow Generator: a chat panel that stays out of your way, full visibility into the model's reasoning, and generated workflows that are easier to read and that actually run correctly on the first try. Plus a round of Brand Kit and canvas improvements. Here's everything that's new!


💬 A refreshed Workflow Generator panel


✨ The generator now adapts to what you're doing instead of fighting for space.

  • Status pill: when a blocking modal opens, the chat panel automatically shrinks into a compact status pill underneath it? You never lose sight of generation progress. Collapse it manually whenever you want, and re-expand it above the modal when you need the full conversation back.
  • Visual refresh: a new skin with a white header, an elevated composer surrounded by an animated ring while generating, and a clearer meta bar hierarchy (assertive model chip, ghost new-chat button).
  • Reasoning trace for everyone: the model's thinking panel, previously admin-only, is now visible to all users. Follow the reasoning behind every workflow the generator builds.

🧠 Smarter, more reliable generation


âš¡ Generated workflows are easier to read and behave correctly at run time.

  • Readable node names: LLM, Agent, code block and merge-with-template nodes are now given meaningful business names instead of raw type labels. The completion card also tells real nodes apart from annotations, sticky notes are no longer counted.
  • Brand kit already linked = question already answered: if your workflow has a brand kit attached, the clarifier stops asking and pre-fills the choice from your default brand.
  • Scalar default values: the wDSL compiler now emits scalar defaults for Text and Date inputs instead of arrays, and the run form unwraps arrays that were already persisted fixing default values that broke at execution.

🎨 Canvas & Brand Kit polish


🛠 A few adjacent improvements shipped alongside the generator work.

  • Readable edges at low zoom: connections stay legible when you zoom out, thanks to capped stroke compensation.
  • Brand Kit generation hardening: a more reliable logo extraction cascade, honest competitor lists (an "AI-suggested" flag plus an abstention notice when confidence is low), and automatic tinting for light logos.
  • Required-params validation fix: dimension params are now left empty so mode conditions correctly flag them.

📚 Plus plenty of bug fixes and performance improvements.