Skip to content

v0.29.0

Choose a tag to compare

@shiroyasha shiroyasha released this 16 Jul 14:23
1b8c377

What's new since v0.28.0

Console dashboard widgets

  • Scorecard widget joins the panel lineup — a single-KPI view with a status-colored trend, an optional target with a progress bar, and a colored sparkline.
  • Table widget gains avatar, progress bar, trend, and value + trend columns.
  • Runs data sources (table/chart/number/scorecard panels, plus markdown and HTML run variables) can now be filtered by status and trigger, reusing the same vocabulary as the canvas runs sidebar.
  • The timestamps tooltip was consolidated across dashboard widgets, and the console dashboard UI got a dark-mode and Storybook polish pass.

Run inspector

  • Queued steps now render in the run inspector with their input and a cancel action, and run detail loading is batched and parallelized to cut added latency.
  • Runner logs are now available directly in the run inspector timeline.
  • Continued polish: faster live-canvas run inspection, clearer runtime actions and navigation controls, fit-to-view for run inspection nodes, focus-on-participants, and the ability to open the run inspector for older node runs.

1 new capability: broadcast messaging

  • Apps can now broadcast messages to each other one-way within the same organization, via a new broadcastMessage action and onBroadcast trigger backed by a new subscriptions table.

App experience

  • Full-page editor: every multi-line configuration field on a component or trigger now has an expand button that opens a large modal editor, so long prompts, messages, and scripts are easier to write and review.
  • Apps gain a configurable default tab.
  • Markdown picks up GitHub-style alerts (NOTE/TIP/IMPORTANT/WARNING/CAUTION) and [!SECTION] accordions, and thematic breaks now render as dividers, shared across Files and Console.

Improvements to existing integrations

  • GitLab adds createDeployment and createDeploymentStatus actions plus acceptMergeRequest and approveMergeRequest, and the project field now accepts expression values.
  • Claude gains a createBatchMessage component for running prompts over an array of inputs via the Message Batches API, a toggle to persist sessions after a run, and file/artifact support for its text-prompt component.
  • OpenAI's text-prompt component also gains file/artifact support.
  • Cursor's launchAgent now returns artifacts, and Discord's sendTextMessage and GitHub's comment action can send a file by URL.
  • GitHub adds a markPullRequestReadyForReview action.

Fixes and polish

Continued stabilization of the staging and edit-mode work from v0.28.0: repository file staging is preserved when re-entering edit mode, staging now uses the canvas update permission, and two HTTP 500s were fixed. Elsewhere, expression autocomplete stays stable on enter and now supports GitLab project fields and Claude Run Code Agent PR URLs, SSH inline scripts stream correctly and normalize line endings, stale execution requests are ignored, runs waiting on deleted approval nodes are cancelled, agent layout controls are respected, survey questions render separately, and agent/account limit errors are handled gracefully.

Bug Fixes

Improvements

New features


Full Changelog: v0.28.0...v0.29.0