Sep 22, 2026
2 min
Changes released on Sep 22, 2026.
New
- Re-run only the failed scenarios. After a run finishes with failures, you can now re-run just the failed scenarios from a dedicated dialog instead of repeating the whole run.
- Run preconditions. Runs can now carry preconditions that scenarios are checked against before they start, and you can add a precondition to several scenarios at once with the new bulk dialog.
- Richer dashboard. The dashboard gains a custom date range picker, user-based KPI widgets and a scheduled runs widget, with schedule filters and run summaries.
- Step wording suggestions in failed reports. When a failed report points to a step whose wording is the likely problem, RafiRun can suggest a rewording (including splitting a step or replacing it) and stage it in the scenario editor. Nothing is saved or re-run until you do it yourself.
- Nested scenario folders for Playwright. Playwright scenarios can now be organised in nested folders.
- PDF and shared reports for Playwright. Playwright runs now have a PDF report and a shareable report view, and report sharing is recorded in the audit log.
Improvements
- Live step progress. Scenario runs now report step-by-step progress as it happens, so you can follow which step is executing.
- Usage and cache details in reports. Reports show AI usage and action cache statistics along with a usage history, and scenario cache details include new stats. Run durations are also calculated more consistently.
- Smoother authoring wizard. The wizard offers the name and summary proposed by the authoring run so you can take them with one click without overwriting what you typed. The run's startup is shown as milestones, and the executed steps stay visible when an authoring run fails.
- Cleaner scenario and run lists. The scenario list now shows how many runs use each scenario and drops the unused Status and Last Run columns and the URL. The runs list shows who created each run and can be filtered by multiple creators.
- Clearer error messages. Appium session failures now explain timeouts and version mismatches, and condition wait timeouts get their own message in reports.
- More control over AI models for admins. Admins can pick from models discovered on the provider, set a semantic reader on AI models, configure Agent Jev in system settings, and check the assigned authoring model for the Playwright module. Testers also get a clear message when a job is refused because no authoring model is set.
Fixes
- Tests lost when a node disconnects are recovered. When a test node disconnects mid-run, the affected tests are now re-queued without using up the scenario's retries.
- Report status filters apply before paging. Status filters on grouped reports are now applied before pagination, so pages no longer show gaps or miss matching results.
- Report and playback fixes. Reports now tell automatic and manual recovery apart with separate badges, step screenshots stay visible after a failed or cancelled authoring run, and step videos in the lightbox no longer stall on non-keyframes.