v0.6.0
v0.6.0
✨ New Features
- Massively expanded MCP server — from 14 to 38 tools: global search, run insights, spec health, slow tests, performance & stability trends, network requests, failure groups, repo commits & diffs, test catalog, traces, tags, links, recent activity and instance stats
- AI failure-cluster workflow via MCP — list open clusters, run cluster diagnosis, explain failures, set cluster status & base commit, submit diagnosis feedback
- Locator healing exposed through MCP with enriched healing context
- Timeline span-type filter — granular, data-driven toggles for Tests, Setup, Hooks, Fixtures & Wasted waits, showing only the span kinds present in the run
- OpenAPI spec in production — generated at build time and properly served at
/_openapi.json
🔧 Improvements
- Full reporter refactor with a modular internal architecture (capture, collect, config, files, support) and new architecture docs
- Improved AI context building: better data aggregation, error fingerprinting, source snippets and trace parsing
- MCP server hardening from external API audit: fixed pagination cursors, enforced project scope, richer flaky-test stats and real diagnosis evidence
- Insights page content improvements, with zero-data and demo-mode error fixes
- Timeline fixes: wasted-time bars no longer vanish after reload
- More data surfaced on test case pages (cluster summary, execution details)
- More realistic demo seed and simulator fixes (accurate test counts)
- Catchier README and docs landing page; docs aligned with the current product
🏗️ Internal
- WorkersTimeline split into composables and presentational sub-components
- Demo seed now generated in CI instead of being committed
- CI: cancel in-progress runs when a new build starts on the same ref
- Code cleanups and dependency optimizations
Full Changelog: v0.5.0...v0.6.0