Changelog

A new opt-in Git sync mode watches your repo, classifies what changed, and acts on it in one round trip — without you driving each step by hand. It's opt-in per repo, and every run leaves a restore point, so it's never a one-way door.

Also fixed this release: phantom zero-test runs from empty report directories, the wrong test's trace/history displaying, revoked ingest keys silently marked "uploaded," a CTRF fallback upload path that never actually worked, and a stuck self-updater.

• tr_diagnose_run listed every failure with a blank test name, an error_type stuck on "Error", an empty message, and zero duration — the count was right, so it looked like a working tool returning useless data. That's fixed.
• Timed-out tests were previously dropped entirely; they're now counted and correctly inferred as TimeoutError.
• Failures now collapse per test instead of per timeline step, so one failing test no longer produces multiple duplicate "failures."
• This also fixes tr_compare_runs and both healing tools.

Released as @testrelic/mcp v3.4.1.

Studio's Tests view can now group by the same AI-classified functional Modules as the cloud platform, in both Local and Cloud mode. See Modules for how the grouping works.

Three additions round out Modules:

• Auto Group with AI — a one-click pass that places any test the free local classifier left Ungrouped, reusing your existing module names instead of minting new ones.
• Ask AI — chat directly with the assistant to merge, rename, or move modules ("merge Checkout into Products") instead of re-running the automatic grouper and hoping.
• Every module now carries a Manual or AI generated badge, and the API returns each test's module so Studio and other clients can render placement without a separate lookup.

The repo Overview tab also gains a Modules panel: grouping coverage, per-module health sorted by failure count, and suggested merges for modules that look like duplicates.

See Modules for the details.

The Test Cases tab gains a compact History strip (pass/fail across recent runs) and a Pass rate column — so a pass rate reads identically here, on Test Runs, and in Test Case detail.

• The Tests view no longer goes stale after a run finishes.
• HTML report popups that were failing to open now work.
• Upload failures no longer fail silently — you'll see them.
• The context-window report now follows whichever engine is currently selected, not just Claude.

A new Modules view in the Test Cases tab groups tests by functional area — Authentication, Dashboard, Checkout — instead of by file or suite:

• Classified automatically, for free — no LLM call, no token cost.
• Fully editable: create, rename, merge, or delete a module, and reassign tests individually or in bulk.
• A per-repo Settings toggle turns automatic classification off if you'd rather group everything by hand.

See Modules for the full picture.

• A new context report shows what's actually filling and freeing the AI's memory window, so a long session doesn't run out of room by surprise.
• Headless testrelic -p now dispatches to whichever engine you've actually selected — Codex, Cursor, or Ollama — instead of always defaulting to Claude Code. Ollama also gained cross-turn chat memory.

See Modes & engines.