Augur
Studio's Augur surface embeds each run's evidence report verbatim, runs a zero-setup demo, and lets you approve or reject references without leaving the pane.
Studio's workbench rail adds a live Augur surface, next to
Live Browser, Live Device, and
Evals — replacing what used to be a disabled, roadmap-only entry. It's
built for testrelic augur, the CLI's text-to-image eval pipeline.
Evidence, embedded verbatim
Every Augur run already ships a self-contained report.html — verdict banner, banded
pHash/SSIM/pixel checks, five evidence viewers (side-by-side, slider, flicker, overlay, diff
mask), region annotations, and plain-language root causes. Rather than re-implement that UI,
the pane embeds the real report through a loopback static server, so what you see in Studio is
exactly what augur open renders on its own.
Close to the limit near the lower-right — worth a look.
- ⚠ 92% similar — close to the limit
- ✓ pixel difference 1.02% — within limits
Run history and verdicts
The pane discovers runs from both .testrelic-augur/runs/ and demo/runs/. Demo runs are
labeled, so a augur demo NO-SHIP is never mistaken for your suite's own result. Each run gets
a verdict chip — SHIP, NEEDS REVIEW, or NO-SHIP — derived the same way the report's
own banner is: any failing case means NO-SHIP, any case needing review (with no failures) means
NEEDS REVIEW, otherwise SHIP.
Zero-setup demo
A one-click Run demo streams augur demo live — 5 scenes, 9 real GenAI failure patterns,
no API keys required — into an activity feed, then opens the finished run's report. It's the
fastest way to see what Augur catches before pointing it at your own generator.
Approve or reject without leaving the pane
The pane prefers the CLI's own review server (augur open <run> with TRC_AUGUR_SERVE=1),
which serves the same report with a per-session token injected into the page — never in the
URL — plus the promote/reject API the report's Apply buttons call. That means references
can be approved or rejected directly from Studio, the same as running augur open from a
terminal. If the CLI can't serve the run, the pane falls back to its read-only static server —
evidence still renders, and the report's copy-commands tray covers approval instead.
The image changed beyond the allowed limits.
per-session token injected into the page by `augur open --serve`, never the URL
Next steps
→ testrelic augur (CLI) · Augur multi-model comparison (cloud
platform) · Workspace
Evals
Studio's Evals pane reads DeepEval test-run JSON straight out of your repo and renders metric scores, agent conversations, and tool calls — no upload required.
Workspace
Explorer, Command Center, a multi-session workbench, Test Paths, run visualization, local issue detection, fix & re-run, an embedded terminal, and Cloud Mode inside TestRelic Studio.