Recording of the visual regression run on eval-ai-interviewer.netlify.app · EVAL — app walkthrough.

In this video· click a step to jump to it

E
Passed ✓

Test recording

EVAL — app walkthrough

56,918 ms · 0 pixels changed · matches baseline

We built this regression test for eval-ai-interviewer.netlify.app — it's yours to keep, free.

Take this test with you

This is a real Playwright test, recorded against eval-ai-interviewer.netlify.app. Claim it into your own workspace and re-run it on every deploy — free.

export async function test(page, baseUrl, screenshotPath, stepLogger) {
  // NOTE: EVAL. is a client-side-only SPA with no persisted session (no auth
  // cookies, no localStorage token) — the "interview session" lives entirely
  // in React component memory. Lastest's setupTestId chaining only replays
  // browser storage (cookies/localStorage/sessionStorage) between tests, so a
  // live in-memory session from Test 1 would NOT survive into a chained Test 2
  // context. This test is therefore SELF-CONTAINED: it independently drives
  // the whole flow (setup form -> interview turns -> report -> coach -> admin)
  // rather than assuming continuation from Test 1's setup test.

  const shot = (n, slug) => screenshotPath.replace('.png', `-${n}-${slug}.png`);
  const settle = function () { return page.waitForLoadState('networkidle').catch(function () {}); };
+204 more lines — claim the test to get the full code.

Checks run

Run
✓
Visual
✓
Text
—
DOM
✓
Network
—
Console
✓
A11y
Review
Perf
—
URL
✓
Variables
—
Full report
Log in →
0
Diff px
57s
Duration
Review
Accessible
WCAG 2.2
A
Fast
Web Vitals · 100

Notes from the demo run

AI-generated

EVAL. skips signup entirely: the homepage IS the app. A single 'Setup Interview.' form (candidate name, email, role track, target company, seniority, JD paste, resume upload) gates straight into a live AI-driven interview session, no account creation required. The interview UI is clean and dark-themed, with adaptive difficulty (visibly escalated from 'easy' to 'medium' after one strong answer) and a live conversational-latency graph. Ending the session produces a genuinely polished 'EXECUTIVE.REPORT' with a synthesized Readiness Index (82% for our test run), a hire/no-hire recommendation, per-dimension rubric scores, strengths/gaps, and a recommended study plan with book references.

Highlights
  • No signup friction
    Zero-auth entry into the core product is a strong choice for an evaluator/demo audience — we were interviewing within seconds of landing, no email verification or OAuth detour.
  • Adaptive difficulty in real time
    The 'Difficulty Level' field visibly moved from easy to medium after the first substantive answer, and the AI's follow-up questions built on details from the candidate's previous answer rather than reading from a fixed script.
  • Executive Report quality
    The post-interview report (readiness index, rubric dimension ratings, key strengths, focus gaps, and a recommended study plan with a real book reference) is genuinely more polished than most first-launch AI recruiter tools we've profiled.
  • Live Admin analytics dashboard
    The Admin tab exposes a System.Analytics view (registered users, interviews initiated, completed sessions, average readiness index, active interview queue) that updated live across our test run — a nice transparency touch for a demo/admin audience.
Friction points
  • AI Coach Room analysis endpoint 500s
    Selecting a candidate response inside the 'Coach.' tab to view its STAR-method coaching review calls priyanshu-kalondia-eval-api.hf.space/api/reports/turn/:id/coach, which returned HTTP 500 on every attempt in our run. The panel silently stays on the empty 'Select a candidate answer...' placeholder instead of surfacing an error, so a visitor may not realize the feature is broken rather than just unclicked.
  • Backend on a cold Hugging Face Space
    The API origin (priyanshu-kalondia-eval-api.hf.space) suggests the backend runs on a free/cold-startable HF Space — worth checking if the coach-endpoint 500s are a cold-start timeout rather than a code bug, since the rest of the interview pipeline (turn scoring, report generation) worked reliably in the same run.

Share this run

Earned · Lastest awards

Proof your app is not AI slop

LASTEST
starter
regressions
0
Badges stay live, ratchet upward, only downgrade on confirmed regression.
57,344 test runs recorded · 1,174 products tested with Lastest

This test was built for eval-ai-interviewer.netlify.app — claim it free

One click copies the test and baseline screenshots into your own Lastest workspace. Re-run on every deploy and catch regressions before your users do — free, no card required.