Personal AI-powered technical interview practice platform — question generation, interview simulation, live coding, flashcards, and personalized study roadmaps.
Live: devinterviewlab.vercel.app
CI: 
DevInterviewLab is a full-stack study platform for technical interviews that uses AI to create a complete practice cycle: generate questions from a personalized roadmap, simulate interviews with detailed feedback, practice live coding with a Socratic pair programmer, review concepts with spaced-repetition flashcards, and track progress with visual metrics.
| Layer | Technology |
|---|---|
| Framework | Next.js 15 (App Router, Server + Client Components) |
| Language | TypeScript — zero errors, no ignoreBuildErrors |
| Auth + DB | Supabase (PostgreSQL + RLS + Auth) |
| AI | Groq llama-3.3-70b-versatile via OpenAI-compatible API |
| Server state | React Query (@tanstack/react-query) |
| Client state | Zustand |
| UI | Tailwind CSS + Radix UI + Lucide React |
| Theming | next-themes (dark / light mode) |
| Editor | Monaco Editor (desktop) / textarea fallback (mobile) |
| Charts | Recharts |
| html2canvas + jsPDF (score-card export) | |
| Document parsing | pdf-parse (CV upload → text extraction) |
| Monitoring | Sentry (tunnel via /monitoring) |
| Testing | Vitest — 331 unit tests |
| Git hooks | Husky — pre-commit (tsc), pre-push (test suite), commit-msg (commitlint) |
| CI/CD | GitHub Actions + Vercel (auto-deploy on merge to main) |
- Personalized roadmap — CV upload + job description generates a gap analysis and 30/60/90-day roadmap
- Topic-based question generation — AI-generated theoretical Q&A and live coding questions per roadmap topic, in EN and PT, without repetition on regeneration
- Direct links — "Practice" button on each topic opens the Questions tab pre-filtered by topic
- Roadmap questions — default tab with topic filter, concept generation from questions, bulk delete
- Flash Topics (Concepts) — quick technical references with summary, when to use, pros & cons (accordion), integrated Q&A, and persistent EN↔PT translation
- AI Interview Coach — dimension-based evaluation (correctness, completeness, clarity, depth) with follow-up and rebuttal
- Live Coding Simulator — 7 languages, configurable timer, Socratic pair programmer with on-demand hints, idle detection
- Question cards by type (Total / Theoretical / Live Coding) per roadmap
- Bar chart of questions by topic — responsive, no horizontal scroll
- Demo Mode —
/demopublic route with mocked data and 4 interactive tabs, accessible from the login screen - Diverse questions — 5-angle taxonomy (Conceptual, Practical, Trade-off, Debug, Evolution) + banned phrases prevent repetition across generations
- PWA — installable, mobile bottom tab bar, iOS safe-area support
- i18n — full EN and PT-BR, synced with database
- Theming — light / dark mode
app/
(app)/ # protected routes (auth layout)
plano/ # roadmap + question generation
revisar/ # roadmap questions + Flash Topics (concepts)
demo/ # public route with mocked data (no auth)
simular/ # interview coach + live coding
stats/ # metrics and charts
api/ # route handlers (Next.js App Router)
features/ # domain logic (React Query hooks + components)
lib/
ai/ # ai.service.ts + prompts/
i18n/ # translations.ts + useT()
supabase/ # client server/browser + types
api/ # rate-limit, brute-force, logger
store/ # zustand stores
__tests__/unit/ # 331 Vitest unit tests
Every API route follows this checklist:
| Item | Implementation |
|---|---|
| Authentication | supabase.auth.getUser() → 401 if unauthenticated |
| Rate limiting | checkRateLimit('endpoint') before any AI call |
| Sanitized errors | sanitizeError() — no stack traces sent to client |
| RLS | All 19 tables with USING + WITH CHECK policies |
| Uploads | validateFileBuffer() — magic bytes + MIME + size |
| Brute force | 10 attempts / 15 min per IP — persisted in Supabase (brute_force_log via SECURITY DEFINER RPCs), in-memory fallback |
| CSRF | Origin validation on all POSTs |
| Security headers | CSP (route-scoped via middleware — unsafe-eval only on /live-coding), HSTS, X-Frame-Options, X-Content-Type-Options |
npm test # 331 unit tests (Vitest)
npm run test:coverage # with coverage report| Test file | Tests | Coverage |
|---|---|---|
brute-force.test.ts |
14 | IP blocking, sliding window, reset |
brute-force-persistent.test.ts |
8 | Supabase RPC path, fallback, retryAfterSec |
rate-limit.test.ts |
13 | checkRateLimit per endpoint |
file-validation.test.ts |
10 | Magic bytes, MIME, size |
spaced-repetition.test.ts |
25 | Full SM-2: EF, intervals, reset |
prompts.test.ts |
63 | Evaluate, behavioral, followup — EN/PT, schema |
generate-prompts.test.ts |
32 | Generate, coding-hint, cn() |
stream.test.ts |
9 | ndjsonStream + readNdjsonStream |
interview-payload.test.ts |
5 | Serialization with transcript |
score-card-utils.test.ts |
6 | aggregateRadar, averageScore |
roadmap-prompt.test.ts |
6 | Language, truncation, JSON schema |
topic-prompt.test.ts |
29 | getTopicSystemPrompt, topicAnalysisPrompt, topicTranslatePrompt — incl. output-language-lock rules |
topic-pairs.test.ts |
10 | groupIntoPairs — pairs, fallback, ordering |
score-card-prompt.test.ts |
11 | EN/PT, JSON-only, strengths/gaps |
coding-generate-prompt.test.ts |
13 | Difficulty, topic, language, fallbacks |
roadmap-questions-prompt.test.ts |
28 | fixJsonNewlines, safeParseJSON, theoretical and live coding prompts |
api-evaluate.test.ts |
4 | 401/429/400 guard layers on /api/ai/evaluate |
api-roadmaps.test.ts |
3 | 401 unauth, 200 success, progress field on /api/roadmaps |
api-topics.test.ts |
11 | 401/400 guards, translation-gap backfill on POST, self-healing + language-mismatch repair + fair cap on GET /api/topics |
topic-language.test.ts |
10 | detectContentLanguage, hasLanguageMismatch — EN/PT detection, loanwords, null-safe |
Coverage: Statements 97% · Branches 90% · Functions 100% · Lines 98%
npm install
cp .env.example .env.local
# fill in the variables below
npm run devNEXT_PUBLIC_SUPABASE_URL=https://idgpscsnbgszhwvhtedy.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=<anon key from Supabase dashboard>
# Groq (free tier) — https://console.groq.com
OPENAI_API_KEY=gsk_...
OPENAI_BASE_URL=https://api.groq.com/openai/v1
OPENAI_MODEL=llama-3.3-70b-versatile
SENTRY_DSN=<optional>
NEXT_PUBLIC_SENTRY_DSN=<optional># Apply in ascending order
supabase db push # or apply manually via the Supabase dashboardnpm run dev # development server (localhost:3000)
npm run build # production build (must pass cleanly)
npm run lint # ESLint
npx tsc --noEmit # TypeScript — zero errors required
npm test # 331 unit testsEvery push and PR automatically runs:
npx tsc --noEmit— zero TypeScript errorsnpm run lint— zero ESLint warningsnpm test— all tests must passnpm run build— clean production build
Merge is blocked if any check fails.
Merge to main → automatic deploy on Vercel.
git checkout -b feat/feature-name
# develop...
npx tsc --noEmit && npm test # or /check in Claude Code
git push origin feat/feature-name
# open Pull Request → CI runs automaticallyFernando Ghiberti — github.com/ghiberti85