IA Gen Maturity Evaluator

Production-ready web app for evaluating IA/GenAI maturity of candidates. Built by Peaksys for Peaksys.

Tech Stack

• Next.js 16 (App Router), React 19, TypeScript, TailwindCSS
• Prisma + SQLite (local) — switch to Postgres/Supabase for production
• Recharts (radar chart), jsPDF (PDF export)

Setup

pnpm install
cp .env.example .env
pnpm db:generate
pnpm db:push   # ou pnpm db:migrate pour une DB vide
pnpm db:seed

Note : Si la DB existe déjà (créée avec db push), pour basculer sur les migrations : pnpm prisma migrate resolve --applied 20250220000000_init

Run

pnpm dev

Open http://localhost:3000.

Seed Data

• 3 candidates with sample evaluations (Alice Chen, Bob Martin, Carol White)
• 2 templates: Full 15-dimensions, Short 8-dimensions
• Admin user: admin@peaksys.local (mock auth)

API Routes

Route	Method	Description
/api/evaluations	GET, POST	List / create evaluations
/api/evaluations/[id]	GET, PUT	Get / update evaluation
/api/templates	GET	List templates
/api/export/csv?id=	GET	Export evaluation as CSV
/api/export/pdf?id=	GET	Export evaluation as PDF
/api/auth	GET, POST	Mock auth
/api/ai/suggest-followups	POST	AI follow-up suggestions (stub)

Export cURL Examples

# CSV export (replace EVAL_ID with actual evaluation id)
curl -o evaluation.csv "http://localhost:3000/api/export/csv?id=EVAL_ID"

# PDF export
curl -o evaluation.pdf "http://localhost:3000/api/export/pdf?id=EVAL_ID"

With auth header (when real auth is added):

curl -H "Authorization: Bearer YOUR_TOKEN" -o evaluation.csv "http://localhost:3000/api/export/csv?id=EVAL_ID"

AI Assistant Stub

The AI assistant is a client-side stub that returns deterministic follow-up suggestions based on:

• Dimension name
• Candidate answer length
• Current score (low scores trigger probing questions)

To plug a real LLM:

1. Create or update /api/ai/suggest-followups to call OpenAI/Anthropic/etc.
2. Pass { dimensionName, candidateAnswer, currentScore } in the request body.
3. Use a prompt like: "Given this dimension and candidate answer, suggest 2–3 probing interview questions."
4. Return { suggestions: string[] }.

The client already calls this API when the user clicks "Get AI follow-up suggestions" in the dimension card.

Tests

# Unit tests (Vitest)
pnpm test

# E2E tests (Playwright) — requires dev server
pnpm test:e2e

Run pnpm exec playwright install once to install browsers for E2E.

Deploy

1. Set DATABASE_URL to Postgres (e.g. Supabase, Neon).
2. Run migrations: pnpm db:migrate (ou pnpm db:push en dev)
3. Seed if needed: pnpm db:seed
4. Build: pnpm build && pnpm start
5. Or deploy to Vercel (set env, use Vercel Postgres or external DB).

Acceptance Criteria

• Auth: mock single-admin login
• Dashboard: list evaluations and candidates
• Create/Edit evaluation: candidate, role, date, evaluator, template
• Templates: Full 15-dim, Short 8-dim
• Interview guide: definition, rubric 1→5, signals, questions per dimension
• AI assistant: stub suggests follow-ups
• Scoring: 1–5, justification, examples, confidence
• Probing questions when score ≤ 2
• Radar chart + findings/recommendations
• Export PDF and CSV
• Admin: view templates
• Warning when all scores = 5 without comments
• Edit after submission (audit log)
• Mobile responsive (Tailwind)

Manual Test Plan

1. Dashboard: Open /, verify evaluations table or empty state.
2. New evaluation: Click "New Evaluation", fill form, select template, submit.
3. Interview guide: On evaluation page, score dimensions, add notes, click "Get AI follow-up suggestions".
4. Low score: Set a dimension to 1 or 2, verify probing questions appear.
5. All 5s: Set all scores to 5 with no justification, submit — verify warning.
6. Aggregate: Click "Auto-generate findings", verify radar chart and text.
7. Export: Click Export, download CSV and PDF.
8. Admin: Open /admin, verify templates listed.

File Structure

src/
├── app/
│   ├── api/           # API routes
│   ├── evaluations/   # Evaluation pages
│   ├── admin/         # Admin page
│   └── page.tsx       # Dashboard
├── components/        # UI components
└── lib/               # Utils, db, ai-stub, export-utils
prisma/
├── schema.prisma
└── seed.ts
tests/e2e/             # Playwright E2E