Rehearse.io

Enterprise async interview platform with AI evaluation for behavioral (voice) and DSA coding rounds.

Monorepo with three services:

• backend: Bun + Express 5 + TypeScript + Drizzle/PostgreSQL, JWT auth, port 9000
• frontend: React 19 + TypeScript + Vite 7 + TailwindCSS 4 (static SPA, Cloudflare Pages)
• ai-service: FastAPI (Python) — scenario generation, STT, TTS, DSA problem generation and code evaluation, port 8000

Honest status lives in STATE.md. Production is live: the API answers GET /health (with checks.database) at api.rehearseio.triptribe.info, and the frontend is served by Cloudflare Pages at https://rehearse.io.shrijit.tech/. Read STATE.md before trusting any other doc. (Also: rehearse.io itself is a parked GoDaddy domain for sale, not this project's.)

Features

• Candidate practice: voice rehearsal + DSA coding practice with AI scoring
• Recruiter interviews: behavioral, DSA-only, or mixed (voice then coding)
• Invite links: shareable candidate links with multi-round progress
• AI evaluation: Whisper transcription + LLM feedback for voice; correctness / quality / complexity for code (static LLM review, no sandbox)
• Results dashboards: per-candidate behavioral and DSA breakdowns
• Organizations: multi-recruiter orgs with member invites
• Consent + GDPR: consent versioning, data export, soft-delete account
• Audit log: security events recorded (write-only, no viewer yet)

All AI features need GROQ_API_KEY (STT + TTS + LLM via LiteLLM). Without it the AI service returns honest 503s and false health flags.

Getting started (local dev)

Prerequisites: Bun, Node 22+, Python 3.11+, and a reachable PostgreSQL (the shared laptop instance is the convention: createdb rehearse_io_dev).

make setup                       # install all three services + create .env files
export DATABASE_URL=postgresql://postgres@localhost:5432/rehearse_io_dev
export JWT_SECRET=$(openssl rand -hex 32)
make migrate                     # create/update tables
make dev                         # backend :9000, frontend :5173, ai :8000

Service URLs: backend http://localhost:9000, frontend http://localhost:5173, AI http://localhost:8000.

Health: backend GET /health (the fleet contract; Postgres is the critical check, 503 when down — there is no /ready route); AI GET /health (reports STT/TTS availability).

Make targets

make setup | dev | test | lint | build | migrate | health — each maps 1:1 to a script in scripts/. CI calls these too, never framework commands directly.

Tests

make migrate
(cd backend && bun src/server.ts) &
SMOKE_BASE_URL=http://127.0.0.1:9000 make test

backend/tests/smoke.ts covers auth, orgs, interviews, invite-token hashing, placeholder claiming, GDPR export, and consent against a real Postgres.

Deployment

Deploy procedure (branches, workflows, secrets, gate failures) is in DEPLOY.md. Live status is in STATE.md.

• backend → GHCR (:<sha>) → Dokploy compose app (rehearse-api), port 9000, health /health
• frontend → static dist/ → Cloudflare Pages
• ai-service → Dokploy, port 8000
• secrets → Doppler platform / prd (one flat namespace, REHEARSE_ prefix), never in git
• database → shared Postgres, rehearse_io_prod / rehearse_io_dev

See platform.yaml, frontend/platform.yaml, and STATE.md for current status.

Docs

• STATE.md — honest current state (read first)
• DEPLOY.md — how this repo deploys (read before deploying)
• AGENTS.md — agent/dev instructions per service
• PRODUCT.md — product purpose and design principles
• backend/README.md — API reference
• frontend/README.md — frontend guide
• ai-service/README.md — AI service reference