Wansom AI — Legal Workspace Platform

AI-powered legal workspace for law firms and solo practitioners. Built with Next.js 16 App Router.

What it does

• AI chat workspace — Gemini-powered conversations scoped to client matters, with document context and tool calling
• AI Associates — Custom AI personas with practice areas, knowledge bases, and structured workflows
• Document Vault — Upload, extract, and semantically search legal documents (PDF, DOCX, Excel, CSV)
• Canvas editor — Rich-text collaborative document drafting with AI inline generation
• Briefly — Periodic AI-curated legal digest emails, jurisdiction-aware across Africa
• RAG — Admin-managed legal knowledge base with pgvector similarity search

Tech stack

Layer	Choice
Framework	Next.js 16 (App Router, Turbopack)
Database	PostgreSQL + Prisma ORM v7 + pgvector
Auth	NextAuth.js v4 (credentials + Google OAuth)
AI	Google Gemini via @google/genai
Storage	Vercel Blob
Payments	Paystack (localized pricing per country)
State	Zustand
UI	Radix UI + shadcn/ui + Tailwind CSS
CMS	Sanity (blogs, legal documents)
Email	Nodemailer (SMTP)

Development

npm install
cp .env.example .env.local   # fill in required vars (see below)
npx prisma generate          # generate Prisma client to src/prisma/client
npx prisma migrate dev --name init
npm run dev                  # http://localhost:3000

Other commands:

npm run build          # prisma generate + next build (sets --max-old-space-size=4096)
npm run lint           # ESLint
npx prisma studio      # Database GUI
npx prisma migrate dev --name <name>   # create + apply a migration

One-off admin scripts in scripts/ — run with npx tsx scripts/<file>.ts or node scripts/<file>.mjs.

Note: No test suite is configured.

Required environment variables

Copy .env.example to .env.local. The minimum set to run locally:

# Database
DATABASE_URL=postgresql://...

# Auth
NEXTAUTH_SECRET=
NEXTAUTH_URL=http://localhost:3000
JWT_SECRET=
JWT_REFRESH_SECRET=

# AI (Gemini)
WANSOM_API_KEY=        # main platform
BRIEFLY_API_KEY=       # Briefly digest pipeline (can be same key)

# File storage
BLOB_READ_WRITE_TOKEN= # Vercel Blob

# Email (SMTP)
EMAIL_USER=
EMAIL_PASSWORD=
EMAIL_HOST=smtp.gmail.com
EMAIL_PORT=587

# Payments
PAYSTACK_SECRET_KEY=
PAYSTACK_PUBLIC_KEY=
PAYSTACK_PLAN_CODE=
PAYSTACK_TEAMS_PLAN_CODE=

# Google OAuth (for calendar/gmail integrations)
GOOGLE_AUTH_CLIENT_ID=
GOOGLE_AUTH_CLIENT_SECRET=

# Sanity CMS
NEXT_PUBLIC_SANITY_PROJECT_ID=
NEXT_PUBLIC_SANITY_DATASET=production

# Admin access
ADMIN_EMAILS=you@example.com
CRON_SECRET=

Optional: GOOGLE_APPLICATION_CREDENTIALS, GOOGLE_CLOUD_PROJECT_ID, GOOGLE_CLOUD_STORAGE_BUCKET for Cloud Vision OCR on scanned PDFs. LII_SCRAPER_URL + LII_SCRAPER_API_KEY for the Pan-African legal scraper on Digital Ocean.

Project structure

src/
  app/
    (account)/        # authenticated app — dashboard, projects, vault, /workflows (AI Associates)
    (auth)/           # login, register, magic-link, password reset
    (landingpages)/   # public marketing pages + /pricing
    (admin)/          # admin dashboard, legal knowledge, email broadcast
    api/              # all API routes
  components/         # shared React components
  store/              # Zustand stores (*.store.ts)
  lib/                # utilities, auth helpers, API middleware, Gemini tools
  services/           # server-side services (RAG, digest, AI document ops)
  hooks/              # custom React hooks
  types/              # TypeScript types
  prisma/client/      # generated Prisma client (do not edit)
prisma/
  schema.prisma

Important: The Prisma client is generated to src/prisma/client. Always import from @/prisma/client (or use the singleton at @/lib/prisma) — never from @prisma/client.

Database

PostgreSQL with pgvector extension (required for RAG embeddings). The Embedding table and legal_knowledge_chunks table both use a native vector column.

# After pulling schema changes from git:
npx prisma migrate dev
npx prisma generate

prisma.config.ts auto-loads .env then .env.local before any Prisma CLI command, so DATABASE_URL from .env.local is picked up without extra flags.

Deployment

Deployed on Vercel. Both vercel.json (root) and src/vercel.json must be kept in sync — they define cron schedules and function timeouts. The root file is primary.

Cron jobs (all require Authorization: Bearer <CRON_SECRET>):

• /api/cron/digest-ingest — ingest legal content every 2–4 hours
• /api/cron/legal-digest — synthesise and send Briefly emails
• /api/cron/digest-cleanup — purge stale digest items
• /api/cron/trial-expiry — mark expired trials
• /api/cron/subscription-renewal — handle Paystack renewals
• /api/cron/notification-cleanup — purge old dismissed notifications
• /api/cron/onboarding-emails — drip email sequence (every 30 min)