AgPerry's picture
Add files using upload-large-folder tool
9d4a3b0 verified
|
Raw History Blame Contribute Delete
6.37 kB

LinkedIn professional network

LinkedIn is an independent, enterprise-deployable professional social network built as a local research replica of common professional-networking interaction patterns. It does not use LinkedIn branding, does not call LinkedIn APIs, and never requests LinkedIn credentials.

Implemented scope

  • Public homepage with original LinkedIn branding and neutral professional imagery
  • Public Learning, Top Content, Games, and directory authwall routes from the observed public interaction matrix
  • Local email/password registration with email verification and PBKDF2 password hashing
  • HTTP-only database-backed sessions
  • Dynamic feed, post composer, likes, comments, optimistic UI states, and API persistence
  • Profile pages, experience data, profile editing API, relationship invitations, suggestions, and notifications
  • Messaging inbox with conversation persistence endpoints
  • Job search, detail view, API-confirmed saved jobs, application API, and recruiter job creation API with RBAC
  • Company pages and company/job APIs
  • Global search across people, jobs, companies, and posts
  • LinkedIn Premium subscription surface backed by Stripe Checkout, Billing Portal, and signed webhooks in Test Mode
  • Multi-tenant foundation through organizations, memberships, roles, tenant-scoped jobs, and audit logs
  • Loading, empty, and error states across the app shell
  • SQLite/D1 schema, migration SQL, deterministic demo seed data, tests, Docker, Nginx, Apache, systemd, health checks, and deployment preflight validation

Stack

  • Next.js 16 App Router + TypeScript + React 19
  • Drizzle ORM + Cloudflare D1/SQLite
  • Zod request validation
  • Node crypto PBKDF2 password hashing
  • Nodemailer transactional email adapter with a local development outbox
  • Stripe server SDK for test-mode subscription checkout and webhook reconciliation
  • Docker Compose with persistent SQLite, app, and Nginx reverse proxy

Local development

npm install
export LINKEDIN_SQLITE_PATH=.data/linkedin.sqlite3
npm run db:migrate
npm run dev

Open http://localhost:3000.

The migration includes the deterministic northstar-systems tenant and demo Feed/Profile records. The TypeScript seeder remains available for idempotent reseeding:

LINKEDIN_SQLITE_PATH=.data/linkedin.sqlite3 npm run db:seed

For a local WebsiteBench-style registration check on port 3103:

LINKEDIN_SQLITE_PATH=.data/linkedin.sqlite3 \
DEFAULT_TENANT_SLUG=northstar-systems \
SESSION_SECRET=$(openssl rand -hex 32) \
SESSION_COOKIE_SECURE=false \
PUBLIC_CLONE_AUTH_MODE=local-dev \
PUBLIC_CLONE_AUTH_LOCAL_DEV_CODE=123456 \
PORT=3103 HOSTNAME=127.0.0.1 npm run start

Then open http://127.0.0.1:3103/public-register and use local verification code 123456. Public deployment uses the authenticated LINKEDIN_D1_PROXY_URL bridge instead of a local database path.

Seeded test account:

maya@example.com / LinkedInDemo123

Validation

npm run typecheck
npm run lint
npm run build
npm test
npm run test:flow
npm run preflight -- --production

npm run verify runs the full validation sequence.

To run the database-backed registration journey after SQLite is initialized:

BASE_URL=http://localhost:3000 npm run test:e2e:registration

That E2E script registers a new account, consumes the local .linkedin-mail verification link when EMAIL_DELIVERY_MODE=console, creates a post, submits a job application, and checks the Premium guardrail. The main test suite also runs a real standalone production server against HTTP Redis/Resend and D1 mocks, then restarts it to verify session and post persistence.

Docker deployment

docker compose up --build

Compose runs the one-shot SQLite migrate service against a persistent volume, then starts the app and Nginx. To rerun the migration job manually:

docker compose run --rm migrate

Open http://localhost:8080 through Nginx.

WebsiteBench public deployment

WebsiteBench public deployment is packaged separately in sources/w2w_workflow/examples/linkedin-public. That flow uses /public-register, /api/auth/send-code, /api/auth/verify-code, and the vendored websitebench.public_clone_auth runtime. It does not use the local SMTP/nodemailer registration path or require Stripe checkout for core validation; Stripe remains workflow-pending for public deployment until WebsiteBench provides an approved adapter.

Environment

Required production values:

LINKEDIN_D1_PROXY_URL=http://d1.internal/linkedin-db
SESSION_SECRET=<32+ character random secret>
EMAIL_VERIFICATION_SECRET=<32+ character random secret>
SESSION_COOKIE_SECURE=true
APP_URL=https://linkedin.example.com
APP_ENV=production
DEFAULT_TENANT_SLUG=northstar-systems
EMAIL_DELIVERY_MODE=smtp
MAIL_FROM="LinkedIn <no-reply@linkedin.example.com>"
SMTP_HOST=<smtp host>
SMTP_PORT=587
SMTP_SECURE=false
SMTP_USER=<smtp user>
SMTP_PASS=<smtp password>
STRIPE_SECRET_KEY=<Stripe Test Mode secret key>
STRIPE_PREMIUM_PRICE_ID=price_...
STRIPE_WEBHOOK_SECRET=<Stripe Test Mode webhook signing secret>

Before promoting a release, run a strict deployment preflight with production environment values loaded:

npm run preflight -- --production

The preflight prints a sanitized JSON report. It checks the D1 proxy, session and email-verification secrets, HTTPS app URL, secure cookies, SMTP delivery, default tenant, and Stripe Test Mode configuration without printing secret values.

Security boundary

  • This project creates only local LinkedIn accounts.
  • New local accounts must verify email before receiving a session.
  • The login page explicitly states that it is not a LinkedIn sign-in.
  • Billing code rejects non-test Stripe secret keys in this environment.
  • No route asks for LinkedIn credentials, tokens, cookies, or profile export files.
  • Mutating API routes reject cross-origin writes.
  • Tenant write operations check membership role before creating recruiter-owned resources.

Production hardening checklist

  • Put D1 behind managed backup and restore drills.
  • Run migrations in CI/CD before promoting a release.
  • Add centralized logging and trace request IDs before accepting real user data.
  • Add password reset, MFA, and rate limiting before public launch.
  • Complete accessibility, privacy, and security review.