GPU Cloud Dashboard

Built on hosted.ai | Next.js 16 | Prisma/MariaDB | Stripe

A full-featured GPU cloud platform dashboard powered by the hosted.ai GPU infrastructure API. Fork and deploy to run your own GPU-as-a-Service business.

Install

curl -fsSL https://raw.githubusercontent.com/hosted-ai/packet-oss/main/install.sh | sudo bash

Installs to /opt/packet-oss with MariaDB, systemd, and Apache reverse proxy. Requires a fresh Linux server (Ubuntu/Debian).

---

This platform requires hosted.ai. All GPU pod deployment, orchestration, scaling, monitoring, and billing runs through the hosted.ai API. You need a hosted.ai account and API credentials for the platform to function. Visit hosted.ai to get access.

Features

• Customer Dashboard - Deploy & manage GPU instances, SSH keys, team members, billing
• Admin Panel - 30+ tabs for managing customers, pods, products, pricing, providers, and more
• Provider Portal - GPU providers can manage nodes, pricing, and payouts
• Investor Portal - Revenue tracking and performance dashboards
• Stripe Billing - Optional prepaid wallets, subscriptions, and payment processing
• Web SSH Terminal - Browser-based terminal access to GPU instances
• Marketing Pages - Landing page, GPU product pages, comparison pages, docs
• Two-Factor Auth - TOTP-based 2FA for admin security
• Referral & Voucher System - Built-in referral codes and discount vouchers
• Platform Settings - Configure all API keys and settings from the admin UI

Prerequisites

• hosted.ai account - Required for GPU pod management
• Node.js 18+ and pnpm 8+
• MariaDB 10.6+ (or MySQL 8.0+)
• Domain + SSL/HTTPS - Required for any non-localhost deploy. Admin session cookies are marked secure, so login breaks over plain HTTP. install.sh provisions this automatically via Let's Encrypt + Apache when you supply a domain.
• Stripe account (optional) - For billing features

Manual Setup

# 1. Clone and install
git clone https://github.com/hosted-ai/packet-oss.git && cd packet-oss
pnpm install

# 2. Create environment file
cp .env.example .env.local

# 3. Configure database URL in .env.local
# DATABASE_URL="mysql://gpucloud:password@localhost:3306/gpucloud"

# 4. Initialize the database
npx prisma db push

# 5. Start development server
pnpm dev

# 6. Set up your admin account
# Visit http://localhost:3000/admin
# Create your admin account with email + password
# Go to Platform Settings to configure your hosted.ai API keys

Configuration

Zero-Config Boot

The app boots and runs with just DATABASE_URL configured. All other settings can be configured through the admin panel at Admin > Platform Settings.

Important: While the app starts without any API keys, GPU features (pod deployment, management, monitoring) require a hosted.ai API connection. Configure your hosted.ai credentials in Platform Settings after your first login.

Environment Variables

*hosted.ai credentials are required for GPU functionality but the app will boot without them. Configure via admin UI or .env file.

Tech Stack

• Framework: Next.js 16 (App Router, React 19)
• Database: MariaDB with Prisma ORM
• Styling: Tailwind CSS 4
• Payments: Stripe (optional)
• GPU Backend: hosted.ai API (required)
• Auth: JWT (jsonwebtoken) with password + optional TOTP 2FA
• Process Manager: PM2

Architecture

src/
  app/                    # Next.js App Router
    (marketing)/          # Public marketing pages
    account/              # Customer account
    admin/                # Admin panel (SPA, tab routing via ?tab=)
    api/                  # API routes
      admin/              # Admin API
      cron/               # Scheduled jobs
      webhooks/           # Stripe webhooks
    dashboard/            # Customer GPU dashboard
    terminal/             # Web SSH terminal
  components/             # Shared React components
  lib/                    # Server-side business logic
    auth/                 # JWT auth (admin + customer)
    email/                # Email templates
    hostedai/             # GPU infrastructure client
    settings.ts           # Platform settings (DB-backed)
  middleware.ts           # POST request blocking
prisma/schema.prisma      # Database schema

CLI Tool

A command-line interface is included for managing GPU instances:

cd cli && npm install -g .

gpu-cloud login
gpu-cloud gpus
gpu-cloud launch --gpu h100 --setup vscode
gpu-cloud ps
gpu-cloud ssh <instance-id>

See cli/README.md for full documentation.

Production Deployment

# Build for production
pnpm build

# Start with PM2
pm2 start ecosystem.config.cjs

# Or start directly
pnpm start

See install.sh for automated server setup with systemd and Apache.

Upgrade

sudo bash upgrade.sh                    # Upgrade to latest
sudo bash upgrade.sh --branch v1.2.0    # Upgrade to specific version

Reconfigure

sudo bash reconfigure.sh                         # Interactive menu
sudo bash reconfigure.sh --domain new.example.com # Change domain
sudo bash reconfigure.sh --ssl-on                 # Enable SSL
sudo bash reconfigure.sh --check                  # Health diagnostics

Development

pnpm dev              # Start dev server
pnpm build            # Production build
pnpm lint             # ESLint
pnpm test:unit        # Vitest tests
pnpm test:e2e         # Playwright E2E tests
npx tsc --noEmit      # Type check

GPU Backend

This platform is built on the hosted.ai GPU infrastructure API. All GPU pod lifecycle management - deployment, scaling, monitoring, billing, and orchestration - is handled through hosted.ai. You must have a hosted.ai account with API credentials to operate GPU features.

To get started with hosted.ai, visit hosted.ai.

License

MIT