Eryx

AI chat platform with multiple LLMs, streaming responses, and transparent pricing.

• Multi-Model - Access GPT-OSS, Llama 4, Llama 3.3, and more from a single interface
• Real-time Streaming - Server-sent events for instant responses
• Pay-as-you-go - Transparent per-token pricing with usage dashboard
• Edge-first - Powered by Cloudflare Workers AI for global low-latency inference

Stack

Frontend: Nuxt 4, Vue 3, Nuxt UI, Tailwind CSS, TypeScript Backend: Nitro, Cloudflare Workers, D1 (SQLite), R2 AI: Cloudflare Workers AI (multi-model) Auth: Google OAuth Payments: Lemon Squeezy

Development

Prerequisites

• Node.js >= 18
• pnpm, npm, or bun
• Cloudflare account (for D1, R2, Workers AI)

Setup

pnpm install

Environment Variables

Create .env from .env.example:

GOOGLE_CLIENT_ID=your_google_client_id
LEMONSQUEEZY_API_KEY=your_api_key
LEMONSQUEEZY_WEBHOOK_SECRET=your_webhook_secret
LEMONSQUEEZY_STORE_ID=your_store_id
SESSION_SECRET=min_32_chars_secret

Commands

# Development
pnpm dev              # Dev server on localhost:3000

# Build
pnpm build            # Build for Cloudflare Pages
pnpm preview          # Preview production build

# Database
npx wrangler d1 migrations apply eryx-db

Architecture

app/
├── composables/      # Vue composables (auth, etc.)
├── layouts/          # App layouts
└── pages/            # Routes
    ├── index.vue           # Home / new chat
    ├── chat/[id].vue       # Chat conversation
    ├── dashboard.vue       # Usage analytics
    ├── pricing.vue         # Pricing info
    └── login.vue           # Google OAuth

server/
├── api/
│   ├── auth/         # Authentication endpoints
│   ├── ai/           # Chat completion (streaming & batch)
│   ├── chats/        # Chat CRUD operations
│   ├── subscription/ # Subscription status
│   └── webhook/      # Lemon Squeezy webhooks
├── middleware/       # Auth middleware
└── utils/            # Session management

migrations/           # D1 database schema

Available Models

Model	Provider	Description
GPT-OSS 120B	OpenAI	Large general-purpose model
GPT-OSS 20B	OpenAI	Efficient general-purpose model
Llama 4 Scout 17B	Meta	Latest Llama architecture
Llama 3.3 70B	Meta	High-performance open model
Llama 3.1 8B	Meta	Fast, lightweight model

Deployment

Cloudflare Pages

# Deploy
npx wrangler pages deploy dist

# Or connect GitHub repo for auto-deploy

Wrangler Configuration

Edit wrangler.toml:

[[d1_databases]]
binding = "DB"
database_name = "eryx-db"
database_id = "your-database-id"

[[r2_buckets]]
binding = "UPLOADS"
bucket_name = "eryx-uploads"

[ai]
binding = "AI"

API Endpoints

Method	Endpoint	Description
POST	/api/auth	Google OAuth login
GET	/api/auth/me	Current user info
POST	/api/ai/chat-stream	Streaming chat completion
GET	/api/chats	List user's chats
POST	/api/chats	Create new chat
GET	/api/dashboard	Usage analytics
GET	/api/subscription/status	Subscription status

License

MIT