Run your own
The app is a single Next.js project. These are the keys it needs.
Environment
| Variable | Needed for |
|---|---|
AI_GATEWAY_API_KEY | Chat, her voice and logo drawing. On Vercel the OIDC token is used and this can be left out. |
ELEVENLABS_API_KEY | Optional. Only for the Japanese library voices, which need a paid ElevenLabs plan. |
BLOB_READ_WRITE_TOKEN | Storing logos, rooms and the list of coins each room launched. |
ROOM_SECRET | Signs room sessions. Any long random string. |
AGENT_MODEL | Optional. The default brain. |
AGENT_FALLBACK_MODELS | Optional, comma separated. Tried when the chosen brain is unavailable. |
LOGO_MODEL | Optional. The image model used by drawLogo. |
NEXT_PUBLIC_TOKEN_ADDRESS, NEXT_PUBLIC_TOKEN_SYMBOL | Optional. The project token. When set, a ticker pill with a copy button and a Pons link appears in the header. |
Commands
pnpm install vercel env pull .env.local pnpm dev
node scripts/slim-models.mjs # assets/models -> public/models
Who pays for what
Visitors pay only for their own launches. The deployment pays for chat, logo drawing and voice. Those endpoints are public and have only a light per-visitor throttle, so set spending limits on the gateway and on ElevenLabs before sharing the link widely.