Docs / Self-hosting
Deploy
Run your own T Chat on Vercel with a Neon database and any OpenAI-compatible model.
Requirements
- A Vercel account (the Hobby plan is enough).
- A Neon Postgres database, added from the Vercel Storage tab.
- An API key for an OpenAI-compatible model provider. DeepSeek is the default.
Environment variables
| Name | Required | Purpose |
|---|---|---|
DATABASE_URL | Yes | Neon connection string. Set automatically by the Vercel integration. |
DEEPSEEK_API_KEY | Yes | Provider API key. Server-side only. |
ADMIN_PASSWORD | Yes | Console password, 8 or more characters. Use a long random one. |
AI_BASE_URL | No | Provider base URL. Default https://api.deepseek.com. |
AI_MODEL | No | Model name. Default deepseek-chat. |
Steps
- Import the repository into Vercel and set the Framework Preset to Next.js.
- Add the Neon integration, then the variables above.
- Deploy, then attach your domain under Settings → Domains.
- Open
/login, sign in and create your first member.
Running locally
npm install
vercel env pull .env.local
npm run devLocal runs use the same database as the variables you pull, so users you create are real.