This repository is actively evolving and not final.
- We are using the Vercel AI SDK as the foundation for agentic workflows.
- We are currently using Gemini (
@ai-sdk/google) as the LLM provider. - We are also integrating Better Auth + Drizzle for authentication and database-backed user/session handling.
- Most architecture and implementation details will continue to change as feature requirements are refined.
- Next.js App Router app
- Agentic base API endpoint at
POST /api/chat - Better Auth route base at
/api/auth/* - Drizzle migrations + PostgreSQL/Neon wiring
- Install dependencies:
pnpm install- Create local env file:
cp .env.example .env.local- Fill required values in
.env.local.
GOOGLE_GENERATIVE_AI_API_KEY=your_google_ai_studio_api_key
DATABASE_URL=postgresql://postgres:postgres@localhost:5432/taskmaster
DB_DRIVER=neon-http
BETTER_AUTH_SECRET=replace-with-a-long-random-secret
BETTER_AUTH_URL=http://localhost:3000
GOOGLE_CLIENT_ID=your-google-client-id
GOOGLE_CLIENT_SECRET=your-google-client-secretpnpm dev
pnpm lint
pnpm build
pnpm auth:generate
pnpm db:generate
pnpm db:migratecurl -N -X POST http://localhost:3000/api/chat \
-H "Content-Type: application/json" \
-d '{
"messages": [
{
"id": "u1",
"role": "user",
"parts": [{ "type": "text", "text": "Help me plan an implementation for a new TaskMaster module." }]
}
]
}'- The current agent implementation is intentionally minimal and generic.
- No domain-specific tools are pre-wired yet.
- Extend from
agent/base-agent.tsas product-specific functionality is defined.