KalamDB Quick Start
September 5, 2026 · View on GitHub
Run a realtime React chat app with a topic worker, then open it in two tabs to see changes arrive live. The worker uses a simulated reply, so you do not need an external AI key.
1. Install the CLI
Use a current Node.js LTS release with npm. The native CLI supports macOS Apple Silicon, Linux x86-64 and ARM64, and Windows x86-64. Initial setup downloads the CLI, starter, and dependencies, so it needs internet access.
npm install -g @kalamdb/cli
kalam version
2. Create and start your app
mkdir my-app && cd my-app
kalam init --yes --template chat-with-ai --languages typescript --package-manager npm
kalam dev
kalam init downloads the chat starter and installs its dependencies. kalam dev starts or reuses a local KalamDB server, applies kalam/schema.sql, generates TypeScript tables, and runs the browser app and topic worker.
3. See a change travel through the app
Open the app URL printed in your terminal in two browser tabs. Send a message such as:
latency spike after deploy
You should see the message in both tabs, live worker progress, and a saved reply. The reply is deterministic demo logic; you can replace it with your own model call in src/agent.ts.
The starter uses local demo credentials (root / kalamdb123) and connects to http://127.0.0.1:2900 by default. These privileged demo sessions are for exploring the data flow. Use ordinary user sessions and the starter's SQL policies when building your application's access controls.
4. Make it yours
- Edit
src/App.tsxto change the interface. - Read
kalam/schema.sqlfor shared rooms, membership policies, STREAM progress events, and the topic route. - Edit
src/agent.tsto change how the worker responds to new messages. - Keep
kalam devrunning to apply schema changes and regenerate types as you work. The demo schema includes reset statements; inspect it before adapting it to data you want to keep.
The chat example guide explains the complete flow. To start smaller, run kalam init in a new empty folder and choose a minimal template. To inspect available starters, run kalam init --list-templates.
Next steps
- CLI workflow — schema generation, migrations, environments, and deployment.
- SQL reference — USER, SHARED, and STREAM tables, policies, and topics.
- API reference — SQL requests from your own client.
- WebSocket protocol — live subscriptions.
- Development setup — build KalamDB from source and contribute.