Kavla CLI

Connect data sources to a Kavla canvas.

Receive SQL from the browser, execute it locally, and relay previews back.

Your credentials never leave your machine.

Architecture

• Every Kavla canvas has its own isolated backend
• A CLI session connects to one canvas over a dedicated websocket
• When it connects the CLI shares its configured sources to the canvas
• Only the CLI owner can execute queries against the CLI sources
• SQL can now be sent from the canvas to the CLI
• Credentials are never sent, they stay local
• When the CLI is terminated, the connection drops instantly
• Preview results are stored on the canvas after the CLI disconnects
• Results follow the canvas access rules and need to be explicitly shared

Supported Sources

Source	Type	Connection
DuckDB	duckdb	Path to an existing DuckDB file
Directory	directory	Path to a directory with CSV, Parquet, or JSON files
BigQuery	bigquery	Google Cloud project ID
Postgres	postgres	Postgres URI

Quick start

Install

curl -fsSL https://raw.githubusercontent.com/aleda145/kavla-cli/main/install.sh | bash

Setup

kavla init

Add a source

kavla config add-source

Login

kavla login

Connect to a canvas

kavla connect

Or connect directly:

kavla connect <room_id>

Config

Config lives in ~/.kavla/config.yaml. Tokens and source definitions stay on your machine.

Support

kavla.dev

Discord