@dbx-tools/ui-lakebox

Render Lakebox lifecycle controls, user services, and live machine metrics in an AppKit React application.

Render the Lakebox page

import { LakeboxIndicator, LakeboxPanel } from "@dbx-tools/ui-lakebox/react";

export function LakeboxPage() {
  return (
    <>
      <header>
        <LakeboxIndicator />
      </header>
      <LakeboxPanel />
    </>
  );
}

Import the stylesheet once after Tailwind and your AppKit UI theme:

@import "@dbx-tools/ui-lakebox/styles.css";

The panel shows start and stop controls, the current user's terminal, desktop, image services, bounded logs, and removal confirmation. It streams CPU, memory, persistent storage, and network metrics while the Lakebox runs.

The indicator fits in a chat header. Its popover has lifecycle controls, terminal and desktop launchers, and compact metrics.

Register the backend plugin

The React components require @dbx-tools/appkit-lakebox in the same AppKit application:

import { createApp, lakebase, server } from "@databricks/appkit";
import { lakebox } from "@dbx-tools/appkit-lakebox";

await createApp({
  plugins: [server(), lakebase(), lakebox()],
});

The backend plugin publishes basePath and routePrefix through AppKit client configuration. The UI reads that configuration, derives routes from @dbx-tools/shared-lakebox, and sends same-origin requests. Don't copy plugin paths into the client.

Use the browser client directly

Use LakeboxClient when you need lifecycle or service actions outside the provided components:

import { useLakeboxClient } from "@dbx-tools/ui-lakebox/react";

export function StartLakeboxButton() {
  const client = useLakeboxClient();
  return (
    <button type="button" onClick={() => void client.start()}>
      Start Lakebox
    </button>
  );
}

The client provides status, start, stop, service create, service start and stop, service removal, bounded logs, current metrics, and an SSE metrics stream.

Understand UI behavior

• Components use the current AppKit session. They never accept or store Databricks credentials.
• The panel reports when Lakeboxes aren't available in the workspace.
• Opening the indicator starts status and metrics activity. Closing it stops the indicator's live metrics stream.
• Machine metrics are sampled only while a client watches them, so the UI doesn't keep an idle Lakebox active by polling metrics.
• Start and stop affect the one Lakebox shared by the app. Service controls affect only the authenticated user's services.
• Terminal and desktop links use authenticated same-origin proxy routes. The browser never connects to the SSH gateway directly.