@computesdk/tenki

Tenki Cloud provider for ComputeSDK - run code in fast microVM sandboxes with native filesystem operations, public preview URLs, snapshots, volumes, and SSH.

Installation

npm install @computesdk/tenki

Requires Node.js 20 or later (the Tenki SDK's gRPC transport depends on it).

Setup

1. Sign up at app.tenki.cloud
2. Create an API key in your workspace settings
3. Set it as an environment variable:

export TENKI_API_KEY=tk_your_api_key

Usage

With ComputeSDK

import { compute } from 'computesdk';
import { tenki } from '@computesdk/tenki';

compute.setConfig({ provider: tenki({ apiKey: process.env.TENKI_API_KEY }) });

const sandbox = await compute.sandbox.create();

const result = await sandbox.runCommand('echo "Hello from Tenki!"');
console.log(result.stdout);

await sandbox.destroy();

Filesystem operations

Tenki exposes native file primitives over its data plane, so filesystem operations do not go through shell commands (no escaping hazards, works with any path or content):

await sandbox.filesystem.writeFile('/home/tenki/app.py', 'print("hi")');
const content = await sandbox.filesystem.readFile('/home/tenki/app.py');
const entries = await sandbox.filesystem.readdir('/home/tenki');
await sandbox.filesystem.mkdir('/home/tenki/data');
const exists = await sandbox.filesystem.exists('/home/tenki/app.py');
await sandbox.filesystem.remove('/home/tenki/app.py');

Public preview URLs

// Start a server (background commands are detached automatically)
await sandbox.runCommand('python3 -m http.server 3000', { background: true });

// Expose the port at a public URL
const url = await sandbox.getUrl({ port: 3000 });
// => https://<slug>.sb.tenki.sh

Configuration

| Option | Env var | Default | Description | |---|---|---|---| | apiKey | TENKI_API_KEY / TENKI_AUTH_TOKEN | required | Tenki API key (tk_...) | | baseUrl | TENKI_API_URL | https://api.tenki.cloud | API endpoint | | workspaceId | TENKI_WORKSPACE_ID | auto-resolved | Workspace for new sandboxes | | projectId | TENKI_PROJECT_ID | auto-resolved | Project for new sandboxes | | timeout | - | none | Default runCommand timeout (ms) | | cpuCores / memoryMb / diskSizeGb | - | Tenki defaults | Default sandbox resources |

When workspaceId/projectId are not set, the provider resolves them once from the API key's identity (first workspace with a project).

Per-sandbox resource overrides are accepted on create:

await compute.sandbox.create({ cpuCores: 4, memoryMb: 8192 });

Feature support

| Feature | Supported | |---|---| | Commands (runCommand) | Yes - via sh -lc, so pipes, globs, and env expansion work | | Filesystem | Yes - native data-plane file API | | Public URLs (getUrl) | Yes - per-port preview URLs | | List / getById / destroy | Yes | | Background commands | Yes - { background: true } detaches stdio automatically | | Custom images | Via Tenki templates (templateId maps to a Tenki image ref) | | Snapshots / pause-resume | In the Tenki SDK (@tenkicloud/sandbox); not yet wired to provider snapshot methods |

Notes

• runCommand wraps the command in sh -lc because Tenki's exec runs argv directly (execve, no shell).
• A long-running process started with a bare & would hold the exec output stream open; use { background: true }, which detaches stdio for you.
• For advanced features (SSH, volumes, tunnels, git operations, snapshots), use the underlying SDK directly via sandbox.getInstance(), which returns the @tenkicloud/sandbox Session.

License

MIT