> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ariacompute.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Aria Compute SDKs and client libraries

> Pick the right client for Aria Compute: REST API from any language, on-device runtimes (aria-engine, aria-router), and agent tooling.

Aria Compute exposes two integration surfaces. Pick the one that matches your workload.

## REST API

Every dashboard action is available as a REST call. Any HTTP client works. Use it to authenticate users, manage API keys, top up the wallet, list and download models, and read billing data.

* Base URL: `https://ariacompute.com/api` (international) or `https://ariacompute.cn/api` (China).
* Authentication: session JWT or `bfvk-` API key. See [Authentication](/authentication).
* Reference: [API Reference](/api-reference/introduction).

<CardGroup cols={2}>
  <Card title="cURL" icon="terminal" href="/sdks/curl">
    Shell recipes for auth, downloads, and billing.
  </Card>

  <Card title="Python" icon="python" href="/sdks/python">
    Minimal `requests`-based client with retries.
  </Card>

  <Card title="Node.js" icon="node-js" href="/sdks/nodejs">
    Fetch-based client for server-side integrations.
  </Card>

  <Card title="Go" icon="golang" href="/sdks/go">
    `net/http` client that mirrors the backend's own patterns.
  </Card>
</CardGroup>

## On-device runtimes

The runtimes run Aria model bundles on user hardware. They ship as release binaries and FFI shared libraries.

| Runtime                              | Purpose                                               | Reference                              |
| ------------------------------------ | ----------------------------------------------------- | -------------------------------------- |
| `aria-engine` + `libaria-engine_ffi` | Inference for language, vision, and VLA/policy models | [Engine runtime](/sdks/engine-runtime) |
| `aria-router` + `libaria-router_ffi` | Model dispatch and multi-model orchestration          | [Router runtime](/sdks/router-runtime) |
| `aria-agent`                         | Agent scaffolding for tool use and planning           | [Downloads](/resources/downloads)      |
| `aria-memo`                          | Persistent memory for Aria agents                     | [Downloads](/resources/downloads)      |

<Note>
  Fetch the latest release binaries programmatically via the [release endpoints](/api-reference/downloads/engine-latest). On the China site, downloads are served from a Gitee mirror.
</Note>

## Model bundles

All runtimes consume the same [Aria model bundle](/concepts/models) format (`weight.bin`, `config.json`, optional tokenizer sidecars). Download bundles with an API key using [`GET /api/models/{slug}/download`](/api-reference/models/download), or pull public mirrors from Hugging Face (`ariacompute/*`) or ModelScope (`AriaCompute/*`).


## Related topics

- [Python client for the Aria Compute API](/sdks/python.md)
- [Go client for the Aria Compute API](/sdks/go.md)
- [Node.js client for the Aria Compute API](/sdks/nodejs.md)
- [aria-router runtime and FFI](/sdks/router-runtime.md)
- [aria-engine runtime and FFI](/sdks/engine-runtime.md)