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

> ## Agent Instructions
> Treat the Dreamscale Python SDK and CLI as the supported public integration surfaces.
> Prefer context-managed sessions and model-specific observation helpers.
> Do not infer or recommend internal control-plane APIs.

# Installation

> Install the Dreamscale SDK and CLI, authenticate, and keep the launch package up to date.

Dreamscale supports Python 3.11 through 3.14 for cloud policy inference. The
current release is `0.1.0a26`. Install only the extras required by your
local robot or simulator stack.

## Choose an install

<CodeGroup>
  ```bash uv (recommended) theme={null}
  uv add dreamscale
  uv tool install dreamscale
  ```

  ```bash pip theme={null}
  python -m pip install dreamscale
  ```
</CodeGroup>

The base package includes the Python SDK, CLI, transport, and observation
helpers.

<Note>
  Dreamscale was previously called Dropbear. The legacy `dropbear` package,
  `import dropbear`, and the `dropbear` command still work as aliases, so
  existing projects keep running unchanged, and an existing `~/.dropbear`
  sign-in is read in place when there is no `~/.dreamscale`. Use `dreamscale`
  for new work.
</Note>

<Note>
  Dreamscale is pre-release, so install the latest version rather than pinning an
  old one. Each release tracks what the hosted models actually serve; an
  out-of-date SDK can disagree with the live control rate for a model. Pin an
  exact version (`dreamscale==<version>`) when you need a reproducible run, and
  refresh the pin when you upgrade.
</Note>

| Install             | Use it for                                               | Compatibility    |
| ------------------- | -------------------------------------------------------- | ---------------- |
| `dreamscale`        | Existing robot controllers and cloud inference           | Python 3.11–3.14 |
| `dreamscale[so101]` | Dreamscale's SO-101 setup, calibration, and dry-run path | Python 3.12–3.13 |
| `dreamscale[sim]`   | The local `dreamscale sim` LIBERO path                   | Linux or WSL2    |

Install an extra in the project that owns the corresponding hardware or
simulator:

<CodeGroup>
  ```bash SO-101 theme={null}
  uv add "dreamscale[so101]"
  ```

  ```bash Local LIBERO theme={null}
  uv add "dreamscale[sim]"
  ```
</CodeGroup>

<Note>
  Your own Franka or simulator integration should use its existing controller
  dependencies plus the base `dreamscale` package. Dreamscale does not provide a
  Franka controller.
</Note>

## Sign in

Browser login is the default:

```bash theme={null}
dreamscale login
```

The CLI opens a one-time approval page. After approval, it verifies the
credential and writes `~/.dreamscale/config.toml` with user-only permissions.
A sign-in from an earlier install keeps working; see the note under
[Choose an install](#choose-an-install).

For a headless machine, paste an API key into the hidden prompt:

```bash theme={null}
dreamscale login --api-key
```

You can also pass an API key as the option value, but shell history and process
inspection can expose it. Prefer the hidden prompt or an existing secret
manager.

<Warning>
  Do not commit `~/.dreamscale/config.toml`, paste keys into source files, or log
  them in CI. Create and revoke keys in the
  [dashboard](https://app.dreamscalelabs.com/dashboard/keys).
</Warning>

## Verify the machine

```bash theme={null}
dreamscale doctor
dreamscale status
```

`doctor` checks the robot-neutral installation, authentication, and network
path. `status` reports live model-region capacity without opening a session.

If either command fails, use the command's printed recovery step and then open
[Troubleshooting](/troubleshooting).

## Upgrade

Update both the project dependency and standalone CLI so their behavior stays
aligned:

<CodeGroup>
  ```bash uv theme={null}
  uv add --upgrade dreamscale
  uv tool upgrade dreamscale
  dreamscale --help
  ```

  ```bash pip theme={null}
  python -m pip install --upgrade dreamscale
  dreamscale --help
  ```
</CodeGroup>

You are ready when `dreamscale --help` lists `login`, `doctor`, `sim`, `status`,
`run`, `robots`, and `sessions`.

Continue to the [Python quickstart](/quickstart) to request one non-actuating
prediction.