> ## 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 CLI and Python SDK as the supported public integration surfaces.
> Prefer `dreamscale run <robot>` for supported robots and `dreamscale.connect()` with a context manager in Python.
> Do not infer or recommend internal control-plane APIs.

# Inspect Robots

> Record Dreamscale robot runs as Inspect Robots eval logs, or use Dreamscale models as a policy inside Inspect Robots.

[Inspect Robots](https://github.com/robocurve/inspect-robots) is an evaluation
framework for robot policies. Use it with Dreamscale in one of two ways:

* **Record a `dreamscale run`** as an eval log. Easiest if your robot is set up
  with Dreamscale.
* **Use `--policy dreamscale`** inside your own Inspect Robots evaluation.

## Record a run as an eval log

Add `--inspect-robots` to `dreamscale run`. The first run installs Inspect
Robots.

```bash theme={null}
dreamscale run yam --model pi05-yam --inspect-robots
```

The run works as usual. After each task, while the robot holds, grade it: `y`
for success, `n` for failure, or a score from 0 to 1, optionally followed by a
note. Then view the log:

```bash theme={null}
inspect-robots view logs/
```

Each instruction becomes a scene and each task a trial.

| Option | What it does |
| - | - |
| `--instruction TEXT --epochs N` | Repeat one task `N` times |
| `--task NAME` | Run a registered Inspect Robots task's scenes in order, at its own horizon |
| `--log-dir PATH` | Where to write logs (default `./logs`) |

For example, five timed trials of one task:

```bash theme={null}
dreamscale run yam --model pi05-yam --inspect-robots \
  --instruction "fold the towel" --epochs 5 --max-actions 600
```

`--inspect-robots` works with every `dreamscale run` option except `--hud`.

## Use Dreamscale as a policy

In the environment where you run Inspect Robots:

```bash theme={null}
dreamscale integrations inspect-robots setup
```

This installs Inspect Robots, signs you in if needed, and registers the
`dreamscale` policy and the `dreamscale_yam` embodiment. Then select the policy
with `-P model=`:

```bash theme={null}
inspect-robots run --policy dreamscale -P model=pi05-yam \
  --embodiment dreamscale_yam "fold the towel"
```

`dreamscale_yam` uses the rig you set up with `dreamscale robots yam setup`. To
use your own YAM embodiment instead, run it at 30 Hz, for example
`--embodiment yam_arms -E control_hz=30`.

Supported models: `molmoact2-bimanual-yam`, `pi05-yam`, and
`perceptron-isaac05-yam` (with `-P checkpoint=ckpt_...`).

| Parameter | Default | What it does |
| - | - | - |
| `model` | `molmoact2-bimanual-yam` | Model to run |
| `keep_warm` | `300` | Seconds to keep the GPU reserved after the run (billed). `0` releases it |
| `region` | `available` | Region preference |
| `speed` | `standard` | `fast` where the model offers it |
| `cameras` | `top_cam,left_cam,right_cam` | Your embodiment's top, left wrist, and right wrist camera names |
| `state_key` | `joint_pos` | State key holding the 14 joint values |
| `timeout_s` | `60` | Per-step deadline |

<Warning>
  Your embodiment sets the pace. If its `step()` returns faster or slower than
  30 Hz, the model's actions arrive at the wrong time and the robot holds more
  often, while the run still reports success. Pace the embodiment at 30 Hz.
</Warning>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.