Skip to main content

Run your first evaluation

1. Install v0.0.1a1​

Use Python 3.10 or newer in a virtual environment. Pin both the SDK and shared interface to this release:

python -m venv .venv
source .venv/bin/activate
python -m pip install --pre "se3labs==0.0.1a1" "se3labs-interface==0.0.1a1"
se3labs --version

On Windows, activate with .venv\Scripts\Activate.ps1 in PowerShell.

2. Run a robot policy​

Ask your operator for the station ID and relay address, and use a station and relay from the same release. This SDK does not attach authentication metadata and cannot connect to the current authenticated relay.

With the station ready for evaluation, run:

se3labs eval run \
  --station yam-1 \
  --address relay.example.com:7443 \
  --secure \
  --request se3labs.sdk.examples.hold_still:HOLD_STILL_REQUEST

Replace the station and address with your operator's values. --secure enables TLS; omit it for a local plaintext connection.

The bundled policy moves to zero joint angles with open grippers and then holds the observed positions for a five-second episode. Confirm that the start pose is appropriate for the setup before the operator begins the rollout.

The CLI prints the session's close reason, episode count, and step count. This version has no auth commands or --output option.

3. Check status​

se3labs eval status \
  --station yam-1 --address relay.example.com:7443 --secure --json

The status includes completed and total episode counts and the success count. See the CLI reference or write a policy next.