Skip to content

Commands

These flags can be used with any command:

OptionDescription
--tokenAPI key (overrides config file and env var)
--base-urlAPI base URL (overrides config file and env var)
Terminal window
qualia --token <API_KEY> project list
qualia --base-url http://localhost:8004 model types

Show CLI and SDK versions.

Terminal window
qualia version

Save your API key for future use. Prompts for the key interactively.

Terminal window
qualia auth login
OptionDescription
--api-keyAPI key (prompted if omitted)
--base-urlOverride API base URL

Check current authentication configuration.

Terminal window
qualia auth status

Show your current credit balance.

Terminal window
qualia credit
qualia credit --json

List all projects.

Terminal window
qualia project list
qualia project list --json

Create a new project.

Terminal window
qualia project create "My Project"
qualia project create "My Project" -d "Optional description"
Argument / OptionDescription
NAMEProject name (required)
-d, --descriptionOptional description

Delete a project. Fails if the project has active jobs.

Terminal window
qualia project delete <PROJECT_ID>

List available VLA model types.

Terminal window
qualia model types
qualia model types --json

List available GPU instance types and pricing.

Terminal window
qualia instance list
qualia instance list --json

Preflight and import one exact Hugging Face dataset commit on the CLI client. The command prints the exact file count, byte count, and tree digest before an interactive confirmation, then uses QDS’s staged direct-to-storage publisher. Success creates a Source and a pinned immutable DatasetVersion with a paired Git audit identity.

Terminal window
export HF_TOKEN="..." # optional for public datasets
qualia dataset import-hf \
qualia-robotics/openarm-cube-in-box-v1 \
17c29c487d886878ddcb3c5649101b2860fbc047 \
openarm-cube-in-box \
--source-format lerobot_v3

The revision must be lowercase 40-hex; branches, tags, URLs, and embedded credentials are rejected. HF_TOKEN is read only from the environment (or the Hub’s protected local credential store) and is never accepted as a command-line option. Use --max-bytes, --max-files, and --work-dir to tighten local resource bounds. Repositories with zero-byte files fail preflight because QDS cannot publish them without changing the source tree. --json requires --yes so stdout remains valid JSON.

The summary separately reports root .gitattributes/.gitignore metadata. Those two exact Git-control names are downloaded, content-verified, and bound in the audit sidecar before exclusion from the QDS payload; arbitrary hidden files are rejected.

Explain bad-data recommendations for immutable DatasetVersions without mutating them. Use --page-token to continue the digest-bound result set.

Terminal window
qualia dataset curation-suggest <DATASET_VERSION_UUID>
qualia dataset curation-suggest <DATASET_VERSION_UUID> --json

Publish the reviewed analysis’s sealed recommendation set as a new immutable DatasetVersion. Supply a stable idempotency UUID. Source bytes and Git history are not deleted.

Terminal window
qualia dataset curate <DATASET_VERSION_UUID> \
--definition-id <DATASET_DEFINITION_UUID> \
--idempotency-key <OPERATION_UUID> \
--yes

Omit --yes for an interactive confirmation.


Render and download one immutable DatasetVersion as a verified LeRobot v3 directory. Transfer bytes go directly from object storage; the CLI validates the paged artifact closure and atomically publishes a new destination.

Terminal window
qualia dataset materialize <DATASET_VERSION_UUID> ./training-set-v7
qualia dataset materialize <DATASET_VERSION_UUID> ./training-set-v7 --json
ArgumentDescription
DATASET_VERSION_IDImmutable dataset-version UUID (required)
DESTNew destination directory; must not already exist

Show the adapter-pinned camera, stream, timebase, and frame-timestamp descriptors for one immutable DatasetVersion.

Terminal window
qualia dataset training-features <DATASET_VERSION_UUID>
qualia dataset training-features <DATASET_VERSION_UUID> --json
ArgumentDescription
DATASET_VERSION_IDImmutable dataset-version UUID (required)

List jobs across all projects, or filter by project.

Terminal window
qualia job list
qualia job list --project-id <PROJECT_ID>
qualia job list --json

Launch a new fine-tuning job.

Terminal window
qualia job launch \
--project-id <PROJECT_ID> \
--vla-type smolvla \
--model-id lerobot/smolvla_base \
--dataset-version-id <DATASET_VERSION_UUID> \
--hours 2.0 \
--camera-mappings '{"cam_1": "observation.images.top"}'
OptionTypeRequiredDescription
--project-idstrYesProject ID
--vla-typestrYesModel type: smolvla, pi05, act, or sarm
--dataset-version-idUUIDYesImmutable QDS DatasetVersion
--hoursfloatYesTraining duration (max 168)
--camera-mappingsJSONYesCamera slot to image key mapping
--model-idstrNoHugging Face model ID (required for smolvla and pi05)
--model-revisionSHANoExact commit SHA; required when --model-id is not built-in
--instance-typestrNoGPU instance type (default: auto-selected)
--regionstrNoCloud region
--batch-sizeintNoTraining batch size (default: 32)
--namestrNoJob name
--hyper-specJSONNoCustom hyperparameters as JSON

Get detailed job status, including phase progression.

Terminal window
qualia job get <JOB_ID>
qualia job get <JOB_ID> --json

Cancel a running job.

Terminal window
qualia job cancel <JOB_ID>

Show default hyperparameters for a VLA model type.

Terminal window
qualia job hyperparams smolvla
qualia job hyperparams smolvla --model-id lerobot/smolvla_base
Argument / OptionDescription
VLA_TYPEVLA model type (required)
--model-idModel ID for type-specific defaults