Introduction

What is Qualia?

Qualia is a cloud platform for fine-tuning Vision-Language-Action (VLA) models. VLAs enable you to build spatial agents that can reason about their surroundings, plan, and execute skill-based tasks.

Upload ROS 2 MCAP, LeRobot 2.1, or LeRobot v3 sources, curate episodes into immutable dataset versions, and fine-tune models such as SmolVLA, Pi0, and ACT on cloud GPUs.

The VLA Workflow

These docs cover the complete workflow for training and deploying VLA models:

1. Data Collection - Record demonstrations using LeRobot and prepare your dataset
2. Dataset versioning - Freeze an exact, reproducible episode selection
3. Fine-tuning - Train on an immutable dataset version using the dashboard or SDK
4. Inference - Deploy your fine-tuned model locally on your robot

Getting Started

1. Sign up at app.qualiastudios.dev
2. Create a project and start a fine-tuning job from the dashboard
3. To use the SDK or API, create an API key in Settings in the dashboard

Install the CLI

The qualia CLI wraps the Python SDK and installs it for you, so this is the only install needed for either.

pip install 'qualia-cli==0.2.0rc5'

# or with uv
uv tool install 'qualia-cli==0.2.0rc5' --with 'qualia-sdk==0.7.0rc4'

This deployment runs ahead of the public release, so it needs the matching pre-release wheel — 0.2.0rc5. Pin it rather than upgrading: the upload protocol is negotiated exactly, so the client and the endpoint it talks to have to be the same generation. A plain pip install qualia-cli resolves to the current public release and will be turned away by this deployment.

Then point it at this environment and sign in:

qualia auth login

See the CLI overview for the config file, the --base-url and --token overrides, and what to do when a deployment refuses your client.

Resources

• SDK Documentation - Python SDK for programmatic access
• API Reference - REST API documentation
• Discord Community - Get help and share your projects