Release Microduck Studio v0.1.0 · microai-lab/microduck-studio

Microduck Studio v0.1.0

Microduck Studio is a local developer console that connects the microduck
robot runtime with microduck_rl simulation and training tools.

This initial release brings browser-based control, runtime visibility, and
repeatable startup into one development workflow.

Bilingual Web controls with Chinese labels and English subtitles.
Movement, enable/stop, and skill commands through the existing robotd interface.
Runtime, MuJoCo, repository, policy, and job status in one place.
One-command macOS startup for Docker-based robotd and Studio, plus a native MuJoCo Viewer.
End-to-end startup verification that confirms control commands move the simulated robot.
Direct access to robotctl monitor through a disposable Docker Compose container.
Optional, allowlisted RL smoke tests, disabled by default.
English and Chinese setup guides, including prompts for local AI coding agents.

microduck owns runtime control, policy inference, hardware I/O, and safety.
microduck_rl owns MuJoCo simulation, RL training, and policy export.
microduck-studio owns the Web interface, status aggregation, and local orchestration.

The complete development stack requires all three repositories in sibling directories.
The one-command launcher currently targets macOS and requires Docker with Compose.
MuJoCo integration uses the sim-remote-io runtime branch.
Studio has no authentication; keep it on localhost or a trusted network.
Backward walking depends on the loaded policy and is not guaranteed by this release.

Follow the README to prepare the repositories and dependencies, then run:

./scripts/dev-stack.sh

Wait for control probe passed before using the Web controls.

Full Changelog: https://github.com/microai-lab/microduck-studio/commits/v0.1.0