Physlint 0.1.0a1 — Public Alpha

Physlint 0.1.0a1 is the first public alpha of a local-first integrity validator for physical-AI recordings and robot-learning datasets.

Highlights

• 17 deterministic LeRobot v3 integrity rules
• Stable CI exit codes and machine-readable JSON reports
• FPS-aware timing and gap validation
• Privacy-safe shared video analysis
• Grouped evidence for repeated gap and black-frame findings
• Hugging Face repository identity and pinned revision provenance
• Separate NaN/Inf and null/empty overlap diagnostics
• Durable validation evidence from four datasets and three controlled corruptions

Validation

The release gate passed:

• 46 automated tests
• Ruff lint and formatting checks
• Strict mypy type checking
• Wheel and source-distribution builds
• Four real public LeRobot datasets
• Three controlled corruption scenarios

Install

pip install "physlint[video]==0.1.0a1"

Until the PyPI upload completes:

pip install "physlint[video] @ git+https://github.com/pranayjoshi/physlint.git@v0.1.0a1"

Current scope

This alpha supports local LeRobot Dataset v3.x directories. MCAP/ROS 2 and Robomimic support are roadmap items and are not yet implemented.

See the README for usage, validation evidence, CI integration, and the roadmap.