v0.0.7

Features

• Add a Codebreaker TUI agentic workflow (#182).
• Add the OLMo 2 model adapter (#176).
• Add repository-local AReno agent skills (#175).
• Support independent rollout policy synchronization (#434).
• Add Qwen3.5 image-to-text support (#152).
• Add Ling Tiny V3 support (#462).

Fixes

• Infer rollout devices from world size in the training CLI (#444).
• Handle empty features in PPO role scoring (#475).

Maintenance

• Align dashboard operations workflows with the PRD (#446).
• Streamline installation and Quick Start documentation (#181, #481).
• Add community meeting notes and DGX Spark training guides (#456, #470, #476).
• Use the published flash-linear-attention==0.5.2 package in distribution metadata.
• Bump the package and container version to v0.0.7.

New Contributors

• @lizongyu0912 made their first contribution in #456.

Full Changelog: v0.0.6...v0.0.7