v0.1.0

This version enables you to quickly train your own agent model.

• Environment decouple: define your tool-use envinroment easily (tools setup and reward function definition)
• Qwen3 Model support: quickly train your agent model using Qwen3 (much better than Qwen2.5 in tool-call)
• Efficient training: 2x faster than existing frameworks for rapid model iteration (mainly through async tool-use)