A: (1) Local deployment: You can follow the instructions in the README to deploy the agent model locally. (2) Production service: For a production-ready service, visit bailian and follow the guided setup.
A: We don’t plan to release the training data, but we have open-sourced the methods for synthesizing it in our series of papers.
A: We’ve already open-sourced the code for ReAct inference in our papers. The Heavy Mode is not fully open-sourced yet, but we plan to release it in the future.
A: The agent model was trained using specific prompts and tools. To reproduce the results, please make sure you’re using the same prompts and tools released in the codebase. We’re also working on developing a more general agent model.