hermes-atropos-environments

To build a Hermes RL environment, implement HermesAgentBaseEnv in environments/hermes_base_env.py and wire an evaluation loop that uses tools with wandb logging for traceability across serve, process, and evaluate CLI modes.

HermesAgentBaseEnv requires setup, get_next_item, format_prompt, compute_reward, evaluate, and wandb_log methods, plus a config class and operator-facing CLI guidelines for end-to-end evaluation.

Yes, Hermes Atropos environments support cross-platform execution on Linux, macOS, and Windows, enabling consistent RL policy testing and tool-enabled evaluation across operating systems.

Wandb logging in Hermes RL environments provides traceability for tool-context based evaluation and reward computation, capturing metrics throughout the serve, process, and evaluate CLI modes.

End-to-end testing of Hermes agent environments uses the full HermesAtropos loop via serve, process, and evaluate CLI modes, validating agent loop orchestration with tool use and reward functions.

Yes, a config class is required alongside the HermesAgentBaseEnv implementation to satisfy setup requirements and enable operator-facing CLI guidelines for end-to-end evaluation.