hermes-atropos-environments

To build RL environments for Atropos training, implement the HermesAgentBaseEnv interface to standardize setup, execution, and evaluation of agent loops. This provides structured reward calculation and tool-enabled evaluation loops.

Wandb logging captures evaluation metrics and training progress for Atropos environments. It enables consistent sandboxed testing and reproducible evaluation across agent loops within the Hermes framework.

The serve, process, and evaluate CLI modes support end-to-end lifecycle testing and benchmarking. They enable creation, review, and debugging of environments to ensure robust lifecycle management and sandboxed evaluation.

Yes, you need the hermes-agent toolkit to set up Atropos RL environments. Installing it and starting with serve mode connects to your Atropos setup for standardized agent loop execution and testing.

Reward functions integrate into a Hermes agent loop through the HermesAgentBaseEnv interface. This structured base handles reward calculation and tool-enabled evaluation loops for reliable RL experimentation.

Yes, you can run sandboxed evaluation with tools in Atropos environments. The framework ensures robust tool integration and rigorous testing to support reproducible evaluation across environments.