Home GRC Operations Security

    Nous Hermes Architecture

    diagram of nous-hermes agent execution dependency

    This is a diagram I created based on my understanding and experience fine-tuning the Nous Research Hermes Agent.

    Coarsely explained;

    • soul.md: The mild to moderately unsettling necessity to imprint a soul, philosophy, and/or persona into the agent that is first referenced prior to any additional data is processed.
    • config.yaml: File where the breadth of the Nous Hermes agent's functionality is configured.
    • Tools: Are various python programs and low-level code functionality to carry out various tasks by the agent.
    • Memories: Statements made in a markdown file of key situational understanding of the agents action and purpose.
    • Skills: are yaml/markdown files written by Nous Hermes as it is asked to perform newer complex tasks. It references the skills when key words are triggered so that the LLM does not have to reason through a complex prompt from scratch.
    • MCP: The Model Context Protocol servers the LLM can be configured to in order to process external platform information.
soulmd config.yaml Tools Memories Skills MCP

WW · Nice.grc

!DH<3