By default, the system or instructions passed in the Hermes Agent API does not replace the original system prompt entirely, but is superimposed on Hermes Agent's own core system prompt. The official API Server documentation makes this very straightforward: the front-end system prompts are just additional commands, and tools, memories, and skills don't disappear.
What does this mean when docking the front end?
- You can add front-end preferences such as "Concise answers" and "Always annotate Python types" to the Open WebUI.
- But don't expect the front-end system alone to completely erase the boundaries of Hermes Agent's tools and capabilities.
- If you want to really limit the capabilities, you should change Hermes' toolsets, platform configuration, or instance configuration.
Many people mistakenly think that "front-end system = take over the entire assistant personality", so when they see Hermes calling tools, they think there is a bug in the API. In fact, according to the official design, it is originally layering, not replace.
In a word: the system in the API is just adding layers, not changing the base.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
