If the Hermes Agent custom endpoint cannot be connected, check three things first: whether the base URL has a /v1, whether the API key is non-empty, and whether the model name is consistent with the server. The official Providers documentation states that Hermes can connect to any OpenAI-compatible endpoint, but only if the server actually implements /v1/chat/completions.
Shortest configuration example
OPENAI_BASE_URL=http://localhost:8000/v1
OPENAI_API_KEY=local-key
LLM_MODEL=your-model-name
Common pits
- The local service is only open inside the container and cannot be accessed by the Hermes host.
- base URL is missing
/v1. - The local service does not require a key, but the Hermes/SDK side still requires a non-null placeholder key.
- The model name is the display name, not the server registration name.
In a word: if the custom endpoint doesn't work, first use curl to open /v1/chat/completions, and then let Hermes connect.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
