Does Hermes Agent not work after receiving vLLM?

If the model only outputs the tool call as text after receiving the vLLM, it is usually not that Hermes is not connected, but that the tool call parameters are missing when the vLLM service is started. Hermes usestool_choice: autoby default, and the vLLM side also needs to be explicitly turned on.

Check the launch command first

depends on whether these items
_CODE_INLINE_4__can be accessed by Hermes.
_CODE_INLINE_5__Whether the memory is exceeded, whether the service has failed to start.
Does the model itself support tool calls? Unsupported models will be unstable.
One sentence: vLLM tool does not work, first add--enable-auto-tool-choice and correct--tool-call-parser.
How to determine whether it is a vLLM or a model problem
Use the same model to directly request/v1/chat/completions, bring the simplest tool schema, and see if there are structured tool calls in the return. If the API layer has returned only ordinary text, the problem is most likely vLLM parser or model capabilities; if the API layer is normal but Hermes is abnormal, check Hermes 'custom endpoint, model name, context length, and log.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/_.