Hermes Agent first popped up content like 💻 pwd and 🔍 Python docs before replying in the Open WebUI, not garbled characters, nor did it string logs on the front-end, but it was the progress prompt of the official streaming tool. Both the API Server and Open WebUI documentation say that when the tool is executed, it inserts short progress information into the stream so that the front-end knows what the agent is doing.
What's the use of this
It allows you to tell if the model is in a daze or is running a tool. This is especially useful for complex tasks, such as reading files, searching web pages, and running commands, so you know that Hermes Agent is not stuck.
Do you need to turn it off?
Generally not. It just adds a layer of visible progress when streamed back and doesn't change the final answer. If you are looking for a particularly clean streaming interface, consider how to deal with the front-end display layer.
In a word: those 💻🔍 are not exceptions, but Hermes Agent deliberately shows you the progress of the tool.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
