Hermes + Local LLM

Setup LM-Studio 0.4.24-2

Setup Hermes TUI

Install Hermes through Omarchy Menu (SUPER+SPACE, Hermes, Enter)

Run setup = Y




Select ‘Full Setup’




Select LM-Studio




Hit enter on LM_API_KEY to accept default




Hit enter on Base URL to accept default




Select default model as ‘qwen3-8b-128k’




Select terminal backup ‘keep current local’




Select ‘Telegram’ from the platforms




Select ‘Automatic’ for the Telegram bot




Open up the link on my mobile to link to telegram

Get the API token in telegram

Go to BotFather in Telegram

Type ‘/mybots’

Select the hermes one

Select API Key

Copy the API manually back into hermes terminal




Leave the Home channel ID empty for cron job results, etc




Setup systemd telegram = ‘Y’




Run gateway as a user service




Start the service now

Select the Tools, I added ‘Home Assistant’ to the defaults




Choose a provider = Local Browser




Went with defaults for Home assistant, just hit enter

Choose a provider (LLM) = skip




Select a TTS = Microsoft Edge




Choose DuckDuckGo for search provider




Will have to setup tools later as some are not connected

Enable browser

Install agent-browser
$ npm install -g agent-browser

Test

Start ‘hermes’ TUI

Create a test file
Prompt
Create a file /home/map7/foo.txt with bar in it

Then do a ls /home/map7 in terminal to check

Speed it up!

Using qwen/qwen3.5-9b from lmstudio-community model do the following through lm-studio

set context to 65000

turn off reasoning

turn off K & V Quantization

Install hermes-desktop and turn off reasoning in there, then try.

t