Hermes Model Presets

Named model, provider and reasoning presets for Hermes messaging.

/model deep
→ /model gpt-5 --provider openai-api --reasoning high --session

The ax alias switching to Astra with xhigh reasoning effort

ax switches both the model to gpt-6-astra and the reasoning effort to xhigh.

Setup

hermes plugins install will-lynas/hermes-model-presets --enable
hermes plugins doctor model-presets --ci

Add presets to your profile's config.yaml, then restart the gateway:

plugins:
  entries:
    model-presets:
      settings:
        aliases:
          deep:
            model: gpt-5
            provider: openai-api
            reasoning_effort: high

Usage

/model deep                   # this session only
/model deep --reasoning low   # override the preset's effort
/model deep --global          # save as the default
/model another-model          # unconfigured names pass through

Names match exactly, including case. reasoning_effort is optional.

Session presets need reapplying after a gateway restart. --once isn't expanded.

Requirements

Hermes 0.22.5+
Python 3.11+
A Hermes messaging gateway; local CLI/TUI, Desktop and API turns aren't supported.

How it works

The pre_gateway_dispatch hook rewrites matching /model aliases into full commands. Hermes then handles the model switch normally. Other commands pass through unchanged.

Development · MIT

model-presets

Screenshots

What it adds

Hooks 1

README