Install ollama

Run in the background to download a Mistral Instruct 0.2 based model:

ollama run ifioravanti/mistral-grammar-checker

Note: If you have base model installed in Ollama, only a very small additional layer will be downloaded. You can easily create new custom models using ModelFiles following instructions on Ollama

Install the workflow. This essentially copies the workflow to ~/Library/Services/autogram-ollama.workflow.

Go to System Preferences -> Keyboard -> Shortcuts -> Services -> General -> autogram, and set the shortcut, in my case I set it to ⌃⌥⌘G.

In any OSX application, select some text, and press the shortcut you set. The selected text will be replaced with the generated text from the model.

Install autogram-mlx.workflow the same way as above, no need to edit and add token, however you may edit MODEL="mistral", in case you want to use a different model

This worfklow is based on mlx-community/Mistral-7B-Instruct-v0.2-4bit-mlx model with prompt defined directly in the workflow. You can edit it directly from Automator.

Original idea LLM-Automator Radamés Ajna