Paragraph Rephrasing Vim Plugin

A Vim plugin to rephrase the selected paragraph inlinely.

Demo

ollma-vim-plugin.mov

Feature

1. Streamline daily task, rephrasing paragraph is my daily task especially writing commit message.
2. Modern, leveraging the modern LLM deployment framework (ollama)
3. Security, leveraing the local LLM model.

Prerequisites

1. Ollama server

For macOS user, install via:

$ brew install ollama

For Linux user, install via:

$ curl -fsSL https://ollama.com/install.sh | sh

Reference

2. jq command for JSON processing, install via:

For macOS user, install via:

$ brew install jq

For Linux user, I think it shall be installed by default.

Install the plugin

I use vim-plug as the vim plugin manager. This is feasible if you use the same manager as well. Open your ~/.vimrc, Add the below to your Plugin manager.

call plug#begin()

...
Plug 'ChinYikMing/inline-message-rephraser'
...

call plug#end()

Reload the file or restart Vim, then you can, run :PlugInstall to install the plugin.

Run your local ollama server

$ ollama serve

Create message rephraser model

$ ollama create msg-rephraser -f Modelfile

Note: Modelfile can be used to set the system prompt.

Sample Modelfile for git commit message rephrasing:

FROM llama3.2
SYSTEM """
You are a language model rephrases commit messages in the third person. Each line should be limited to 72 characters; if this limit is exceeded, continue the text on a new line. Format the output as a single paragraph.

Example:
Instruct: "Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum is simply dummy"

Response: "Lorem Ipsum is simply dummy text of the printing and typesetting\n
industry. Lorem Ipsum is simply dummy"
"""

Fire the plugin

1. Open your vim
2. Place your cursor at the point where you want to rephrase.
3. Run :R to rephrase

TODO:

1. Use an excellent Modelfile to create a more powerful message rephraser.
2. More text selection granularity for rephrasing. E.g., line by line.
3. Add processing status indicator inside vim when inferencing LLM.

Reference

• ollama RESTful API