p33m5t3r / vimiq

a dead simple neovim llm extension

vimiq

A Neovim plugin for interacting with language models

This is a work-in progress, and may have bugs. But I didn't love the existing options, so i wrote an llm extension for myself. it aims to be extremely simple and easy to modify. you can see the demo here:

setup

1. clone/move the repo to .config/nvim/lua/vimiq (or wherever you nvim plugin lua lives)
2. if you want to use anthropic or openai, then:

   pip install anthropic openai
   

3. also set your api keys in your shell:

   export ANTHROPIC_API_KEY="your_anthropic_api_key"
   export OPENAI_API_KEY="your_openai_api_key"
   

4. in your init.lua

    require('vimiq').setup({
        python_path = "/path/to/your/python/installation/or/virtualenv"
    })

5. on a per-project basis, you can create a .vimiq.json file to put context into your chats

   {
       "model": ["default"],
       "system": "You are a helpful assistant.",
       "includes": ["./backend.py", "./init.lua"]
   }

6. you can edit backend.py to include your own llm implemenations:

   LLM_IMPLS = {
       "default": llm_anthropic,
       "anthropic": llm_anthropic,
       "openai": llm_openai,
       "claude": llm_anthropic,        # example alias
       "chat": llm_openai              # example alias
   }

usage

commands:

• :InitBuffer: start new llm buffer
• :InitBufferWithLSPError: start buffer with current line and lsp errors
• :EvalBuffer: send buffer to llm
• :StopStreaming: stop current stream

structure

• init.lua: handles neovim stuff (buffers, commands, data passing)
• backend.py: llm logic (message formatting, api calls, response streaming)

separated like this so you can modify llm logic without touching neovim code.

customization

• edit backend.py to change llm providers, message formatting, etc.
• update init.lua if you need new commands or data passing changes
• use .vimiq.json in project root for custom prompts or context files

TODO

• add custom filetype syntax
• add locallama support