⚡️ prompt-ai - configure vim with natural language

prompt-ai is a super lightweight Neovim plugin that lets you configure your editor using natural language prompts like:

• :P use a light theme and turn on line numbers
• :P optimize for viewing large files
• :P replace word "foo" with "bar"
• :P use the kanagawa-lotus colorscheme
• :P format file
• Or really anything you can think of.

---

✨ Features

• Easily apply prompt config via natural language (:P)
• Caches responses to avoid redundant API calls
• Remembers previous conversations (limited to recent history)
• Saves history and cache to disk across sessions
• Configurable model, temperature, and history length
• Commands to clear history and cache

Tip

This plugin focuses solely on one thing: on-the-fly Vim prompt generation. If you're looking for AI-powered file editing or manipulation, check out vim-ai. prompt-ai and vim-ai compliment each other well.

---

📦 Requirements

• Neovim 0.8+
• OpenAI API key in your environment:

export OPENAI_API_KEY="sk-..."

🚀 Installation (Lazy.nvim)

{
  "psaia/prompt-ai.vim",
  config = function()
    require("prompt_ai").setup({
      model = "o4-mini",    -- or "o4-mini" (default)
      temperature = 1  ,    -- controls creativity (default: 1)
      max_history = 8,      -- max prompt/response pairs remembered (default: 8)
    })
  end,
  cmd = { "P", "PClearHistory", "PClearCache" },
}

---

Usage

Configure with Natural Language

:P use gruvbox with transparent background

Clear History

:PClearHistory

Clear Cache

:PClearCache

---

Where Data is Stored

File	Purpose
~/.local/share/nvim/ai_config_history.json	Prompt/response history
~/.local/share/nvim/ai_config_cache.json	Prompt → code cache
/tmp/nvim_ai_config.log	Log file (debug)

---

Privacy & Safety

No telemetry or analytics. Your prompts and configs are never stored or sent anywhere beyond OpenAI's API.

---

Ideas for the Future

• Intelligently contextualize pre-installed plugins for more relevant prompts
• Different models other than OpenAI
• Telescope integration for history
• Floating preview of generated code

---

🖤 Created by @psaia