LLM Agent for Neovim (Work In Progress - MVP)

A context-aware AI assistant for Neovim.

⚠️ Note: This plugin is under active development as a Minimum Viable Product (MVP) focusing on local Ollama interaction. It is primarily intended for local development and testing. Many features described historically may not be present in the current local development version.

Features (Current MVP)

• Basic chat interface within Neovim (right-side vertical split).
• Asynchronous communication with Ollama using curl via Neovim's jobstart().
• Toggle keymap to show/hide/focus the chat window.

Requirements

• Neovim 0.7.0+
• curl command-line tool installed.
• Ollama running locally (e.g., ollama serve or ollama run <model>)

Installation (Local Development Setup)

This plugin is currently developed and used locally.

1. Clone the repository: Clone this repository directly into your Neovim configuration's Lua path. For a standard setup, this would be:

   git clone <repository_url> ~/.config/nvim/lua/llm_agent_new

   (Replace <repository_url> with the actual URL of your development repository. If you are already working within the cloned repo located at ~/.config/nvim, you can skip this step.)

2. Add to Packer: Use the local path in your packer.nvim configuration (init.lua or equivalent):

   use {
     '~/.config/nvim/lua/llm_agent_new', -- Path to the locally cloned plugin
     as = 'llm_agent_new',
     config = function()
       require("llm_agent_new").setup()
     end,
   }

3. Sync Packer: Run :PackerSync in Neovim.

Usage (MVP)

• Press <Leader>ac (default: Space + a + c) to toggle the chat window:
  • If closed, it opens and focuses.
  • If open and focused, it hides.
  • If open and unfocused, it focuses.
• In the chat window, type your message at the > prompt and press Enter.
• The user message and the assistant's response will appear in the buffer.
• Press <Esc> while in insert mode in the chat window to enter Normal mode.
• Use standard Neovim window commands (e.g., Ctrl-W h/l/j/k) to navigate away from the chat window when in Normal mode.

Configuration (MVP Options)

The plugin can be configured via the setup() function. Only the following options are currently used by the MVP:

require("llm_agent_new").setup({
  -- Debug mode (enables detailed logging in :messages)
  debug = false, 

  -- UI Configuration
  ui = {
    width = 80, -- Width of the chat split window
  },

  -- API Configuration
  api = {
    -- Only 'ollama' provider is currently implemented
    provider = "ollama", 
    ollama = {
      enabled = true, -- Must be true to use Ollama
      host = "localhost:11434", -- Ollama API endpoint
      model = "gemma3:4b"      -- Default Ollama model to use
    }
    -- Other settings like openrouter are ignored in the MVP
  }
})

Ensure your Ollama server is running before use.

License

MIT

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.