llmnvim

A neovim plugin built on top of llm

For more on llm: https://simonwillison.net/tags/llm/

Features

Chat

Start a long-running conversation.

Ask

Ask a one-off prompt and supply context via fragments

Models

Manage available models

Logs

Explore logged prompts and responses

Installation

1. Install python (if applicable): https://docs.astral.sh/uv/getting-started/installation/

# MacOS
`curl -LsSf https://astral.sh/uv/install.sh | sh`

# Windows
`powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"`

2. Install llm: https://llm.datasette.io/en/stable/index.html#quick-start

uv tool install llm

3. Install llm-mlx plugin for MLX model support (if applicable): https://github.com/simonw/llm-mlx

llm install llm-mlx

4. Add llmnvim plugin to neovim:

return {
  "llmnvim",
  dir = "~/.config/nvim/lua/llmnvim",
  config = function()
    require("llmnvim").setup({
      no_stream = false,
      system_prompt = "You are a helpful assistant.",
      template = nil,
      float_opts = { border = "rounded", width = 80, height = 20, title = "LLM Chat" },
      key = nil, -- Set your API key here, e.g., "sk-..."
      continue_conversation = false,
      conversation_id = nil,
      params = {}, -- e.g., { var1 = "value1" }
      options = {}, -- e.g., { temperature = "0.7" }
    })
  end,
}

Reference

Commands

• LLMAsk: Opens the LLM ask window, allowing you to interact with models and generate text.
• LLMChat: Opens the LLM chat window, enabling you to engage in conversations with models.
• LLMLogs: Opens the LLM log window, providing a way to analyze and review model logs.
• LLMSelectModel: Opens the LLM model selector window, allowing you to select the best model for your task.

Keybindings

• <leader>aa: Opens the LLM ask window.
• <leader>ac: Opens the LLM chat window.
• <leader>al: Opens the LLM log window.
• <leader>am: Opens the LLM model selector window.
• <leader>lf: Opens the LLM fragment window.

TODO

Essentials

[x] chat: interactive prompt (right-side sidebar0 [x] models: list/select models (floating buffer) [x] ask: one-shot question on files (open, glob, search, etc.) (floating buffer

• file
• files
• keybindings [x] logs: view logs

V1.1

• upgraded ask:

  • scratch
  • git - staged,unstaged,untracked
  • git - commit
  • git - branch vs master
  • selected cursor
  • no file, just regular prompt
  • all files in current directory
  • all files in root directory
  • search for files using snacks/telescope

• keybindings / keybindings ui

  • ask keybindings
  • ask keybindings ui
  • chat keybindings
  • chat keybindings ui
  • models keybindings ui

Planned upgrades

Expanded LLM CLI Usage

• prompt Execute a prompt
• chat Hold an ongoing chat with a model.
• logs Tools for exploring logged prompts and responses
• models Manage available models
• collections View and manage collections of embeddings
• embed Embed text and store or return the result
• embed-models Manage available embedding models
• embed-multi Store embeddings for multiple strings at once in the...
• fragments Manage fragments that are stored in the database
• similar Return top N similar IDs from a collection using cosine...
• templates Manage stored prompt templates
• schema Manage I/O format
• tool Manage tool calls

LLM Plugins

• Embeddings via https://github.com/simonw/llm-embed-jina

Cursor-like Features

• Expand ask to entire codebase - use embeddings?
• Inline code writing / suggestions