Awesome

🦙 nvim-llama

Ollama interfaces for Neovim: get up and running with large language models locally in Neovim.

https://github.com/jpmcb/nvim-llama/assets/23109390/3e9e7248-dcf4-4349-8ee2-fd87ac3838ca

🏗️ 👷 Warning! Under active development!! 👷 🚧

Requirements

Docker is required to use nvim-llama.

And that's it! All models and clients run from within Docker to provide chat interfaces and functionality. This is an agnostic approach that works for MacOS, Linux, and Windows.

Installation

Use your favorite package manager to install the plugin:

Packer

use 'jpmcb/nvim-llama'

lazy.nvim

{
    'jpmcb/nvim-llama'
}

vim-plug

Plug 'jpmcb/nvim-llama'

Setup & configuration

In your init.vim, setup the plugin:

require('nvim-llama').setup {}

You can provide the following optional configuration table to the setup function:

local defaults = {
    -- See plugin debugging logs
    debug = false,

    -- The model for ollama to use. This model will be automatically downloaded.
    model = llama2,
}

Model library

Ollama supports an incredible number of open-source models available on ollama.ai/library

Check out their docs to learn more: https://github.com/jmorganca/ollama

When setting the model setting, the specified model will be automatically downloaded:

Model	Parameters	Size	Model setting
Neural Chat	7B	4.1GB	`model = neural-chat`
Starling	7B	4.1GB	`model = starling-lm`
Mistral	7B	4.1GB	`model = mistral`
Llama 2	7B	3.8GB	`model = llama2`
Code Llama	7B	3.8GB	`model = codellama`
Llama 2 Uncensored	7B	3.8GB	`model = llama2-uncensored`
Llama 2 13B	13B	7.3GB	`model = llama2:13b`
Llama 2 70B	70B	39GB	`model = llama2:70b`
Orca Mini	3B	1.9GB	`model = orca-mini`
Vicuna	7B	3.8GB	`model = vicuna`

Note: You should have at least 8 GB of RAM to run the 3B models, 16 GB to run the 7B models, and 32 GB to run the 13B models. 70B parameter models require upwards of 64 GB of ram (if not more).

Usage

The :Llama autocommand opens a Terminal window where you can start chatting with your LLM.

To exit Terminal mode, which by default locks the focus to the terminal buffer, use the bindings Ctrl-\ Ctrl-n