Codecomp

Copilot like code suggestions using RAG. Currently only plugins for Lem editor and Vim (with textprop support, > 9.0.0185).

Demo

Lem

Screen.Recording.2024-03-06.at.6.33.44.PM.mov

Vim

Setup

• Clone the repo

$ git clone https://github.com/asutoshpalai/codecomp
$ cd codecomp

• Download the embedding model

$ curl -OL 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5-GGUF/resolve/main/nomic-embed-text-v1.5.Q8_0.gguf?download=true'

• Export the following environment variables or create an .env file contianing the following variables.

export OPENAI_BASE_URL="https://api.openai.com/v1"
export OPENAI_API_KEY="skynet-launch-codes"
export OPENAI_MODEL_NAME="gpt-3.5-turbo-0613"

Note: Strongly using Groq becuase of how fast it is. Groq has an OpenAI compatible API which can be used here.

• Launch the Python server from the same shell. I haven't commited the dependency file yet. Please install anything that it complains about. Some errors might show up while serving the request.

$ python server.py

• Test that the server is working properly.

curl -X POST -v http://localhost:8000/code_complete/invoke --header 'Content-Type: application/json' --data-raw '{
        "input": {
        "repo_path": "/home/user/projects/lem-codecomp/",
        "language": "python",
        "partial_code": "def start_server(host: str, port: int):\n"
  }
}'

Vim

• Install the plugin. You can use vim-plug to install as follows:

Plug 'asutoshpalai/codecomp'

• Add keybindings to invoke code completion

imap <C-a> <esc>:Codecomp<cr>i
imap <C-e> <esc>:CodecompAccept<cr>i
imap <C-s> <esc>:CodecompReject<cr>i

Lem

• Load this in your Lem's init.lisp file. Add the following lines to the init file.

(push "/home/remote/Learning/llm/anyscale/" asdf:*central-registry*)
(ql:quickload :lem-codecomp)