vim + semantic search

runs a local vec db and embedding model and watches configured directories of your fs to reindex on change and provides an api for semantic search + telescope frontend

• chromadb

  python -m venv venv
  source ./venv/bin/activate
  pip install -r ./requirements.txt
  

• lm studio

  • download an embedding model, eg text-embedding-nomic-embed-text-v1.5-embedding (./emb.ts)
  • run headless https://lmstudio.ai/docs/app/api/headless

• node project

  pnpm install
  

with lazy.vim

{
    "vem",
    dir = "/users/devidw/desktop/code/vem",
    dependencies = {
        'nvim-telescope/telescope.nvim',
    },
    config = function()
        require("telescope").load_extension("vem")

        vim.keymap.set("n", "<leader>v", function()
            require("telescope").extensions.vem.search({
                repo_path = "/Users/devidw/Desktop/code/vem",
                config_path = "/Users/devidw/Desktop/code/vem/config.json",
            })
        end)
    end
},

• hit control + space to search
  • (why not enter? bc once results are there, enter is already reserved for opening them)

for configuration see ./config.ts

pnpm dev ./path/to/my/config.json

• starts the vec db
• maybe loads the emb model
• reindex all collections
• start fs watchers to re-embed

macos launch agent setup:

# first update vem.plist to point to your daemon script, then:
cp vem.plist ~/Library/LaunchAgents/gdn.wolf.vem.plist
launchctl load ~/Library/LaunchAgents/gdn.wolf.vem.plist

• run chromadb
• run some embedding model
• watch folder and index into db
• api to query
• telescope plugin