VectorCode
VectorCode is a code repository indexing tool enabling semantic search across your local projects. This extension integrates VectorCode with CodeCompanion, allowing LLMs to query your indexed repositories for enhanced context during chat sessions.
Features
- Provides the
@vectorcodetool for use in the chat buffer. - Enables LLMs to perform semantic searches across multiple indexed local code repositories.
- Supplies relevant code snippets from your projects as context to the LLM.
Prerequisites
NOTE
VectorCode requires initial setup outside of CodeCompanion. You must install the Python backend and index your project files using the VectorCode CLI.
Please refer to the VectorCode CLI documentation for detailed setup instructions.
Installation
First, install the VectorCode Neovim plugin. A minimal installation is sufficient if you only plan to use VectorCode via CodeCompanion:
{
"Davidyz/VectorCode",
version = "*", -- optional, depending on whether you're on nightly or release
build = "pipx upgrade vectorcode", -- optional but recommended. This keeps your CLI up-to-date.
dependencies = { "nvim-lua/plenary.nvim" },
}Next, register VectorCode as an extension in your CodeCompanion configuration:
require("codecompanion").setup({
extensions = {
vectorcode = {
opts = {
add_tool = true,
}
}
}
})With add_tool = true, the @vectorcode tool becomes available in the CodeCompanion chat buffer. For further configuration options, see the VectorCode wiki.
Usage
To grant the LLM access to your indexed codebases, simply mention the @vectorcode tool in the chat buffer. The LLM can then query any projects indexed by VectorCode (verifiable via vectorcode ls in your terminal) to retrieve relevant context for your prompts.
Example: Using VectorCode to Explore the VectorCode Repository
