davidyz_vectorcode

davidyz_vectorcode

by Davidyz
A tool that indexes code repositories to enhance LLM prompt quality by providing contextual information.

VectorCode: Code Repository Indexing for LLMs

Overview

VectorCode is a code repository indexing tool designed to enhance the experience of working with Large Language Models (LLMs) for coding tasks. It indexes and provides information about the code repository you're working on, enabling better prompt generation and context-aware completions. This tool also includes a Neovim plugin, making it highly accessible for developers using this popular editor.

Features

  • Repository Indexing: Indexes code repositories to provide context for LLMs.
  • Neovim Integration: Comes with a Neovim plugin for seamless integration.
  • RAG Context: Retrieval-Augmented Generation (RAG) context improves LLM completions.
  • Chunking Support: Supports chunking with configurable chunk sizes and metadata.
  • Persistent Client: Utilizes Chromadb for persistent storage of indexed data.
  • Project Root Detection: Automatically detects project roots using .git or custom .vectorcode.json files.
  • Respects .gitignore: Ignores files and directories specified in .gitignore.

Why VectorCode?

LLMs often struggle with understanding close-source projects, less-known projects, or cutting-edge developments. VectorCode addresses this by providing project-specific context, significantly improving the quality of LLM completions.

Example

Without VectorCode, LLM completions for a sudoku-solving project were random guesses:
Sudoku without RAG

With VectorCode, completions were accurate and relevant:
Sudoku with RAG

Documentation

TODOs

  • [x] Query by excluded paths
  • [x] Chunking support
  • [x] Persistent client
  • [x] Project root detection
  • [ ] View and delete files in a collection
  • [x] Joint search with codecompanion.nvim/MCP

Credit

About

VectorCode is a tool to supercharge your LLM experience by providing better context and completions for your coding projects.

Topics

Resources

License

MIT License

Activity

  • Stars: 193
  • Watchers: 5
  • Forks: 7

Contributors

Languages

  • Python: 75.5%
  • Lua: 24.4%
  • Makefile: 0.1%

Features & Capabilities

Categories
mcp_server model_context_protocol python lua neovim-plugin rag retrieval-augmented chromadb embedding code_indexing

Implementation Details

View on GitHub

Stats

0 Views
199 GitHub Stars

Repository Info

Davidyz Organization

Similar MCP Servers

continuedev_continue by continuedev
25049
cherryhq_cherry_studio by CherryHQ
21423
bytedance_ui_tars_desktop by bytedance
9300
2025© BestAI