VectorCode is a command-line tool and Neovim plugin designed to enhance Large Language Model (LLM) interactions with codebases. It indexes code repositories, allowing users to programmatically inject relevant context into LLM prompts, thereby improving output quality and reducing hallucinations for closed-source, niche, or cutting-edge projects.

How It Works

VectorCode leverages a vector database (ChromaDB) to store embeddings of code chunks. It supports various embedding engines, with SentenceTransformer being the default. The tool offers features like configurable chunking, metadata inclusion, and smarter, syntax-aware chunking using py-tree-sitter. It also respects .gitignore and uses project root anchors for better detection.

Quick Start & Requirements

• Install: pip install vectorcode
• Prerequisites: Python 3.8+, ChromaDB.
• Documentation: CLI Documentation, Neovim Plugin Documentation, Wiki

Highlighted Details

• Syntax-aware chunking via py-tree-sitter.
• .gitignore support for indexing.
• Neovim plugin integration for seamless workflow.
• Configurable chunk size and document selection.

Maintenance & Community

The project acknowledges contributions from @milanglacier and @olimorris. Discussions are available for questions and usage sharing.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README. Users should verify licensing for commercial or closed-source integration.

Limitations & Caveats

The project is in beta quality with basic retrieval and embedding functionalities. Remote ChromaDB support with authentication is not yet implemented, and there's no current method to view or delete individual files within a collection without re-vectorizing the entire project.