A code repository indexing tool that enhances LLM coding assistance by providing project-specific context to reduce hallucinations.
VectorCode is a code repository indexing tool that enhances the performance of Large Language Models (LLMs) for coding tasks. It works by indexing a codebase and providing relevant context to LLMs, which helps reduce hallucinations and improve the quality of generated code, especially for closed-source or niche projects.
Developers using LLMs for coding assistance, particularly those working on proprietary, less-known, or rapidly evolving projects where LLMs lack inherent knowledge.
It offers a programmatic way to inject precise, project-specific context into LLM prompts, making AI coding tools more accurate and reliable compared to generic LLM responses.
A code repository indexing tool to supercharge your LLM experience.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Uses tree-sitter for syntax-aware chunking, preserving code structure and improving LLM context relevance over naive splitting, as highlighted in the TODOs for smarter chunking.
Provides a Neovim plugin with APIs to enhance AI plugins like codecompanion.nvim, making it seamless for Vim-centric workflows, as noted in the documentation.
Automatically detects project roots using .git or custom .vectorcode/ anchors and respects .gitignore to exclude irrelevant files, ensuring focused indexing.
Supports ChromaDB for caching embeddings, reducing redundant processing and speeding up retrieval, as mentioned in the features for efficient caching.
The project is in beta with rapid iterations and breaking changes in minor versions, making it risky for production use, as warned in the versioning notes.
Primarily integrated with Neovim; users of other editors lack native support and must rely on CLI tools, limiting accessibility for broader teams.
Remote ChromaDB with authentication is a TODO item, forcing users to manage local instances or work around limitations for distributed setups.