VectorVFS, your filesystem as a vector database

What is VectorVFS, your filesystem as a vector database?

VectorVFS is a lightweight Python package that transforms a Linux filesystem into a vector database by storing vector embeddings directly as extended attributes on files. Rather than requiring a separate database or indexing system, VectorVFS integrates semantic search capabilities directly into the existing directory structure, allowing users to query files by meaning rather than by filename or metadata alone. The package leverages Meta's Perception Encoders for embedding generation, supporting image and video analysis with vision-language understanding capabilities. It operates on both CPU and GPU, though GPU acceleration is recommended for large image collections during initial embedding. VectorVFS currently supports Perception Encoders and image data types, with plans to expand model support and add additional data types. The tool is designed for developers and organizations seeking to add semantic search functionality to their file systems without infrastructure overhead. It is particularly suited for teams working with large image collections, visual content management, or applications requiring similarity-based file discovery. As an open-source project in its initial release phase, VectorVFS offers a practical alternative to traditional vector databases for filesystem-based workflows, with ongoing development to broaden its capabilities.