Show HN: VexFS – A Kernel-Native Semantic Filesystem for AI Agents
May 26, 2025 - news.ycombinator.com
VexFS is a newly developed kernel-native filesystem designed to store vector embeddings alongside files, enabling semantic search capabilities. Currently, it supports brute-force search with plans to implement HNSW later. The system provides a minimal interface through IOCTL and mmap, aiming to serve as a semantic memory layer for local AI agents without relying on a vector database. The concept envisions vector search as an operating system primitive, offering a native mountable memory layer for AI applications.
The project is in its early stages and is described as barely functional but operational. The creator seeks feedback on whether this innovative approach could enhance AI systems by providing an OS-level semantic context layer, potentially leading to more powerful and efficient AI applications. The project is open for contributions and discussions on its GitHub page.
Key takeaways:
VexFS is a kernel-native filesystem designed to store vector embeddings alongside files and support semantic search.
The filesystem aims to serve as a semantic memory layer for local AI agents without relying on a vector database.
VexFS is in its early stages, with basic functionality, but it has the potential to enhance AI systems by providing an OS-level semantic context layer.
The project is open to feedback and contributions, with its source code available on GitHub.
