Skip to content

HIT-ReFreSH/VeryFS

Repository files navigation

VeryFS

GitHub GitHub last commit GitHub repo size GitHub code size

Project Hub | Call for Contributions

VeryFS is an AI-native Virtual File System for connecting applications, automation tools, and AI agents to heterogeneous knowledge and storage sources.

It provides a Unix-style namespace over mounted resources such as local directories, Git repositories, WebDAV endpoints, remote VeryFS instances, and knowledge libraries. The project is designed around a small core, pluggable drivers, language clients, and workspace export paths that are friendly to agentic coding and knowledge workflows.

Why AI-Native

Most filesystems are built for human-operated applications first. VeryFS treats AI agents as first-class users of the filesystem contract:

  • Stable workspace views: external tools can receive a predictable workspace layout instead of bespoke application APIs.
  • Mount-aware provenance: files can retain their source driver, mount path, ETag, version hints, and cache policy so agents can reason about where data came from.
  • Capability-based access: drivers declare supported operations such as list, stat, read, write, manifest, recent, materialize, and versioning before an agent attempts an action.
  • Least-privilege integration: user-scoped mounts, ACLs, signed requests, secret references, and remote mount cache policy are part of the core design.
  • Agent-friendly discovery: manifest and recent queries help agents inspect large workspaces without recursively crawling every mounted source.
  • Data-safety bias: cache, versioning, and retention work is designed around preserving user data first, even when storage usage is higher in early deployments.

Goals

  • Provide a stable VFS API for files, directories, mounts, and links.
  • Keep the core independent from application-specific business logic.
  • Support drivers for common storage and knowledge systems.
  • Make resources available to AI agents and external tools through predictable workspaces.
  • Prefer mature protocols and implementation techniques over custom formats.

Repository Index

The project is split into focused repositories:

Repository Purpose
VeryFS.Core Core library, metadata storage, ACLs, virtual namespace, local driver, and driver abstractions.
VeryFS.Server ASP.NET Core HTTP server and Docker packaging for VeryFS.
VeryFS.Drivers Driver packages for Git providers, WebDAV, Zotero, remote VeryFS, and shared driver utilities.
VeryFS.Clients.CSharp C# SDK for the VeryFS API.
VeryFS.Clients.Python Python SDK for the VeryFS API.
VeryFS.Clients.TypeScript TypeScript SDK for the VeryFS API.
VeryFS Project overview, roadmap, contribution guide, and repository index.

This repository includes the implementation repositories as Git submodules. Clone with --recurse-submodules to get a complete source checkout.

Current Status

VeryFS is pre-1.0 software. APIs and driver contracts may change while the project converges on a stable plugin and client model.

The current implementation focuses on:

  • ASP.NET Core based VeryFS Server.
  • PostgreSQL-backed metadata and secret storage.
  • User-scoped access control and signed request support.
  • Virtual directories and links.
  • Repository, WebDAV, Zotero, remote VeryFS, and local filesystem drivers.
  • C#, Python, and TypeScript client repositories.

License

VeryFS is released under the MIT License.

About

AI-native Virtual File System project hub

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors