Repolode
Предложить
Ephemeral-AI-Lab

Ephemeral-AI-Lab/layerfs

A versioned filesystem for AI agents that gives each agent a writable, forkable copy of a project with deduplicated, durable history.

Открыть на GitHubСайт
Звёзды
851
Форки
39
Открытые issues
37
Последний коммит
58 минут назад

Что это?

LayerFS gives each AI agent its own complete, writable copy of a project without duplicating files. Branches are cheap names for existing states, so an agent can fork from main, from another branch, or from any past commit, and forks can be forked again to any depth. Each fork works independently and merges back when ready.

Storage relies on content-addressed storage, content-defined chunking, copy-on-write directory trees and delta encoding, so a new state costs roughly the size of what changed. Commits merge with the current branch head, and a real conflict writes markers into the agent's workspace instead of overwriting. The README reports a measured case where a project's full commit history of 770 MB fit in a 6.3 MB database.

Compute and state are separate: file operations run in the caller's process against a shared store, while disposable Sandboxes (Docker containers) mount workspaces and run commands. The project is a Rust core with a JavaScript SDK, targets Linux with ext4, xfs or btrfs, and is still pre-1.0 with the 1.0 release scheduled for October 21, 2026.

Ключевые возможности

  • 01Forks a branch from main, another branch, or any past commit without copying files.
  • 02Deduplicates identical files and chunks across all branches and agents.
  • 03Merges commits with the branch head and reports conflicts without overwriting work.
  • 04Traces each line of a file to the commit that wrote it via blame.
  • 05Separates disposable Docker sandboxes from the durable shared store.
  • 06Provides a JavaScript SDK with file, branch, project and sandbox operations.

Сценарии использования

  • Running several AI agents in parallel on the same codebase, each in its own isolated workspace.
  • Trying alternative solutions in recursive agent workflows and merging only the winning branches.
  • Running environment or dependency experiments without rebuilding or copying the whole project.
  • Recording agent tool-call history as rewindable checkpoints.

Темы

#ai-agents#version-control#filesystem#deduplication#copy-on-write#sandbox#rust#sdk

Частые вопросы

Что такое layerfs?
A versioned filesystem for AI agents that gives each agent a writable, forkable copy of a project with deduplicated, durable history. LayerFS gives each AI agent its own complete, writable copy of a project without duplicating files. Branches are cheap names for existing states, so an agent can fork from main, from another branch, or from any past commit, and forks can be forked again to any depth. Each fork works independently and merges back when ready.
layerfs можно использовать бесплатно? Какая у него лицензия?
layerfs распространяется с открытым исходным кодом под лицензией MIT. Перед использованием в коммерческих продуктах проверьте условия лицензии.
layerfs всё ещё поддерживается?
Активно поддерживается. Последний коммит в layerfs — 19 окт. 2026 г..
Для чего можно использовать layerfs?
Running several AI agents in parallel on the same codebase, each in its own isolated workspace. Trying alternative solutions in recursive agent workflows and merging only the winning branches. Running environment or dependency experiments without rebuilding or copying the whole project. Recording agent tool-call history as rewindable checkpoints.
Где скачать layerfs?
Исходный код и инструкции по установке — на GitHub: https://github.com/Ephemeral-AI-Lab/layerfs

Описание создано ИИ на основе README проекта. Статистика GitHub обновляется ежедневно.

Похожие проекты