Skip to content

Releases: web3infra-foundation/scorpiofs

ScorpioFS v0.1.1 — The Filesystem Layer Between Monorepo and Agent

07 Mar 02:15
@genedna genedna
Immutable release. Only release title and notes can be modified.
v0.1.1
2ac3178
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
ScorpioFS v0.1.1 — The Filesystem Layer Between Monorepo and Agent Latest
Latest

ScorpioFS v0.1.1 — The Filesystem Layer Between Monorepo and Agent

We're excited to announce the first public release of ScorpioFS, a FUSE-based virtual filesystem that bridges Mega monorepos with local development and AI agent workflows. ScorpioFS lets you mount any subtree of a petabyte-scale monorepo as a local directory — no full checkout required.

What is ScorpioFS?

Monorepos solve the context problem for AI agents, but they introduce a scale problem: no agent (or developer) wants to clone terabytes of code just to work on one module. ScorpioFS eliminates this tradeoff.

Built in Rust on top of FUSE (Filesystem in Userspace), ScorpioFS presents remote monorepo content as a native local filesystem. File reads are fetched on demand from the Mega server; writes are captured locally and can be pushed back. The result: agents and developers get the full-context illusion of a monorepo with the I/O footprint of a sparse checkout.

Think of it as what VFSForGit did for Microsoft's Windows monorepo, or what EdenFS does for Meta's Sapling — but open source, Rust-native, and designed for the Mega ecosystem.

Why ScorpioFS for Agents?

AI coding agents need filesystem access to read, write, and build code. But they also need it to be fast, isolated, and resource-efficient:

  • On-demand fetching — Only the files an agent actually touches are downloaded. An agent working on services/auth/ never pays the cost of pulling frontend/ or ml-models/.
  • Mount/unmount via API — ScorpioFS exposes HTTP endpoints for mounting and unmounting subtrees programmatically. Agents and orchestrators can spin up workspaces dynamically without CLI interaction.
  • Antares overlay — A writable overlay layer on top of the read-only monorepo view, giving each agent an isolated, copy-on-write workspace for builds and experiments.
  • Lightweight footprint — No background Git object downloads, no multi-gigabyte .git directories. Just the files you need, when you need them.

What's in v0.1.1

FUSE Virtual Filesystem

Mount any Mega monorepo path as a local directory. File operations (read, write, stat, readdir) are transparently proxied to the Mega server with local caching.

HTTP Control API

Manage mount points programmatically:

# Mount a subtree
curl -X POST http://localhost:2725/api/fs/mount \
  -H "Content-Type: application/json" \
  -d '{"path": "third-party/mega/scorpio"}'

# List active mount points
curl -X GET http://localhost:2725/api/fs/mpoint

# Unmount
curl -X POST http://localhost:2725/api/fs/unmount \
  -H "Content-Type: application/json" \
  -d '{"path": "third-party/mega/scorpio"}'

Antares Overlay Layer

A writable overlay filesystem on top of the FUSE-mounted monorepo view, supporting isolated workspaces for builds and agent sandboxing.

TOML Configuration

Simple configuration via scorpio.toml and config.toml for mount paths, store paths, and server endpoints.

Published on crates.io

ScorpioFS is available as the scorpiofs crate for integration into Rust-based toolchains and agent runtimes.

Where ScorpioFS Fits in the Ecosystem

ScorpioFS completes the third leg of the Mega infrastructure stack:

Component Role What it does
Mega Server Centralized monorepo engine — stores and serves the entire codebase
Libra Git Client Lightweight, embeddable Git client with SQLite storage for agents
ScorpioFS Filesystem On-demand virtual mount of monorepo subtrees for local access

A typical agent workflow: an orchestrator calls the ScorpioFS API to mount services/payment/ into the agent's workspace, the agent uses Libra to track its changes as Git commits with structured metadata, and pushes back to Mega — all without ever checking out the full repository.

Get Started

# Clone and build
git clone https://github.com/web3infra-foundation/scorpiofs.git
cd scorpiofs
cargo build --release

# Configure
# Edit scorpio.toml to set mount_path and store_path

# Run (requires a Mega server at localhost:8000)
cargo run -- -c scorpio.toml

What's Next

  • Agent sandbox mode — Per-agent isolated mount namespaces with resource limits and filesystem-level access control.
  • Prefetch hints — Allow agents and build systems to declare file access patterns ahead of time, enabling predictive fetching.
  • macOS support — Extend FUSE support beyond Linux via macFUSE for broader development environments.
  • Deeper Mega integration — Push-back workflows from the overlay layer directly to Mega, closing the read-modify-push loop without leaving the filesystem.

Contributing

We welcome contributions. ScorpioFS is a Rust project — clone the repo and get started:

cargo clippy --all-targets --all-features -- -D warnings
cargo +nightly fmt --all --check

License

ScorpioFS follows the licensing of the Mega ecosystem.

ScorpioFS v0.1.1 — Mount your monorepo. Don't clone it.

Assets 3
Loading