vixen

A Rust build engine with correct caching. The CLI is vx.

Getting Started

cargo xtask install

This installs vx and vxd. On Linux, you also need libfuse3 and fusermount.

Then try it on your project:

vx build

Brand

• Primary brand color: #E88602
• Brand guide: BRAND.md

What it does today

Builds single-crate Rust projects with no dependencies:

vx build           # debug build
vx build --release # release build
vx clean           # remove .vx/ directory

Outputs go to .vx/build/ (not target/).

Caching

vixen uses content-addressed storage and picante (incremental queries) to cache builds correctly:

• Second build with unchanged inputs = instant (zero rustc invocations)
• Change source, edition, or profile = rebuild
• Different checkout path = still a cache hit (--remap-path-prefix)
• Cache persists across sessions

Global cache lives at ~/.vx/ (or $VX_HOME).

macOS FSKit host

The FSKit extension host lives in apps/mac as Vixen.app.

• App bundle id: eu.bearcove.vixen
• Extension bundle id: eu.bearcove.vixenfs
• App Group path source for vx-vfs: <mount>/.vixenfs/status.json (app_group_path)

Limitations (v0)

vixen explicitly rejects (with clear errors):

• Workspaces
• Dependencies
• Features
• Build scripts (build.rs)
• Proc macros
• Tests / benches / examples
• Multiple binary targets

Debug logging

RUST_LOG=vx_daemon=debug vx build

Tests

Ordinary Rust tests still run in nextest:

cargo nextest run

If you need to cap per-test memory on Linux, use the wrapper-script profile:

cargo nextest run --profile systemd
export NEXTEST_PROFILE=systemd
export VX_NEXTEST_MEMORY_MAX=2G
cargo nextest run
VX_NEXTEST_MEMORY_MAX=2G cargo nextest run --profile systemd -p vx-eval2 -E 'test(session_) | test(prepared_) | test(stale_)'

Full-stack daemon/VFS scenarios run through xtask instead of nextest:

cargo run -p xtask -- integration list
cargo run -p xtask -- integration run
cargo run -p xtask -- integration run cache-basics
cargo run -p xtask -- integration run json-output --jobs 8
cargo run -p xtask -- integration run lua --no-capture
cargo run -p xtask -- integration run json-output --keep-temp

The xtask integration run suite owns the migrated vx build end-to-end scenarios that need explicit fixture setup, daemon lifecycle, and retained artifacts. It bootstraps one shared daemon session, runs an explicit warmup build before parallel scenarios, supports bounded parallelism with fail-fast or --no-fail-fast, and reports live pass/fail/running counts. On failure, the runner prints merged output plus retained temp/artifact directories so logs can be inspected after teardown. --no-capture switches to serial execution and streams scenario output directly for debugging, and --keep-temp keeps scenario temp dirs even on success. Ctrl-C cancels the run and still tears the shared daemon down.

Co-developed with LLMs

This project was co-developed with LLMs such as Claude, Codex, etc. Human maintainers are responsible for their contributions regardless of machine help.

License

Vixen is licensed under AGPL-3.0-or-later.

For commercial licensing inquiries, contact: amos@bearcove.eu