Interactive dynamic instrumentation app built on Frida. All business logic lives in LumaCore, a cross-platform Swift package; the current shipping frontend is a macOS SwiftUI app, with a GTK/Adwaita frontend for Linux on the way.

Sources/LumaCore/   # cross-platform Swift package — engine, sessions,
                    # persistence, disassembly, collaboration, hook
                    # packs, GitHub auth, address annotations, …
Agent/              # TypeScript agent injected into the target process
Luma/               # macOS SwiftUI frontend
Luma.xcodeproj/     # Xcode project (Luma app + LumaBundleCompiler)
Package.swift       # SPM manifest for LumaCore

• macOS ≥ 15.0
• Xcode ≥ 26 (with the Metal toolchain installed — open Xcode once and accept the Metal SDK download prompt, or install it via Settings → Components)

LumaCore itself only needs Swift 6 and the package dependencies listed in Package.swift. It builds on Linux too:

swift build --target LumaCore

1. Open the project:

   open Luma.xcodeproj

2. Ensure the build destination is set to My Mac (Luma currently uses AppKit-only components and does not yet build for iOS).

3. Choose Product → Build (⌘B).

This performs an incremental build and is the most convenient workflow during development.

A Makefile is provided for building Luma without opening Xcode. This build is also incremental, because it uses a persistent derived-data directory.

The output app is produced in ./build/, and intermediate build files are stored in ./build/.derived.

To build:

make

To clean:

make clean

The resulting app will be located at:

build/Luma.app

Install the toolchains and -devel packages for Luma and its native dependencies:

sudo dnf install -y \
    gcc-c++ libstdc++-static patch golang-bin nodejs swift-lang \
    libadwaita-devel atk-devel webkitgtk6.0-devel \
    libgee-devel json-glib-devel libsoup3-devel \
    libunwind-devel libdwarf-devel libnice-devel \
    ngtcp2-crypto-ossl-devel libbpf-devel capstone-devel \
    lzfse-devel

Build and install frida-core into /usr/local (Fedora's libbpf is too old, so force a subproject fallback):

cd ~/src
git clone [email protected]:frida/frida-core.git
cd frida-core
./configure --enable-shared --without-prebuilds=sdk \
    --enable-barebone-backend --enable-compiler-backend \
    -- --force-fallback-for=libbpf
make
sudo make install

Build and install radare2 into /usr/local. The stock sys/install.sh builds without optimization, so override CFLAGS and strip unused code with --gc-sections:

cd ~/src
git clone [email protected]:radareorg/radare2.git
cd radare2
CFLAGS="-O2 -g -ffunction-sections -fdata-sections" \
    LDFLAGS="-Wl,--gc-sections" \
    ./sys/install.sh --install

From LumaGtk/:

make           # incremental build → .build/debug/LumaGtk
make run       # build + launch
make install PREFIX=/usr/local

Requires Swift for Windows, Visual Studio 2022 (for cl.exe), vcpkg with gtk4, and prebuilt frida-core / radare2 prefixes. Launch a Developer PowerShell for VS and run from LumaGtk/:

.\scripts\windows\build.ps1                            # debug
.\scripts\windows\build.ps1 -Configuration release
.\scripts\windows\package-msi.ps1 -Version 0.1.0       # build\Luma-*.msi
.\scripts\windows\run.ps1                              # launch with DLL PATH set

Prefix locations default to C:\vcpkg\installed\x64-windows-release and C:\src\dist; override with -VcpkgPrefix, -FridaPrefix, -R2Prefix (or $env:VCPKG_PREFIX etc.).