Electron SDK — Development

Contributor guide for building and testing @runanywhere/electron from source on Windows.

Prerequisites

Windows x64 (the package is private: true, os: win32 only)
MSVC and the RunAnywhere commons + engines built with the windows-release preset
Node.js dev headers (installed via npm install)

Build steps

From sdk/runanywhere-electron:

bash

npm install
npm run build          # TypeScript -> dist/
npm run bundle:native  # copy the built .node + DLLs into prebuilds/
npm test               # unit tests (hermetic, no native addon needed)

Integration tests need the addon and downloaded models:

powershell

$env:RUNANYWHERE_NATIVE_PATH = '<path>/runanywhere_native.node'
npm run test:integration

See native/CMakeLists.txt and the commons build docs for compiling the native addon against runanywhere-commons.

GPU build (CUDA / NVIDIA)

The default build is CPU-only. To offload inference to an NVIDIA GPU, build commons with -DRAC_GPU_CUDA=ON (CUDA toolkit ≥ 12.4 and MSVC). At load time, llama.cpp auto-offloads layers that fit VRAM.

powershell

# From the repo root. Set CMAKE_CUDA_ARCHITECTURES to your GPU (86 = RTX 30-series).
$env:NVCC_PREPEND_FLAGS = '-allow-unsupported-compiler'   # only if MSVC is newer than the CUDA toolkit officially supports
cmake -S . -B build/windows-cuda -G "Visual Studio 17 2022" -A x64 `
  -T cuda="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6" `
  -DRAC_GPU_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=86 `
  -DRAC_BUILD_ELECTRON_ADDON=ON -DRAC_BUILD_BACKENDS=ON -DRAC_STATIC_PLUGINS=ON -DRAC_BUILD_SHARED=OFF
cmake --build build/windows-cuda --config Release --target runanywhere_native --parallel

Place the built runanywhere_native.node in prebuilds/win32-x64-cuda/ beside its DLLs: cudart64_12.dll, cublas64_12.dll, cublasLt64_12.dll (from the CUDA bin/) plus onnxruntime.dll, onnxruntime_providers_shared.dll, sherpa-onnx-c-api.dll.

Then run npm run bundle:native or copy manually. The demo's examples/electron/RunAnywhereAI/run-demo-gpu.cmd launches against that prebuild.

Example demo

From the repo root:

cmd

examples\electron\RunAnywhereAI\run-demo.cmd

Or manually:

cmd

set RUNANYWHERE_NATIVE_PATH=sdk\runanywhere-electron\prebuilds\win32-x64\runanywhere_native.node
npx electron examples/electron/RunAnywhereAI

When bundling into an Electron app, unpack native artifacts from the asar:

jsonc

// electron-builder config
"asarUnpack": ["**/node_modules/@runanywhere/electron/prebuilds/**"]