Wafer Overlay (智能叠图)

Wafer Overlay by Sichain is a desktop app for intelligent wafer stacking and AOI. It renders accurate wafer maps, stacks multi-stage data, and can run TorchScript AOI (segmentation first, optional YOLO detection).

Development stack

• Tauri: Tauri + React + Typescript
• Mantine: GUI library for React
• ThreeJS: JS-based 3D library for accurate drawing of wafers
• Rust Backend

IDE setup:

VS Code + Tauri + rust-analyzer

Project structure

Branches

Folder structure

Screenshots & demo

Pre-compiled binaries

Building

Developer notes

Creating icons

pnpm tauri icon --help
pnpm tauri icon public/logo3.png

Building

Windows

MacOS

pnpm tauri build --bundles dmg

文件夹与文件正则表达式

文件夹	正则表达式
衬底	(?<=[/\\])衬底(?=[/\\])
FAB CP	(?<=[/\\])FAB\s*CP(?=[/\\])
CP 1	(?<=[/\\])CP\s*1(?=[/\\])
WLBI MAP	(?<=[/\\])WLBI\s*MAP(?=[/\\])
CP 2	(?<=[/\\])CP\s*2(?=[/\\])
AOI	(?<=[/\\])AOI(?=[/\\])

Database (数据库)

The database file is located in the %APPDATA folder, named data.db.

Admin password via env

• Set a default admin password by creating a .env file at the project root:
  • Copy .env.example to .env
  • Set VITE_ADMIN_DEFAULT_PASSWORD=your-secret
• On first run, if the database still has the seed password (admin), the app updates it to the env value during initialization.
• The “default password” check in the UI uses this env value as the baseline.

References (libraries, dependencies, papers, etc.)

• https://www.sichainsemi.com/
• https://v2.tauri.app/start/
• https://react-redux.js.org/
• https://threejs.org/
• https://github.com/0xtaruhi/ufde-next/
• https://github.com/tabler/tabler-icons/

Authors

• JUN WEI WANG | jwwang2003
• YI TING | ee731

PyTorch / libtorch

macOS ARM Tauri bundle (ship libtorch inside the .app)

• Download the official libtorch-macos-arm64 zip from pytorch.org and unpack it into src-tauri/libtorch/ so you have src-tauri/libtorch/lib/libtorch.dylib, libtorch_cpu.dylib, etc.
• src-tauri/tauri.conf.json already copies libtorch/** into the bundle; the dylibs land in AOI Wafer Stacking.app/Contents/Resources/libtorch/lib.
• If you built libtorch locally at 3rdparty/pytorch/build/install, either copy or symlink it into place so the bundler sees it:

  ln -snf ../3rdparty/pytorch/build/install src-tauri/libtorch
  

• Build the macOS ARM bundle with the vendored libs:

  cd src-tauri
  LIBTORCH=./libtorch \  # or ../3rdparty/pytorch/build/install
  RUSTFLAGS="-C link-args=-Wl,-rpath,@executable_path/../Resources/libtorch/lib" \
  cargo tauri build --target aarch64-apple-darwin
  

  (You can swap cargo tauri build with pnpm tauri build; add LIBTORCH_BYPASS_VERSION_CHECK=1 if needed.)
• Verify the binary can see the libs:

  otool -l target/release/bundle/macos/AOI\ Wafer\ Stacking.app/Contents/MacOS/aoi-wafer-stacking | rg LC_RPATH
  ls target/release/bundle/macos/AOI\ Wafer\ Stacking.app/Contents/Resources/libtorch/lib
  

• If codesign complains, run from src-tauri: codesign --force --deep --sign - target/release/bundle/macos/AOI\ Wafer\ Stacking.app

macOS dev using a Python-installed torch

In a shell before running pnpm tauri dev:

export LIBTORCH_BYPASS_VERSION_CHECK=1
export LIBTORCH_USE_PYTORCH=1
torch_lib=$(python - <<'PY'
import torch, os
print(os.path.join(os.path.dirname(torch.__file__), 'lib'))
PY
)
echo "Using torch lib dir: $torch_lib"
ls "$torch_lib/libtorch_cpu.dylib"  # should exist
export LIBTORCH="$torch_lib"
export DYLD_LIBRARY_PATH="$torch_lib:${DYLD_LIBRARY_PATH}"
pnpm tauri dev

Building libtorch from source

Refer to PyTorch's build-from-source guidelines. Tested on an M1 Pro (2023 MacBook 14").

git submodule init
git submodule update --recursive

cd 3rdparty/pytorch
export BUILD_TEST=0
export USE_DISTRIBUTED=0
export USE_CUDA=0
export USE_MPS=0   # set to 1 if you actually need MPS on macOS
export DEBUG=0
python tools/build_libtorch.py

mkdir -p build && cd build

cmake .. \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_INSTALL_PREFIX="$PWD/install" \
  -DBUILD_SHARED_LIBS=ON \
  -DBUILD_PYTHON=OFF \
  -DBUILD_TEST=OFF \
  -DUSE_CUDA=OFF \
  -DUSE_MPS=OFF \
  -DUSE_DISTRIBUTED=OFF

cmake --build . --target install -j"$(sysctl -n hw.ncpu)"

ln -snf ../3rdparty/pytorch/build/install ../src-tauri/libtorch 

Dev without libtorch:

pnpm tauri dev -- --no-default-features