maint: move prebuilt binaries to GitHub Releases

.gitignore

@@ -40,9 +40,10 @@ echoframe/cpp/src/py3.10_wheel_build
 echoframe/cpp/src/py3.11_wheel_build
 echoframe/cpp/src/py3.12_wheel_build
 echoframe/cpp/src/tempbuild
-# Build trees from build_scripts/build_echoframe.bat. Unanchored: the script
-# writes them at the repository root by default, and EF_BUILD_ROOT can move
-# them anywhere.
+# Build trees and packaged zips from build_scripts/build_echoframe.bat.
+# Unanchored build_CUDA_*: the script writes them at the repository root by
+# default, and EF_BUILD_ROOT can move them anywhere.
+/binaries/
 build_CUDA_*/
 echoframe/cpp/src/testing
 documentation/_build/

README.md

@@ -17,7 +17,7 @@ EchoFrame delivers high‑throughput **Fourier‑domain beamforming** and **Powe
 * Real‑time performance (> 10 kFPS on RTX 40‑series)
 * Self‑contained build: CMake ≥ 3.21 (used to build the project) + a recent CUDA toolkit are all you need
 
-**Prebuilt binaries**: 
+**Prebuilt binaries**: available from [GitHub Releases](https://github.com/BrainEchoLab/EchoFrame/releases/latest).
 
 | Versions | CUDA 12.8 | CUDA 12.9 | CUDA 13.1 |
 |----------|-----------|-----------|-----------|
@@ -59,9 +59,9 @@ We offer a range of pre-built binaries for EchoFrame. This means you can simply
 
 **⚠️ IMPORTANT:** If you do not have one of the CUDA and Matlab combinations mentioned in the **Prebuilt binaries** table in [Overview](#overview), then you may run into issues using EchoFrame with them!
 
-You can find these binaries in the `binaries` directory. There you can pick from the available builds whichever matches your CUDA and Matlab setup. Each folder in `binaries` contains a compressed `build` folder.
+Download the matching `build_CUDA_<version>_MATLAB_<version>.zip` from [GitHub Releases](https://github.com/BrainEchoLab/EchoFrame/releases/latest). Each asset contains a compressed `build` folder.
 
-Once you have the EchoFrame repository in your local machine, you can navigate to `EchoFrame\echoframe\cpp\src` and paste the **uncompressed** `build` folder inside `src`. After that, you should be good to go.
+Once you have the EchoFrame repository on your local machine, unzip the asset, navigate to `EchoFrame\echoframe\cpp\src`, and paste the **uncompressed** `build` folder inside `src`. After that, you should be good to go.
 
 You can then explore [how to use EchoFrame](#using-echoframe).
 
@@ -483,8 +483,6 @@ Each folder has its own README with the details.
 
 - **build_scripts/**: developer helper scripts — the Windows build, vcpkg setup, and the build-verification harness. See [build_scripts/README.md](./build_scripts/README.md).
 
-- **binaries/**: prebuilt builds, one zip per CUDA/MATLAB combination. See [Plug and Play](#plug-play).
-
 - **docker/**: the Ubuntu/CUDA development image.
 
 - **justfile**: task runner recipes. Stays at the repository root so `just` can find it.

build_scripts/README.md

@@ -45,7 +45,8 @@ set EF_BUILD_ROOT=C:\efb
 ```
 
 The packaged `binaries\build_<CONFIG>\build.zip` is written to the repository
-either way. Enabling long paths (`LongPathsEnabled` under
+either way. These generated zips are ignored by git; attach them to the matching
+GitHub Release instead. Enabling long paths (`LongPathsEnabled` under
 `HKLM\SYSTEM\CurrentControlSet\Control\FileSystem`, needs a reboot) or using the
 Ninja generator also works -- see the main README.