# Choose your path

Choose the path that matches the project you already have. Keep the runtime source or release bundle version consistent with the integration instructions you follow.

## Zephyr

For applications built with west and Kconfig.

Select a backend and integrate source or prebuilt libraries with your Zephyr application.

[View setup guide](/helia-rt/getting-started/zephyr/)

## neuralSPOT-X

For applications built with Ambiq’s development SDK.

Add the nsx source module, configure its features and link the runtime into the SDK’s CMake build.

[View setup guide](/helia-rt/getting-started/neuralspot-x/)

## CMSIS-Pack

For projects managed with CMSIS tooling.

Generate, validate and install a local source pack. Pack validation does not establish support for every IDE or board application.

[View setup guide](/helia-rt/getting-started/cmsis-pack/)

## Build from source

For control over compiler options and kernel selection.

Build Make archives or CMake source targets when you need to configure features or debug the runtime.

[View setup guide](/helia-rt/getting-started/source/)

## Prebuilt archive

For firmware that links a released runtime library.

Choose an archive with matching architecture, toolchain and features, then link it with compatible headers and ABI settings.

[View setup guide](/helia-rt/getting-started/cmake/)

## One source of truth

CMake, Zephyr, neuralSPOT-X source mode and the CMSIS-Pack generator consume the same [source manifest and backend-selection rules](https://github.com/AmbiqAI/helia-rt/blob/main/cmake/README.md). This single source of truth (SSoT) keeps runtime file lists and kernel selection consistent across those integrations.

Make-built release archives use a separate build path. Match an archive's toolchain, architecture and feature settings to your application; shared source selection does not make build configurations interchangeable.

## Runtime or ahead-of-time compilation

Use heliaRT for a LiteRT interpreter integration. For supported models, [heliaAOT](https://ambiqai.github.io/helia-aot/) is the recommended option when targeting latency, power and memory efficiency. Check its compatibility and deployment requirements before choosing the compiler path.