This repo hosts MicroPython-based deployment enablements for running embedded AI models on PSOC™ Edge boards. It currently supports DEEPCRAFT™ Voice Assistant models.
A single Python script that takes a DEEPCRAFT™ Voice Assistant model and gets it running on a PSOC™ Edge KIT_PSE84_AI board — firmware is built and flashed in one command.
From the repository root, switch to the tools/ folder where the deploy script and its config template live:
cd toolsSet your LLVM toolchain path in deepcraft-voice-assistant-model-deploy.ini:
[tools]
llvm_dir = C:/llvm/LLVM-ET-Arm-19.1.5-Windows-x86_64This avoids the interactive LLVM setup prompt. OpenOCD is downloaded automatically if not found.
python deepcraft-voice-assistant-model-deploy.py \
--config-file deepcraft-voice-assistant-model-deploy.ini \
all path\to\your_modelall clones the firmware repo, builds with your model, and flashes to the board.
The model can be a folder or a .zip exported from the DEEPCRAFT™ cloud tool.
Other commands:
| Command | Description |
|---|---|
build <MODEL> |
Build only (also clones/updates the repo) |
flash |
Flash the last built .hex |
clean |
Remove build artifacts for a clean rebuild |
You don't need a full checkout of this repo — just download the script itself:
curl -s -L -o deepcraft-voice-assistant-model-deploy.py https://raw.githubusercontent.com/Infineon/micropython-psoc-edge-ai-model/main/tools/deepcraft-voice-assistant-model-deploy.pyLeave repo_dir blank in the config file (the default). The tool then manages its
own private copy of the firmware — cloning it under va-mpy/ next to the script
on first run and updating it on later runs — so your model builds against a
known-good checkout regardless of what else is on disk.
Once build or all has finished, flash the board (all already does this for you):
python deepcraft-voice-assistant-model-deploy.py all <model-name>Clone the full repo instead of just the script, so you have cm55_firmware/ and
tools/ to edit:
git clone --recurse-submodules https://github.com/Infineon/micropython-psoc-edge-ai-model.git
cd micropython-psoc-edge-ai-model/toolsThen point the tool at this checkout instead of letting it manage a separate copy, so your local changes are picked up immediately without pushing or re-cloning:
[project]
repo_dir = ..Or, without touching the config file:
python deepcraft-voice-assistant-model-deploy.py \
--repo-dir .. \
--config-file deepcraft-voice-assistant-model-deploy.ini \
build test_gpio_control--repo-dir skips the clone/update step entirely and builds cm55_firmware/ as it
sits in your working tree, so re-running build after an edit picks it up right away.
Choose one of the following workflows.
Use this workflow when Git is installed on the host. Clone the firmware and open its root directory before running Docker.
git clone --recurse-submodules https://github.com/Infineon/micropython-psoc-edge-ai-model.git
cd micropython-psoc-edge-ai-modelFrom this directory, which contains cm55_firmware/, run the following
commands. Docker mounts the checkout at /workspace; -w selects the
project's build directory inside it.
WSL with docker installed:
docker run --rm \
-v "$PWD":/workspace \
-w /workspace/cm55_firmware \
ifxmakers/psoc-embedded-ai-toolchain:deepcraft-0.1.0 \
make CONFIG=Debug FRAMEWORK=deepcraftmkdir firmware-build
docker run --rm \
-v "$PWD/firmware-build":/workspace \
--entrypoint sh \
ifxmakers/psoc-embedded-ai-toolchain:deepcraft-0.1.0 \
-c 'git clone --recurse-submodules https://github.com/Infineon/micropython-psoc-edge-ai-model.git . &&
cd cm55_firmware &&
make CONFIG=Debug FRAMEWORK=deepcraft'The checkout and build outputs remain in firmware-build after the container
exits.
The image includes Infineon's OpenOCD fork, which supports the KitProg3 interface and PSE84 target configs.
Flashing requires USB access to the board's debug probe, which build commands do not need. On Linux, pass the USB bus through explicitly:
docker run --rm \
--device=/dev/bus/usb \
-v "$PWD":/workspace \
-w /workspace/cm55_firmware \
ifxmakers/psoc-embedded-ai-toolchain:deepcraft-0.1.0 \
make CONFIG=Debug FRAMEWORK=deepcraft flashOn WSL, attach the debug probe to the WSL distribution first using usbipd-win before running the command above. Direct USB passthrough to Windows Docker Desktop containers is not generally supported.
| Guide | Contents |
|---|---|
| Quick Start | CM33 MicroPython over IPC — requirements, steps, example code |
| Deployment Tool | Installation, commands, options, config file |
| Requirements | Hardware, toolchain, and software prerequisites |
| MicroPython Interface | CM33-side MicroPython API |