Skip to main content

QAI AppBuilder

Quick AI Application Builder (QAI AppBuilder) helps developers easily use the Qualcomm® AI Runtime SDK to deploy AI models and design AI applications on Qualcomm® SoC platforms equipped with Qualcomm® Hexagon™ Processor (NPU). It encapsulates model deployment APIs into a set of simplified interfaces for loading models to the NPU and performing inference. QAI AppBuilder significantly reduces the complexity of model deployment for developers and provides multiple demos as references for designing their own AI applications.

Supports Python API / C++ API

QAI-APPBUILDER Architecture

Supported Devices

DeviceSoC
Dragon Q6AQCS6490
Fogwise® AIRbox Q900QCS9075

Installation Methods​

tip

You can install via wheel package or build from source

Clone the Repository​

Device
git clone -b radxa-dev https://github.com/ZIFENG278/ai-engine-direct-helper.git && cd ai-engine-direct-helper

Install via Wheel Package​

Create a Virtual Environment​
tip

Create a virtual environment to isolate the Python libraries

Device
sudo apt install python3-venv
python3 -m venv .venv
source .venv/bin/activate
Device
pip3 install https://github.com/ZIFENG278/ai-engine-direct-helper/releases/download/radxa-dev-2.38.0/qai_appbuilder-2.38.0-cp312-cp312-linux_aarch64.whl

Build from Source​

Device
git submodule update --init --recursive

Download QAIRT SDK​

tip

Please download and extract the QAIRT SDK according to Download QAIRT SDK Package

Configure QAIRT Environment Variables​
Device
export PRODUCT_SOC=9075 DSP_ARCH=73
Device
cd qairt/2.42.0.251225
source bin/envsetup.sh
export ADSP_LIBRARY_PATH=$QNN_SDK_ROOT/lib/hexagon-v${DSP_ARCH}/unsigned
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$QNN_SDK_ROOT/lib/aarch64-oe-linux-gcc11.2

Compile qai_appbuilder​

Create a Virtual Environment​
tip

Create a virtual environment to isolate the Python libraries

Device
sudo apt install python3-venv python3-dev cmake gcc g++
python3 -m venv .venv
source .venv/bin/activate
Compile qai_appbuilder Wheel​
Device
pip3 install setuptools
cd ai-engine-direct-helper
python3 setup.py bdist_wheel

After compilation, a qai_appbuilder-2.38.0-cp312-cp312-linux_aarch64.whl file will be generated in the dist directory

Install qai_appbuilder​

Install qai_appbuilder Wheel​
Device
cd dist
pip3 install ./qai_appbuilder-2.38.0-cp312-cp312-linux_aarch64.whl

Create a qai_libs symlink in ai-engine-direct-helper/samples/python to link to the lib directory of the QAIRT SDK

tip

If you haven't downloaded the complete QAIRT_SDK, you can install qcom-qairt and then create the symlink.

qcom-qairt is a slimmed-down version of QAIRT-SDK, installing only the required files

Device
sudo apt install qcom-qnn-sdk-v73 qcom-genie-sdk-v73
Device
ln -s /usr/lib/aarch64-linux-gnu samples/python/qai_libs

Create ADSP Environment Variables​

tip

If you installed the qcom-qairt package, use the following command to create the ADSP_LIBRARY_PATH environment variable

Device
export ADSP_LIBRARY_PATH=/usr/lib/aarch64-linux-gnu

Application Examples​

tip

For QCS9075 Demos, please refer to Demo Examples

Device
cd ai-engine-direct-helper/samples/python

  • Image Generation / Inpainting

    ModelRun Command
    stable_diffusion_v2_1python3 stable_diffusion_v2_1/stable_diffusion_v2_1.py --prompt "the prompt string ..."
    stable_diffusion_v1_5python3 stable_diffusion_v1_5/stable_diffusion_v1_5.py --prompt "the prompt string ..."
    lama_dilatedpython3 lama_dilated/lama_dilated.py
    aotganpython3 aotgan/aotgan.py

  • Image Super-Resolution

    ModelRun Command
    real_esrgan_x4pluspython3 real_esrgan_x4plus/real_esrgan_x4plus.py
    real_esrgan_general_x4v3python3 real_esrgan_general_x4v3/real_esrgan_general_x4v3.py
    quicksrnetmediumpython3 quicksrnetmedium/quicksrnetmedium.py

  • Image Recognition

    ModelRun Command
    inception_v3python3 inception_v3/inception_v3.py
    beitpython3 beit/beit.py
    googlenetpython3 googlenet/googlenet.py

  • Pose Estimation

    ModelRun Command
    openposepython3 openpose/openpose.py
    mediapipe_handpython3 mediapipe_hand/mediapipe_hand.py

  • Image Segmentation / Depth Estimation

    ModelRun Command
    unet_segmentationpython3 unet_segmentation/unet_segmentation.py
    depth_anythingpython3 depth_anything/depth_anything.py

  • Object Detection

    ModelRun Command
    yolov8_detpython3 yolov8_det/yolov8_det.py

  • 3D Modeling

    ModelRun Command
    facemap_3dmmpython3 facemap_3dmm/facemap_3dmm.py
    face_attrib_netpython3 face_attrib_net/face_attrib_net.py

  • Audio Models

    ModelRun Command
    yamnetpython3 yamnet/yamnet.py
    whisper_base_enpython3 whisper_base_en/whisper_base_en.py
    whisper_tiny_enpython3 whisper_tiny_en/whisper_tiny_en.py

  • Text / Multimodal Embeddings

    ModelRun Command
    nomic_embed_textpython3 nomic_embed_text/nomic_embed_text.py
    openai_clippython3 openai_clip/openai_clip.py --text "mountain"

  • 3D Action Recognition

    ModelRun Command
    resnet_3dpython3 resnet_3d/resnet_3d.py

  • OCR (Text Recognition)

    ModelRun Command
    easy_ocrpython3 easy_ocr/easy_ocr.py

    You need to be logged into GitHub to post a comment. If you are already logged in, please ignore this message.

    Radxa-docs © 2026 by Radxa Computer (Shenzhen) Co.,Ltd. is licensed under CC BY 4.0