qai-hub-apps
CLI for browsing and downloading Qualcomm® AI Hub Apps — sample apps for deploying ML models on Qualcomm® devices.
Installation
pip install qai-hub-apps
Quick Start
qai-hub-apps list # browse available apps
qai-hub-apps info whisper_windows_py # inspect an app
qai-hub-apps fetch whisper_windows_py # download source to current directory
Commands
list
List all available apps.
qai-hub-apps list
+----------------------------------------------------------------------+
| Qualcomm® AI Hub Apps |
+---------------------+------------------+---------------+-------------+
| ID | Name | Domain | Languages |
+---------------------+------------------+---------------+-------------+
| whisper_windows_py | Whisper Windows | Audio | Python |
| stable_diffusion_py | Stable Diffusion | Generative AI | Python, C++ |
+---------------------+------------------+---------------+-------------+
Total: N apps
Filters
Different flags are ANDed, values within one flag are ORed, and matching is
case-insensitive (_ and - are treated as spaces).
| Flag | Filters on | Example values |
|---|---|---|
--type |
App type | android, windows, ubuntu |
--language |
Implementation language | Python, C++ (or cpp), Java, Kotlin, Go |
--runtime |
Inference runtime | tflite, onnx, genie |
--domain |
Domain | Audio, Computer Vision, Generative AI |
--use-case |
Use case | Object Detection, Speech Recognition |
--precision |
Model precision | float, w8a8, w4a16 |
--model |
A model the app supports (substring match) | whisper, llama |
qai-hub-apps list --type windows --model whisper
+------------------------------------------------------------------------------------+
| Qualcomm® AI Hub Apps |
+--------------------+-----------------+--------+-----------+---------+--------------+
| ID | Name | Domain | Languages | Type | Models |
+--------------------+-----------------+--------+-----------+---------+--------------+
| whisper_windows_py | Whisper Windows | Audio | Python | windows | whisper_base |
+--------------------+-----------------+--------+-----------+---------+--------------+
Total: 1 of 2 apps
Each flag takes one or more values, and may be repeated:
qai-hub-apps list --type android # only Android apps
qai-hub-apps list --language Python C++ # Python or C++ apps
qai-hub-apps list --type android --language kotlin # Android apps written in Kotlin
qai-hub-apps list --model whisper --runtime onnx
info
Show details for an app.
qai-hub-apps info <app_id>
+=============================================================================+
| Whisper Windows |
| https://github.com/qualcomm/ai-hub-apps/tree/main/apps/whisper_windows_py |
+=============================================================================+
+---------------------------+
| Headline |
+---------------------------+
| Speech to text on Windows |
+---------------------------+
+-----------------------------------+
| Description |
+-----------------------------------+
| Run Whisper on-device using ONNX. |
+-----------------------------------+
+--------------------------------------------+
| Metadata |
+-------------------+------------------------+
| ID | whisper_windows_py |
| Type | windows |
| Languages | Python |
| Runtime | onnx |
| Domain | Audio |
| Use Case | Speech Recognition |
| Precision | float |
| Models | whisper_base |
| Supported Devices | Snapdragon X Elite CRD |
| AI Hub Models | 0.53.1 |
+-------------------+------------------------+
fetch
Download and extract an app's source to a local directory.
# Download app source only
qai-hub-apps fetch <app_id>
# Download app source + a model to download for a specific chipset
qai-hub-apps fetch <app_id> --model <model_id> --chipset <chipset>
# Download app source + bundle a locally-exported model
qai-hub-apps fetch <app_id> --model <path/to/model>
| Flag | Description |
|---|---|
--output-dir PATH |
Output directory (default: current directory) |
--model MODEL_ID_OR_PATH |
Model to bundle: a model ID to download (must be supported by the app), or a path to a locally-exported model (directory or .zip). Use --model-id/--model-path to be explicit |
--model-id MODEL_ID |
Model ID to download (must be supported by the app) |
--model-path PATH |
Path to a locally-exported model (directory or .zip) |
--chipset CHIPSET |
Target chipset for the model download. Only applies to a downloaded model (--model <id> or --model-id); not valid with a local model path |
--device DEVICE |
Target device for the model download. Only applies to a downloaded model (--model <id> or --model-id); not valid with a local model path |
--model, --model-id, and --model-path are mutually exclusive.
On success, the path to the fetched app directory is printed.
Example — fetch app with model:
qai-hub-apps fetch stable_diffusion_windows_py --model stable_diffusion_v2_1 --chipset qualcomm-snapdragon-x-elite
# or
qai-hub-apps fetch stable_diffusion_windows_py --model stable_diffusion_v2_1 --device "Snapdragon X Elite CRD"
Logging
Control verbosity with a global flag or the QAI_HUB_APPS_LOG_LEVEL environment
variable. The command-line flag takes precedence over the environment variable.
The default level is info.
| Flag | Description |
|---|---|
--log-level {debug,info,error} |
Set log verbosity |
-v, --verbose |
Shorthand for --log-level debug |
-q, --quiet |
Shorthand for --log-level error |
qai-hub-apps -v list
QAI_HUB_APPS_LOG_LEVEL=debug qai-hub-apps list
Metadata
Release files for qai-hub-apps 0.38.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| qai_hub_apps-0.38.0-py3-none-any.whl | Python 3 | none | any | Details |
Release files / qai_hub_apps-0.38.0-py3-none-any.whl
| Download URL | qai_hub_apps-0.38.0-py3-none-any.whl |
|---|---|
| Size | 92.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
a12c1973bc63a7298c979b4471f2beeaa5925f0f723af051a70b9de39fcf0518
|
|
BLAKE2b-256 checksum How to use checksums |
b1c056423a281cf06f51c33716197e3d1ed5b68be38d72222d66a86e6a5782eb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.10.20
|