chore: import upstream snapshot with attribution
CI / lint (push) Has been cancelled
CI / js-syntax (push) Successful in 10m24s
CI / deps-audit (push) Has been cancelled
CI / test (push) Failing after 24m55s
CI / sast-bandit (push) Failing after 11m13s
CI / trivy (push) Failing after 9m32s
Docker Publish / build-and-push (push) Failing after 34m3s

This commit is contained in:
wehub-resource-sync
2026-07-13 12:32:09 +08:00
commit 8f10353f0c
135 changed files with 37786 additions and 0 deletions
+77
View File
@@ -0,0 +1,77 @@
StemDeck Linux Portable Alpha
=============================
This comes in two variants. Pick one:
- StemDeck-Linux-x64.tar.gz CPU-only (smaller; runs anywhere)
- StemDeck-Linux-x64.NVIDIA.tar.gz NVIDIA/CUDA (larger; much faster on an
NVIDIA GPU, falls back to CPU if no GPU)
Run
---
1. Extract the tarball, e.g.:
tar -xzf StemDeck-Linux-x64.tar.gz
2. Install the runtime prerequisites (see below).
3. Run the launcher:
cd StemDeck-Linux-x64 # or StemDeck-Linux-x64.NVIDIA
./StemDeck
4. Let first-run setup prepare local runtime assets.
Prerequisites
-------------
This portable package bundles its own Python runtime (torch + demucs), and
StemDeck downloads FFmpeg automatically on first launch (or uses a system
`ffmpeg` if one is already on your PATH). The only system libraries you need
are your distro's WebKitGTK + GTK, which the desktop shell links against.
Debian / Ubuntu:
sudo apt update
sudo apt install libwebkit2gtk-4.1-0 libgtk-3-0
Fedora:
sudo dnf install webkit2gtk4.1 gtk3
Arch:
sudo pacman -S webkit2gtk-4.1 gtk3
NVIDIA variant
--------------
To use the GPU you need a working NVIDIA driver on the host such that
`nvidia-smi` runs and reports your GPU.
Check your driver:
nvidia-smi
On first launch, the NVIDIA build detects your GPU and downloads the matching
CUDA-enabled PyTorch (a few GB) into your data directory — so the first run
needs an internet connection and some disk space. You do NOT need a separate
CUDA toolkit install, only the driver.
If no usable GPU is detected, the NVIDIA build still runs and falls back to CPU.
If you do not have an NVIDIA GPU, use the CPU-only tarball instead — it skips
the CUDA download entirely.
Notes
-----
- This is a portable folder, not a system package. No .desktop entry, service,
or package-manager integration is created.
- User data lives under $XDG_DATA_HOME/stemdeck (or ~/.local/share/stemdeck).
- Your stem library is written to ~/Documents/StemDeck/.
- Demucs model weights download from the backend on first use into the data
directory under models/.
Troubleshooting
---------------
- "./StemDeck: error while loading shared libraries" — install the WebKitGTK
and GTK packages listed above.
- A job failing immediately with an FFmpeg error — first-run setup downloads
FFmpeg automatically; check internet access and retry, or install a system
`ffmpeg` so `ffmpeg -version` works in your shell.
- If setup fails, check internet access and retry.
- Inspect logs under the data directory's logs/ folder.
- Deleting the data directory forces first-run setup to recreate runtime state.
+78
View File
@@ -0,0 +1,78 @@
THIRD-PARTY NOTICES
===================
StemDeck includes third-party open-source software. Each component is
copyrighted by its respective authors and distributed under its own license.
This starter notice is not a substitute for the full license inventory that
must be generated from the final packaged Python runtime before a public
release.
Bundled Components
------------------
Python
License: Python Software Foundation License
Website: https://www.python.org/
Tauri
License: MIT or Apache-2.0, depending on component
Website: https://tauri.app/
FastAPI
License: MIT
Website: https://fastapi.tiangolo.com/
Uvicorn
License: BSD
Website: https://www.uvicorn.org/
yt-dlp
License: Unlicense
Website: https://github.com/yt-dlp/yt-dlp
Demucs
License: MIT
Website: https://github.com/facebookresearch/demucs
PyTorch / Torch
License: BSD-style
Website: https://pytorch.org/
torchaudio
License: BSD-style
Website: https://pytorch.org/audio/
librosa
License: ISC
Website: https://librosa.org/
pyloudnorm
License: MIT
Website: https://github.com/csteinmetz1/pyloudnorm
soundfile
License: BSD
Website: https://github.com/bastibe/python-soundfile
System Requirements (Not Bundled)
---------------------------------
FFmpeg is not bundled in the Linux package. On first launch StemDeck downloads a
static FFmpeg build into your user data directory (or uses a system `ffmpeg`
already on your PATH). The download is fetched directly from the upstream
provider, not redistributed by StemDeck. FFmpeg may be distributed under LGPL or
GPL terms depending on how the build was compiled.
WebKitGTK and GTK shared libraries are provided by your Linux distribution and
are not bundled. They are distributed under LGPL terms.
Demucs model weights are not bundled. They are downloaded during first use.
StemDeck should display the model source and license/usage terms before or
during download.
Disclaimer
----------
This notice file is not legal advice. Before public release, verify the exact
licenses of every bundled package and generated binary artifact.