Install, build and run from source

Download: the latest release. On Windows take SignalViewer-Setup-X.Y.Z.exe (installs for you only, no administrator rights); to run without installing, the …-Windows-x64-portable.zip (extract, double-click Signal Viewer.bat). Linux and macOS: the -portable.7z for your system.

With Python (3.9 or newer): pip install signalviewer, then signalviewer starts the app - with every reader and report (pip install "signalviewer[desktop]" for its own window instead of the browser). The light install: pip install signalviewer-lite (about 350 MB instead of 760) is the core - CSV, MATLAB v5, ArduPilot .bin, CAN, binary layouts, every plot and analysis, HTML reports - and the rest are extras: [parquet] (also a faster CSV reader), [hdf5] (.h5, MATLAB v7.3), [mavlink] (.tlog, live MAVLink), [px4] (.ulg), [ros] (.mcap), [reports] (PDF, Word, images), [desktop], [all]; a file that needs one says which. Both install the same signalviewer Python package: install one of them, not both. signalviewer-cli and import signalviewer read and export recordings from scripts. rec.signal("ATT.Roll") finds ATT.Roll [deg] (the unit may be left out), and signalviewer.show(t, {"alt": alt}) sends arrays from a notebook into the open app.

First start: Windows and macOS warnings

The packages are not code-signed yet, so the first start may be stopped once:

• Windows (Windows protected your PC, SmartScreen): More info → Run anyway. Shown for the installer and for SignalViewer.exe from the portable package, once per version.
• macOS ("SignalViewer" cannot be opened or is damaged): macOS marks files that came from the internet. Remove the mark from the extracted folder once, in Terminal: xattr -dr com.apple.quarantine /path/to/SignalViewer, then start SignalViewer in it (or right-click it → Open → Open). The macOS package is for Apple silicon (M1 and later).

Building the packages

Nothing needs the internet at run time. The installer, the portable folder and the single file contain Python, every library, the UI with its fonts, the report assets and the icon. Verified by running the built app with no network interface at all.

Build command (Windows)	Result	When
build.bat installer	dist\SignalViewer-Setup-1.13.0.exe	Recommended. One file; installs for the current user (no admin), Start-menu and desktop icons, Open with for data files, uninstaller
build.bat portable	dist\SignalViewer-portable.zip	Unzip anywhere and run SignalViewer.exe (fast start)
build.bat onefile	dist\SignalViewer.exe	A single file to copy; starts a few seconds slower (it unpacks itself each time)
build_windows_portable.bat (Windows) / python tools/make_windows_portable.py	dist\SignalViewer-Windows-portable.7z	Windows portable build on a Windows PC: double-click the .bat (needs Python 3.8+ with pip and internet, once, only to build). Downloads the official embeddable Python + the Windows builds of the libraries; the result needs no Python and no internet. 7z (about 80 MB instead of 126 MB as zip): made with 7-Zip when installed, else with the py7zr library fetched once; Windows 11 opens .7z in Explorer, Windows 10 needs 7-Zip. --format zip for a zip, --keep-folder keeps the assembled folder, --clean forces a fresh download, --build-ui rebuilds the web UI first (Node.js; done automatically when the UI sources are newer than the built UI)
tools/make_windows_portable.sh (Linux/macOS)	dist/SignalViewer-Windows-portable.7z	Builds the same Windows package without a Windows PC (runs the same Python builder; options are passed on)
build.bat wheels	wheels\	On a connected PC: downloads the libraries once, so the build itself can run on an offline PC

The installer needs Inno Setup 6 on the BUILD PC only (free).

Windows window engine. The app window uses Microsoft Edge WebView2 (built into Windows 11 and updated Windows 10). If it is missing (e.g. an old offline Windows 10 LTSC), the app opens in the local browser instead — still with no internet — and stops by itself when no tab has been open for 30 s.

Ready-made packages (no Python and no installation needed, except the Windows installer itself) are built by the GitHub Actions workflow in .github/workflows/build.yml for Windows (x64: the installer, a portable .zip and .7z), Linux (x64) and macOS (Apple silicon): run it manually (Actions → build → Run workflow) and download them from the run's artifacts, or push a tag such as v1.0.0 to also get a draft Release with every file and a "which file?" note. Windows: the embeddable-Python package described above; Linux / macOS: the PyInstaller folder (SignalViewer.spec) - extract it and run SignalViewer/SignalViewer.

Build it yourself: build.bat (Windows) or ./build.sh (Linux / macOS) → dist/SignalViewer[.exe].

Run from source

pip install -r requirements.txt
python run.py              # native window (pywebview), or the browser if pywebview is missing
python run.py --browser    # always use the browser
python run.py flight.bin   # open files (or a .svsession) - in the app that is already running, if there is one

The UI is already built (frontend/dist). When the UI code is newer than that build (after a pull, a patch or an edit), run.py builds it once by itself - Node.js needed - into frontend/.local-dist (ignored by git; the tracked frontend/dist is not touched) and serves that; the next start sees it is current and starts at once. python run.py --build forces a build, --no-build (or SV_NO_BUILD=1) never builds. Without Node.js it says so and shows the last build. For working on the UI:

python run.py --port 8765 --no-open   # backend
cd frontend
npm install
npm run dev      # dev server with hot reload, proxies /api to port 8765
npm run build    # rebuilds frontend/dist

Tests:

python -m pytest backend/tests          # ~790 server tests (incl. damaged copies of every format)
pip install -r requirements-dev.txt && python -m playwright install chromium
python -m pytest tests/e2e              # ~195 browser tests, the whole suite in one run (~10-15 min)
python -m pytest tests/e2e/test_features.py    # or one file: one test per catalog feature
cd backend && mypy --config-file ../pyproject.toml app svcore signalviewer   # type check
python tools/benchmark.py big.csv       # performance against the targets (results in docs/)