PLASMA

PLASMA: Platform for LSL-based Acquisition of Sensor Metrics and Analytics

📖 Operator handbook: https://yuyichang.github.io/PLASMA/ — install, configure, pair MSense wristbands, run a session, read the memo panel, recover from faults.

quickstart

• conda create -n plasma python=3.12
• conda activate plasma
• pip install -e ".[all]" — or a lean subset, e.g. ".[msense]" / ".[qb2,pupil]"
• python -m plasma (or the plasma console script)

pip install -r requirements.txt still works — it's a shim for -e .[all,build,test].

Want a double-clickable icon instead of running plasma from a terminal? pip install -e ".[desktop]" then plasma-install-shortcut — see Install & first launch.

Writable state (device config, gyro-bias calibration, data/ recordings, session log) location, in order: $PLASMA_HOME if set → a per-user app-data dir when running as a packaged app (~/Library/Application Support/PLASMA on macOS, %LOCALAPPDATA%\PLASMA on Windows, ~/.local/share/plasma on Linux) → the working directory when running from source. Resolved once at import — set PLASMA_HOME before launching to override.

use PLASMA from another project

pip install "plasma-app[msense] @ git+https://github.com/YuyiChang/PLASMA@v2.0.0"

The distribution is plasma-app; the import package is plasma. Extras map to plugins: msense, qb2, pupil, shimmer, obs (and all).

Known issue / future works

• need manually set lidar ip addr
• the built-in LSL→XDF recorder captures every stream on the default liblsl session, so a concurrent pytest run or another lab tool leaks into the recording; re-Initialize doesn't fully tear down old outlets. Direction (SessionID scoping, outlet lifecycle): docs/reference/lsl-lifecycle-and-scoping.md
• interactive=false on ses/sub when collection started
• linux aarch64 build — PLASMA_Linux_arm64 for Jetson Orin (JetPack 6); CI job build-linux-arm64
• infrequent stream (e.g., PLASMA journaler) shows green after recording stop
• doc for msense workflow with better screenshots
• memo stage transition logic optimization (e.g., timing of "ready to start" msg wrt sensor init status)

developing a sensor plugin

Every sensor is a plugin: a PlasmaDevice subclass plus a registration entry in plasma/plugins.py. Core code never imports a concrete device.

full guide: plasma/devices/README.md base class: PlasmaDevice in plasma/devices/template.py reference package (config section + tabs + multi-unit): plasma/devices/msense/

documentation

The playbook is a MkDocs Material site under docs/, deployed to GitHub Pages by .github/workflows/docs.yml on every push to main.

pip install -e ".[docs]"        # mkdocs-material + mkdocs-glightbox + playwright
mkdocs serve                     # preview at http://127.0.0.1:8000

Screenshots (docs/assets/screenshots/*.png, Git LFS) are regenerated by a Playwright script that drives PLASMA with the simulated MSense device:

brew install git-lfs && git lfs install     # one-time
python scripts/capture_screenshots.py        # ~4 min; needs Google Chrome

It uses PLASMA_HOME=~/Library/Application Support/PLASMA by default so the paths shown in the memo panel match a packaged install; pass --home /tmp/x for a disposable run.

references

• qb2: https://docs.blickfeld.com/qb2/Qb2/v1.10/guides/api.html
• pupil-labs: https://pupil-labs.github.io/pl-realtime-api/dev/