Olive's Bark Logger — PWA (zero-hardware variant)

A browser version of the monitor for when you don't have a Raspberry Pi. It uses the Web Audio API to measure sound levels and logs events to IndexedDB. Audio is processed in memory and never recorded, stored, or uploaded — the same guarantee as the Python service. It works fully offline after first load.

This is a parallel implementation of the same idea; it shares the detection semantics and the honest methodology/limitations framing with the Python core:

Concern	Python (monitor/, report/)	PWA (pwa/)
Level math	level.py	level.js
Detector (threshold/min-dur/debounce)	detector.py	detector.js
Aggregation + report + CSV	aggregate.py, render.py, export.py	report.js
Storage (events only, no audio)	SQLite	IndexedDB
Cover block + no-verdict line	render.py (cover_html, cover_text_lines)	report.js (coverHtml, coverTextLines, csvPreamble)

The caveats travel with every file this page produces. The report HTML and both CSV downloads carry the same "what this can and cannot prove" cover block as the Python exports — in the CSVs as a leading # comment preamble, so the caveat travels with a file that gets emailed on while the data rows below it still parse. Anything reporting a quiet-hours count also carries the no-verdict line: being within quiet hours is not the same as a violation, and only the relevant authority can decide whether a rule was broken. The quiet-hours CSV names its recorded monitoring gaps, since a gap removes events and would otherwise make an unmonitored stretch read as a quiet one. This edition has no calibration step, so its report says plainly that its readings are relative dBFS and never dB(A).

The exact strings live in spec/report/cover.json and are replayed against both implementations, the way spec/detector/*.json keeps the two detectors from drifting — see spec/SEMANTICS.md.

Run it

It must be served over http(s):// (microphone access and service workers don't work on file://). Any static server works:

cd pwa
python3 -m http.server 8000
# open http://localhost:8000

Click Start monitoring, grant microphone permission, and adjust the threshold and the quiet-hours window while watching the live level. Use Download report (HTML, with a day×hour calendar heatmap and a quiet-hours summary), Download CSV (the event log), or Download quiet-hours CSV (every event flagged within/outside quiet hours — an honest export for a neighbor/landlord/HOA submission). Clear events resets. Install it as an app from your browser's "Install" option.

Test

node --test pwa/*.test.mjs

Covers the detector port and the aggregation/report/CSV logic, including the export caveat gate: every export path this module has must carry every string in spec/report/cover.json, and tests/test_export_caveats.py discovers the module's export paths from source so a new one fails the gate until it is exercised here too. The report must still contain the methodology + limitations + no-audio statements.