Foul Flock · ALPR Siting Compliance Map

Live at www.foulflock.com. Made by lowlydba.

A small, fully static map that cross-references crowdsourced ALPR camera locations (OpenStreetMap, DeFlock tagging convention) against a hand-curated table of state camera-siting statutes, and flags cameras sitting within a screened distance of a protected place: schools, places of worship, healthcare facilities, courts, food banks.

A few states have decided license-plate cameras don't belong near certain places. Laws like that only matter if someone checks, and checking by hand is slow enough that it rarely happens. This map automates the tedious part.

To be clear about what it is: a screening tool for advocacy and compliance research. Not a router, not a camera-avoidance app, and not legal advice. Every flag carries a statute citation and the law's own placement language, and the UI walks the user toward a verify-first inquiry or records-request letter. Rule provenance (confidence, last_verified) stays in the rules file for maintainers; the UI deliberately doesn't dress heuristic candidates up as verified findings.

What the colors mean

The map deliberately distinguishes:

1. 🔴 Flagged - camera sits within the screened distance of a protected place. Comes with the statute citation; needs human verification.
2. 🔵 Checked, nothing found - a siting law exists in this state, and no mapped protected place turned up within the screened distance.
3. ⚪ Reviewed, no siting law - grey on the map, with a last-checked date on hover. States that haven't been reviewed get no overlay at all, because "we looked and there's nothing" and "we haven't looked" are different claims.

Heuristic distances

When a statute uses vague language ("immediately surrounding") instead of a number, the build applies a configurable "close by" default of 100 m. That number is ours, not the law's, and the UI labels it as such everywhere it appears (buffer_specified: false in the rules file).

Two legend sliders adjust the heuristics live, no rebuild needed (the build pre-computes candidates out to --scan-max, 400 m):

• "Close by" (50-400 m, default 100) for statutes with no distance language at all.
• "On premises" (10-150 m, default 50) for on-property language like VA's "on or within the boundaries of" (heuristic_buffer_meters in the rules file).

Statutory numeric distances are fixed and unaffected by either slider. Change the defaults with --default-buffer or default_heuristic_buffer_m in rules/alpr_state_rules.yaml.

Whenever a heuristic distance is used, the UI and the drafted report letter also quote the statute's actual placement language ("on or immediately surrounding", "on or within the boundaries of", etc.) from the statutory_language field, so the on-property vs. nearby distinction is never hidden behind an invented number.

View cones

Where OSM records a camera:direction bearing, the map draws an illustrative view cone (60 degrees wide, 60 m deep, zoom in to see them). OSM has no field-of-view width/range tags for ALPRs, so cones show mapped bearing only - they are not real detection zones.

Layout

rules/alpr_state_rules.yaml   Hand-curated legal core. Never scraped/automated.
build/build.py                Fetch (cached) -> match -> emit web/data/*.geojson
build/vendor/                 Pinned Overture places taxonomy (CC-BY-4.0)
web/                          Static MapLibre app (no backend, no runtime APIs)
docs/research/                Archived state-law survey outputs
cache/                        Overpass + Overture response cache (safe to delete)
AGENTS.md                     Maintenance runbook (state re-reviews, overrides)

Build & run

pip install pyyaml overturemaps
python build\build.py                        # fetches (cached), then matches
python -m http.server 8080 -d web            # then open http://localhost:8080

Useful flags: --offline (cache only), --refresh [scope] (ignore cache and refetch; scope is all or a comma-separated subset of cameras,places,divisions), --states WA (subset).

A scheduled workflow (.github/workflows/data-refresh.yml) refreshes cameras weekly and Overture places/divisions monthly (matching Overture's monthly release cadence), then opens a PR with the regenerated web/data/*, so the published data stays current without manual runs.

Tests

The Python build pipeline is unit-tested with pytest, and WCAG 2.1 A/AA is asserted with axe-core + Playwright in both light and dark color schemes (see AGENTS.md for the theming rules):

pip install -r tests\python\requirements.txt
python -m pytest tests\python

npm ci
npx playwright install chromium
npm run test:a11y

CI runs both suites on every push/PR (.github/workflows/ci.yml, axe against fixture data in tests/fixtures/data) and rolls them into a single required check. Workflow security is linted by zizmor at its pedantic persona (.github/workflows/zizmor.yml), and dependabot watches actions, npm, and pip weekly.

Data sources

Source	Provides	License
OpenStreetMap via Overpass	ALPR camera nodes (surveillance:type=ALPR, DeFlock convention)	ODbL
Overture Maps places theme	Protected places (schools, worship, healthcare, courts, food banks)	CDLA-Permissive-2.0
Overture Maps divisions theme	State boundary polygons + GERS division IDs	Contains OSM data: ODbL
Overture places taxonomy (vendored, build/vendor/)	Category codes mapped to the five protected categories	CC-BY-4.0
rules/alpr_state_rules.yaml	Hand-curated statutes, buffers, citations	Project

Places and divisions are streamed per state from the Overture S3 release via the official overturemaps package (anonymous reads, no credentials) and filtered locally against the vendored taxonomy; category-to-protected-place mapping lives in category_overture_mapping in the rules YAML. Cameras stay on OSM/Overpass because Overture has no ALPR layer. state_division_id on each rule records the state's Overture GERS division ID.

License

Code is MIT. The data is not: OSM-derived files (cameras, and state boundaries via Overture divisions) remain under ODbL 1.0 © OpenStreetMap contributors, and Overture places remain under CDLA-Permissive-2.0 © Overture Maps Foundation. The build stamps a DATA_LICENSE.txt into web/data/ alongside the artifacts, and the vendored taxonomy in build/vendor/ is CC-BY-4.0. Keep those notices intact when redistributing.

Maintaining the rules file

• Human-authored only. Do not scrape legal text.
• Every entry needs source_url, confidence (verified/inferred/needs_review), and last_verified.
• Re-review quarterly - ALPR siting law is an actively moving target.
• If a statute gives no numeric distance, set buffer_meters: null and buffer_specified: false; never invent a number and present it as law.

Honesty notes

These are load-bearing, not boilerplate:

• Siting-specific ALPR laws are rare; most state ALPR statutes govern retention/access, not placement.
• Camera positions are crowdsourced and can be wrong, stale, or mis-tagged.
• A flag is a candidate for human verification, never a violation determination. Verify camera, place, and current statute text before contacting any agency.
• Camera data © OpenStreetMap contributors (ODbL). Places and boundaries © Overture Maps Foundation (places CDLA-Permissive-2.0; divisions contain OSM data, ODbL).