Skip to content

Python tool for aircraft Doppler audio geolocation using historical ADS-B flight data (Source-available)

10.5 KB
Raw Blame

Public website refresh, 2026-11-07

Follow-up after PR #437

The owner requested merging the first pass and continuing; PR #437 was merged at 771897f. The follow-up adds actual method plots with scroll and keyboard stage navigation, compact model settings, an aligned page grid, and a screenshot of the real empty upload interface. No audio was submitted to capture that UI. Figure qualifications are preserved in visible prose and in the enlarged view.

  • Completed homepage Taste build: GATE: PASS, zero waivers. The earlier 35 findings are historical. A subsequent invocation returned GATE: OFF because the successful build had already closed; it was not restarted.
  • All three pages: 18 layouts from 320 to 1440 pixels and 111 recorded interaction observations passed with zero unexpected browser errors.
  • Six axe WCAG A/AA scans passed. All three method stages work by keyboard at 1024x768 and 1440x900; reduced motion restores the static layout and all links.
  • Local performance check: zero failures; warnings for the original large Livermore figure and the upload screenshot. A smaller screenshot variant was added afterward. Field Web Vitals and geolocation accuracy were not measured.
  • Remaining nonblocking Taste warnings concern the instructional pinned moment, visual coverage/spacing, original plot palettes, contextual action labels, image resolution and the live pitch readout. Scientific plots keep their data colors; pitch text is a working measurement readout. Responsive sources serve the method plots and upload screenshot.
  • Impeccable's remaining warnings concern the retained Geist family and padding detection: rendered method rows have no left border, and the footer has 30px of top padding. No guide/vendor source was modified. Its MIT notice is retained.
  • Deployment verification is omitted at the owner's explicit request.

Local follow-up screenshots and checks are in /home/cinna/dopplergeist-site-refinement-preview/. The rest of this document records the first pass at 742b89d, before these follow-up changes.

Scope: the three static GitHub Pages pages and their shared stylesheet/script. Base: a1307b621b28b5c677fa88d5793bb4d8f43726f1. Scientific code, evidence, experiment results and acceptance criteria are outside this change.

Direction and sources

The incumbent serif/dark/wave identity is retained, with a split introduction and interactive model, a charcoal/amber palette, clearer controls and evidence hierarchy, and a method list instead of three feature cards. Reasons and motion dials are recorded in DESIGN.md.

User-requested guides were read from these source revisions:

No new images, brand marks, client claims, testimonials, accounts or dependencies were added to the website. Fonts remain local. Existing scientific captions and qualifications are retained.

Rendered and interaction verification

Chromium 1243 was run locally against an HTTP-served checkout. The system browser adapter supplied --no-sandbox because this host does not support the browser's default namespace sandbox. No skill source was modified.

  • All three pages: 320x800, 390x844, 768x1024, 820x1180, 1024x768 and 1440x900. Eighteen layouts passed root-overflow and content-boundary checks. Command examples scroll inside their own focusable region.
  • 103 interaction observations passed at mobile and desktop widths, with zero unexpected browser console errors or JavaScript exceptions.
  • Every figure on every page opens its actual original, loads, closes with both Escape and the Close button, and returns focus to its link.
  • Both simulations: play, pause, replay, timeline scrubbing and keyboard changes to speed/distance update the curve and named readouts.
  • Every local navigation, section link and page transition was clicked. All nine distinct external URLs returned HTTP 200; external anchors were not verified as destinations within the GitHub page.
  • The physics disclosure opens with the pointer and closes by keyboard.
  • Delayed original image: visible loading status followed by the actual figure. Blocked original image: recovery text, original-image link and working close. The deliberately blocked request's console error is recorded separately.
  • Reduced motion: the simulation starts still and accepts explicit playback; the entrance and hover transitions are disabled.
  • Normal motion: automatic playback advances, stops advancing offscreen and resumes when the model returns. Switching the preference to reduced motion stops playback and removes the entrance animation.
  • With JavaScript disabled: headings, explanatory SVG, content and original figure links remain; unavailable enhancement controls stay hidden.
  • Axe-core WCAG 2 A/AA and 2.1 AA: zero detected violations across all three pages at 390 and 1440 pixels. The scan found and prompted the keyboard fix for the mobile command region. Automated scans do not certify accessibility.
  • node --check docs/site.js and git diff --check passed.
  • Taste's local performance run (three runs per size, 1440 and 390): zero FAILs and one WARN for the existing 457.9 KiB full-resolution Livermore WebP. Total measured transfer was 589.3 KiB on desktop and 210.9 KiB on mobile; the mobile figure variant was 90.4 KiB. No broken requests, measured layout shift or long tasks were recorded in that local run. These are lab observations, not field performance claims. Original scientific plot assets were retained.

Local screenshots, per-interaction receipts, accessibility results and link responses are in /home/cinna/dopplergeist-site-refresh-preview/. These are website checks, not validation of geolocation accuracy. Scientific tests were not rerun for this static-site change. Field Web Vitals were not measured.

Anti-slop delivery gate

Each line records a scoped check of the changed interface.

  • R-02 PASS: new interface copy contains no em dash; original metadata is preserved.
  • R-03 PASS: eighteen layouts fit 320 through 1440 pixels; navigation and controls have 44px targets.
  • R-17 PASS: the existing 466-aircraft example remains tied to the linked experiment; no statistics added.
  • R-18 PASS: no testimonials or invented identities.
  • R-23 PASS: existing assets, name and navigation are retained; layout work is within the user's refresh request.
  • R-24 PASS: all local navigation destinations were clicked and resolved.
  • R-25 PASS: six axe A/AA scans found no contrast violations; hover/focus colors were inspected.
  • R-26 PASS: 103 recorded interaction checks; no new dead controls.
  • R-27 PASS: the viewer has unopened, loading, loaded and recoverable error states.
  • R-28 PASS: no FAQ added.
  • R-32 PASS: skip link, sliders, disclosure and dialog keyboard paths verified; visible focus retained.
  • R-33 PASS: UI changes were written directly in the source; no source-rewriting helper shipped.
  • R-34 PASS: one fixed theme, consistent across all three pages; no theme toggle.
  • R-35 PASS: HTTP-served Chromium run, clicked local navigation and controls, zero unexpected console errors.
  • R-36 PASS: no invented claims or measured-performance numbers.
  • R-37 PASS: incumbent direction and explicit dials recorded before UI edits in DESIGN.md.
  • R-38 PASS: original research content and plot assets preserved; no fabricated content.
  • R-39 PASS: native controls, labels, focus, image alternatives and reduced-motion handling verified.
  • R-40 PASS: images retain dimensions, scripts defer, fonts are local, no third-party runtime scripts added.
  • R-01 PASS: flat charcoal surfaces and amber signal emphasis; no decorative gradients/glows.
  • R-04 PASS: existing aircraft/microphone geometry explains the model; no generic icon set added.
  • R-06 PASS: retained editorial serif and readable sans; measurement labels identify axes and controls.
  • R-07 PASS: wavefronts exist inside the simulation; no decorative page grid.
  • R-08 PASS: primary action is text; retained arrows denote destinations or diagram direction.
  • R-09 PASS: no badges added.
  • R-10 PASS: no glass or backdrop blur.
  • R-12 PASS: no floating card shadows.
  • R-13 PASS: no glow layer.
  • R-14 PASS: research figures and a method list replace a uniform feature-card rhythm.
  • R-19 PASS: one brief introduction and explicit simulation playback; no scroll reveals.
  • R-22 PASS: only the existing domain model and actual experiment imagery.
  • R-05 PASS: existing research narrative; split model, evidence figures and method list have different composition.
  • R-11 PASS: 12px frames and 4px controls; no pill system.
  • R-15 PASS: primary action says Read the method and opens that page.
  • R-16 PASS: no new marketing buzzwords.
  • R-20 PASS: the physical wave/curve instrument and actual aircraft evidence anchor the identity.
  • R-21 PASS: dark canvas preserves the incumbent world and continuity with measurement figures.
  • R-29 PASS: neutral surfaces, paper text and one amber accent; plot data retains original colors.
  • R-30 PASS: no reference website cloned.
  • R-31 PASS: one-line reasons for palette, layout, type, spacing, containers, imagery and motion in DESIGN.md.
  • Liveliness PASS: ENERGY 2 / RHYTHM 3 / MOTION 2; the instrument is the focal point and identity motif.
  • C-1..C-5 PASS: documented decisions, working controls, real content, responsive/keyboard paths and factual claims.

Other guide checks

Impeccable context and its manual detector were run. The detector identified a real skip-link hover contrast issue, which was fixed. Its padding warnings were checked against computed/rendered padding (24px for method rows and physics details, 36px for page transitions). Explicit longhands now express that padding. The generic Geist warning is an intentional retention of the incumbent, self-hosted type system, as documented in DESIGN.md.

The strict taste-skill template gate ended GATE: OPEN 35 (rounds used up). It improved from 46 to 38 to 35 findings across the three allowed rounds. No waivers were created. Every remaining finding is recorded in SITE_REFRESH_TASTE_GATE.md, including fixed screen geometry, mandatory visuals in prose sections, the original headline, container conventions and the retained scientific captions. The guide says to stop fixing after the third round. This gate is open, independently of the passing browser and accessibility checks; the proposed change is left unmerged for review.