CommonerLLP · skishchampi · May 10, 2026 · May 10, 2026 · May 10, 2026 · May 10, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,177 +1,33 @@
 # Changelog
 
-This file is append-only. Entries record public project changes that are
-too detailed for the README but useful for maintainers, reviewers, and
-future contributors.
-
-## 2026-05-08 — UI Redesign & Article 16 Political Palette
-
-The listing card UI has been redesigned to improve scannability and accessibility. 
-- **Typography & Hierarchy**: Increased institution font size to 28px and adjusted line-heights to establish a clearer visual order. The listings column is now capped at 740px to improve line length for readability.
-- **Article 16 Palette**: Shifted the reservation status pill to a Bahujan-political color register:
-    - **Ambedkar Blue**: Confirmed roster disclosure or Special Recruitment Drive (SRD).
-    - **Saffron**: Institutional exclusion / No reservation (Private Universities).
-    - **Grey**: Unclear or undisclosed status.
-- **Accessibility**: Increased tap areas for the star/save buttons (WCAG 2.5.5 AA) and re-anchored footer popovers to prevent viewport overflow.
-
-## 2026-05-08 — `sansad-semantic-crawler` bumped to v0.4.0
-
-`requirements.txt` now pins the upstream crawler at
-[`v0.4.0`](https://github.com/CommonerLLP/sansad-semantic-crawler/releases/tag/v0.4.0).
-This release automates the **politician enrichment layer**:
-
-- **Automated Party/State Lookup**: Question records now include an `asker_details`
-  block (party, party_name, state, house) pulled from the latest official
-  member lists. No more manual party-mapping in `consolidate_corpus.py`.
-- **Committee Composition Rosters**: The crawler can now fetch the full
-  roster of parliamentary standing committees using a hybrid API and
-  PDF/LLM strategy. This enables tracking how committee membership
-  (and the political balance within) changes from one report to the next.
-- **Refactored Base Architecture**: Improved provenance tracking and
-  PDF sanity checks.
-
-This bump is **behaviour-preserving**: the single-schema assumption for
-questions remains intact, but manifests will now contain richer metadata
-by default.
-
-## 2026-05-06 — `sansad-semantic-crawler` bumped to v0.2.0
-
-`requirements.txt` now pins the upstream crawler at
-[`v0.2.0`](https://github.com/CommonerLLP/sansad-semantic-crawler/releases/tag/v0.2.0)
-(was `v0.1.0`). The new release ships a **pluggable-classifier
-architecture** — regex (default, back-compat), embeddings (Sentence
-Transformers anchor similarity), LLM (OpenAI-compat chat-completions
-JSON tagging), or ensemble (combine modes). Optional pip extras:
-`[embeddings]`, `[llm]`, `[all]`. The package never ships model
-weights; users supply their own runtime (Ollama, vLLM, llama.cpp,
-mlx-lm, transformers, or any OpenAI-compatible hosted service).
-
-This bump is **behaviour-preserving for this project**:
-`notes/topics/cei-vacancies.json` does not declare a `classifier`
-block, so v0.2.0 transparently falls back to the regex classifier
-that drove every Gap chart prior to today. Adopting embeddings or
-LLM modes is a separate, opt-in editorial decision.
-
-`make test` (153 + 9 skipped) and `npm test` (126 across 11 files)
-unaffected. The bump touches one line in `requirements.txt`.
-
-## 2026-05-06 — Parliamentary-corpus crawler extracted and externalised
-
-The three legacy scripts that built *The Gap*'s parliamentary corpus
-— `scripts/sansad_crawl.py` (291 lines), `scripts/sansad_rs_crawl.py`
-(215 lines), and `scripts/sansad_download_pdfs.py` (125 lines) — are
-**retired**. Their behaviour (LS DSpace API + RS rsdoc.nic.in API +
-PDF discovery + dedup-on-resume) now lives in a separately-released
-public-good package, **`sansad-semantic-crawler`**, hosted at
-[github.com/CommonerLLP/sansad-semantic-crawler](https://github.com/CommonerLLP/sansad-semantic-crawler)
-and pinned at `v0.1.0` in `requirements.txt`.
-
-The package is config-driven: it expects a topic-profile JSON that
-encodes search groups, ministry filters, and tag rules. The faculty-
-vacancy / reservation / Mission-Mode lens that drove the legacy
-scripts is now `notes/topics/cei-vacancies.json` — gitignored, since
-the analytical lens is project-specific and the public package ships
-only a `libraries.json` example for `theright2read`.
-
-What the host project picks up in exchange:
-
-- **One canonical schema for both houses.** The legacy LS manifest
-  used `questiontype` / `questionno` / `members`; the legacy RS
-  manifest used `qtype` / `qno` / `asker`. The package emits
-  `qtype` / `qno` / `askers` directly for both houses.
-  `scripts/consolidate_corpus.py` is rewritten to consume that single
-  schema, which dropped roughly half its branching logic.
-- **Stable `key` field on every record** (`LS|U|178|2024-11-25` /
-  `RS|S|365|2025-07-23`), so dedup is no longer a per-script
-  computation. This was always how `consolidate_corpus.py`
-  internally normalised — it is now a guaranteed property of the
-  upstream manifest.
-- **Re-usable PDF naming** — the package writes LS PDFs to
-  `data/_sansad_crawl/pdfs/ls/` and RS PDFs to
-  `.../pdfs/rs/`. Filenames match the legacy convention
-  (`{qtype-letter}{qno}_{slug}.pdf`), so existing PDFs can be moved
-  into the new tree without re-download if the maintainer chooses
-  to skip the full re-crawl.
-
-Operational entry points: `make corpus-refresh` (full pipeline:
-crawl → parse → consolidate). See `make help` for the per-step
-targets.
-
-## 2026-05-06 — Test counts + repo-layout refresh
-
-The 2026-05-05 entry below describes a frontend test floor of 81 Vitest
-tests across 4 files (`sanitize`, `classify`, `excerpt`, `schema`) and
-notes that 5 lib modules still lacked coverage. Both numbers are
-superseded:
-
-- **Python**: 153 tests + 9 skipped (was 119).
-- **Vitest**: 117 tests across 11 files (was 81 across 4).
-- 11 of 13 `docs/lib/` modules now have at least smoke / contract
-  coverage: `sanitize`, `classify`, `excerpt`, `schema`,
-  `current-validator`, `card-helpers`, `render-card`, `filters`,
-  `map`, `render-tabs`, `search`. The two without dedicated unit
-  tests are `charts.js` (chart data + Resources-tab payload) and
-  `state.js` (a thin shared mutable-state holder); both are exercised
-  indirectly by the higher-level tests but warrant direct contracts
-  next time they're touched.
-
-The Repository-layout block is also updated above to reflect the
-public-tree files added since the original was written:
-`docs/lib/`, `docs/MISTAKES.md`, `docs/PARSER-ARCHITECTURE.md`,
-`LICENSE`, `CITATION.cff`, `package.json`, `tests/`. The orphaned
-`TECHDEBT.md` line is removed: that file is part of the maintainer's
-private working notes (`/notes/` is gitignored), not the public
-tree, so the original layout entry was always pointing at a path
-that GitHub never sees.
-
-## 2026-05-06 — Project relicensed to non-commercial terms
-
-The Licence section above is rewritten as of this date. Prior to
-2026-05-06 the project shipped under MIT (code) and CC BY-SA 4.0
-(data); both permitted commercial use. From 2026-05-06 forward,
-the project is non-commercial source-available: PolyForm
-Noncommercial 1.0.0 for code, CC BY-NC-SA 4.0 for data and corpus.
-
-The change is not retroactive against existing users — both MIT and
-CC BY-SA 4.0 are perpetual for any recipient who exercised rights
-under them. New copies of the code and data going forward are
-governed by the new terms.
-
-The intent: this work is funded indirectly by the Indian public,
-exists to surface a public-interest argument, and should never
-become a commercial product — anyone's, including the maintainer's.
-The site footer and the colophon disclaimer on every page are
-updated to match. A `LICENSE` file with the canonical PolyForm
-text and a `CITATION.cff` are added at the repository root.
-
-## 2026-05-05 — Phase 2 frontend refactor
-
-The "Repository layout" block above describes `docs/app.js` as "all
-SPA logic" — that was true at the time of writing. As of commit
-`2d50c7c`, `app.js` is **728 lines of orchestration only** (imports,
-`loadData`, `render()`, tab routing, event wiring); the bulk of the
-SPA logic now lives in **9 ESM modules under `docs/lib/`**:
-
-| Module | Purpose |
-|---|---|
-| `lib/sanitize.js` | `escapeHTML` / `safeUrl` / URL allowlist *(tested)* |
-| `lib/schema.js` | Zod schemas for runtime + test-time validation *(tested)* |
-| `lib/classify.js` | Field tags / position rank / listing quality *(tested)* |
-| `lib/excerpt.js` | `raw_text_excerpt` sanitiser *(tested)* |
-| `lib/charts.js` | Vacancies tab + The Gap charts + resources data |
-| `lib/state.js` | Shared mutable state holder (`state.ADS`, `state.SAVED`, etc.) |
-| `lib/card-helpers.js` | Per-card cue extractors and rank/discipline formatters |
-| `lib/render-card.js` | `renderAd()` + hiring-trap detection + card wiring |
-| `lib/filters.js` | Filter / sort / search + reactive facet counts |
-| `lib/map.js` | Leaflet init + marker updates |
-| `lib/render-tabs.js` | Resources / Saved / Coverage tab renderers |
-
-Frontend tests live under `tests/` (Vitest); 81 tests across 4 files
-covering `sanitize`, `classify`, `excerpt`, and `schema`. Run with
-`npm test`. The remaining 5 lib modules ship without unit tests yet —
-backfilling those is deliberate next-step work.
-
-The "Running it locally" `make test` command above runs the 119-test
-**Python** scraper suite. Frontend tests run separately via `npm test`.
-Both should be green before opening a PR that touches their respective
-trees.
+## 2026-05-10
+- **feat: comprehensive accessibility remediation for whoseuniversity.org**
+  - Implemented findings from the design audit to ensure the forensic record is accessible to researchers using assistive technologies.
+  - **Semantic Navigation:** Corrected heading hierarchy in "The Gap" report.
+  - **ARIA Labeling:** Added `aria-label` to search inputs and descriptive `<title>` tags to SVGs.
+  - **Focus Management:** Updated filter popovers to capture focus on open and return it on close.
+  - **Map Accessibility:** Enabled keyboard navigation for Leaflet markers.
+- **feat: custom SVG map markers with institution-specific icons**
+  - Replaced `L.circleMarker` with custom `L.divIcon` SVG pins for improved visual hierarchy.
+  - Assigned unique icons for Universities (🎓), Technical HEIs (⚙️), and Management Institutions (📊).
+  - Implemented dynamic 'Active' state styling.
+- **feat: accessible data-pill map markers with representative palette**
+  - Implemented Airbnb-style 'Data Pills' showing the number of jobs.
+  - Added institutional symbols to active pills for color-blind accessibility.
+  - Applied representative color palette: Saffron for IIM/Private, Light Blue for IIT/IISc, Ambedkar Blue for Central Universities.
+- **feat: Airbnb-style marker clustering**
+  - Integrated `Leaflet.markercluster` to bunch markers at low zoom levels.
+  - Implemented layered discovery: National -> Regional -> Institutional.
+  - Enhanced cluster pills to show both institutional count and total job count.
+- **fix: test suite hardening**
+  - Centralized `localStorage` mock in `tests/setup.js` to resolve conflicting mocks.
+  - Created `vitest.config.js` to standardize the test environment.
+
+## 2026-05-09
+- **chore: security posture and privacy purge**
+  - Rotated keys, purged git history, and enforced local-only policy for `CLAUDE.md`, `AGENTS.md`, and `MISTAKES.md`.
+  - Updated `.gitignore` across all repos.
+- **docs: unified READMEs and handoff protocols**
+  - Consolidated per-repo instructions into a single `_org/` source of truth.
+
+... (older entries)
diff --git a/docs/index.html b/docs/index.html
@@ -37,6 +37,9 @@
 <meta name="twitter:image" content="https://whoseuniversity.org/og.png" />
 
 <link rel="stylesheet" href="https://unpkg.com/leaflet@1.9.4/dist/leaflet.css" />
+<link rel="stylesheet" href="https://unpkg.com/leaflet.markercluster@1.4.1/dist/MarkerCluster.css" />
+<link rel="stylesheet" href="https://unpkg.com/leaflet.markercluster@1.4.1/dist/MarkerCluster.Default.css" />
+
 <link rel="preconnect" href="https://fonts.googleapis.com" />
 <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
 <!-- Typography: switched the display face from Source Serif 4 (the Google
@@ -50,6 +53,7 @@
      script-src locked to 'self' without 'unsafe-inline'. -->
 <script src="theme-init.js"></script>
 <script src="https://unpkg.com/leaflet@1.9.4/dist/leaflet.js"></script>
+<script src="https://unpkg.com/leaflet.markercluster@1.4.1/dist/leaflet.markercluster.js"></script>
 <link rel="stylesheet" href="styles.css" />
 </head>
 <body>
@@ -108,7 +112,7 @@ <h1><span class="masthead-title">Whose University?</span><span class="masthead-s
   <span class="strip-label">Filter:</span>
   <label class="filter-search" for="search">
     <span class="search-ico" aria-hidden="true">⌕</span>
-    <input id="search" type="search" value="" autocomplete="off" placeholder="Search ML, postdoc, STS, policy…" />
+    <input id="search" type="search" value="" autocomplete="off" placeholder="Search ML, postdoc, STS, policy…" aria-label="Search advertisements" />
   </label>
 
   <!-- Reserved posts toggle: a first-class chip for the dashboard's most
@@ -342,6 +346,7 @@ <h3 class="about-h3">Corrections</h3>
         Open an issue at
         <a class="github-link" href="https://github.com/commonerllp/academiaindia" target="_blank" rel="noopener" aria-label="GitHub repository: commonerllp/academiaindia">
           <svg class="github-icon" viewBox="0 0 16 16" aria-hidden="true" focusable="false">
+            <title>GitHub</title>
             <path d="M8 0C3.58 0 0 3.67 0 8.2c0 3.62 2.29 6.69 5.47 7.77.4.08.55-.18.55-.4 0-.19-.01-.84-.01-1.53-2.01.38-2.53-.5-2.69-.96-.09-.24-.48-.96-.82-1.16-.28-.16-.68-.56-.01-.57.63-.01 1.08.59 1.23.83.72 1.24 1.87.89 2.33.68.07-.53.28-.89.51-1.09-1.78-.21-3.64-.91-3.64-4.04 0-.89.31-1.62.82-2.19-.08-.21-.36-1.04.08-2.16 0 0 .67-.22 2.2.84A7.4 7.4 0 0 1 8 3.95c.68 0 1.36.09 2 .28 1.53-1.06 2.2-.84 2.2-.84.44 1.12.16 1.95.08 2.16.51.57.82 1.3.82 2.19 0 3.14-1.87 3.83-3.65 4.04.29.25.54.76.54 1.54 0 1.11-.01 2-.01 2.27 0 .22.15.48.55.4A8.09 8.09 0 0 0 16 8.2C16 3.67 12.42 0 8 0Z" />
           </svg>
           <span>GitHub</span>

diff --git a/docs/lib/card-helpers.js b/docs/lib/card-helpers.js
@@ -17,12 +17,13 @@ import { safeUrl, resolveUrl, escapeRegExp } from "./sanitize.js";
 
 /** Type-to-colour map used by the card type chip and the map markers. */
 export const TYPE_COLORS = {
-  IIT: "#1F4E79", IIM: "#2d6a4f", IISc: "#6b21a8", IISER: "#b45309",
-  NIT: "#64748b", IIIT: "#0e7490", CentralUniversity: "#92400e",
+  IIT: "#58a6ff", IIM: "#F47C20", IISc: "#58a6ff", IISER: "#58a6ff",
+  NIT: "#64748b", IIIT: "#0e7490", CentralUniversity: "#000080",
   StateUniversity: "#9a3412",
-  PrivateUniversity: "#7c3aed",
+  PrivateUniversity: "#F47C20",
 };
 
+
 export function detectAdCampus(ad) {
   const text = `${ad.title || ""} ${ad.raw_text_excerpt || ""}`;
   if (!text) return null;