Move the generic extractor out of event-extractors/

extension/event-extractors/ is the extensibility point for PER-SITE
extractors. The generic extractor is a fundamentally different thing — one
fixed, site-agnostic reader that every per-site extractor layers over, not
another entry in the set — so it doesn't belong in that folder:

  event-extractors/core/generic.js       -> generic-extractor.js
  event-extractors/core/generic-sites.js -> generic-sites.js

Both now sit at the extension root beside config.js / fallback-policy.js, and
event-extractors/ holds only the pipeline's own registry/orchestrator/helpers
and the custom/ sources.

The load-order generator emits extension-root-relative paths throughout (it
already did for the output) and pins the two generic files by name, since
they're the only injected files outside that folder; the ordering rule is
unchanged, with generic-sites.js still pinned behind every custom source so a
dedicated source always wins the host match.

Also added the two files to the release shipping set — the zip guard caught
their absence, since it was keyed on the event-extractors/ directory entry.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VE29ox6D2VwKLJJEy4bmfy

Author: claude

.claudinite/local/packs/gcec/RULES.md

@@ -176,7 +176,7 @@ on that pipeline). Adding a cached live case is the
   `og:locale`, or a non-English language; a bare `en` (region unknown) stays
   month-first rather than guess. Resolution lives in `helpers/dates.js`
   (`parseDateFromText` / `normalizeDateValue` take a `dayFirst` flag, threaded
-  from `core/generic.js`'s `pageUsesDayFirstDates`), mirroring
+  from `generic-extractor.js`'s `pageUsesDayFirstDates`), mirroring
   `derive-timezone.js`'s locale read; unambiguous dates (a part > 12) and the
   `.` / `-` separators are always day-first regardless (#686).
 - **`fetch-page.yml` is attended by its dispatcher** (the workflow-failure
@@ -206,28 +206,33 @@ on that pipeline). Adding a cached live case is the
   hints, see `helpers/derive-timezone.js` — `ctz`). This is the comparison the
   fallback-coverage gate automates — and a case where it shows NO gap is a
   candidate for deleting the per-site source and listing the host in
-  `core/generic-sites.js`.
+  `generic-sites.js`.
 
 ## Architecture rules of the road
 
 Whenever we agree on a new or changed top-level architectural guideline, update
 this section as part of the same change (the design doc itself is
 `dev/procedures/highLevelDesign.md`):
 
-- **The core generic extractor is the base layer of every extraction**
-  (`extension/event-extractors/core/generic.js`) — it runs on every page, and a
-  per-site source is only a layer of *overrides* merged over it, stating the
-  fields it gets better. A source never re-reads what the page already says about
-  itself, and nothing in `core/` may know about a specific site.
+- **`extension/event-extractors/` is the extensibility point for PER-SITE
+  extractors, and nothing else belongs in it.** The generic extractor is a
+  fundamentally different thing — one fixed, site-agnostic reader, not another
+  entry in the set — so it and its host list live at the extension root
+  (`extension/generic-extractor.js`, `extension/generic-sites.js`).
+- **The core generic extractor is the base layer of every extraction** — it runs
+  on every page, and a per-site source is only a layer of *overrides* merged over
+  it, stating the fields it gets better. A source never re-reads what the page
+  already says about itself, and the generic extractor may never know about a
+  specific site.
 - **A site we fully support gets no extractor file when there is nothing to
-  override.** Its host goes in `core/generic-sites.js` (a `matches`-only
+  override.** Its host goes in `generic-sites.js` (a `matches`-only
   registration) and stays fully supported — green icon, no correction prompt.
   When a `custom/<site>.js` shrinks to nothing, delete it and move the host
   there; when a generic-site starts needing a fix the generic extractor can't
   make site-agnostically, move it back. Never keep a file whose only content is
   a restatement of the generic base, and never annotate one as "nothing to add".
 - Adding support for a new host is the most common change — the architecture
-  must keep it a single-file change: a host added to `core/generic-sites.js`,
+  must keep it a single-file change: a host added to `generic-sites.js`,
   or one self-contained new `extension/event-extractors/custom/<site>.js`, plus
   regenerating the load list — touching nothing else and assuming nothing about
   other extractors.

.claudinite/local/packs/gcec/tasks/fallback-extractor-improvements/postconditions.sh

@@ -17,7 +17,7 @@
 #
 # Checks (all deterministic):
 #   1. SCOPE — the diff touches ONLY the core generic extractor and its allowed
-#      companions: core/generic.js and the shared helpers/*.js (the generic
+#      companions: generic-extractor.js and the shared helpers/*.js (the generic
 #      extractor), a covering test in extension-test/event-extractors/extraction.test.js,
 #      and the regenerated fallback-coverage GENERATED artifacts. A custom/<site>.js
 #      edit or a host special-case is out of scope — it isn't a GENERIC win. (This
@@ -55,7 +55,7 @@ git rev-parse --verify --quiet "$ref" >/dev/null 2>&1 || ref=main
 git rev-parse --verify --quiet "$ref" >/dev/null 2>&1 || ref=HEAD
 
 # ── 1. SCOPE ──
-allowed='^(extension/event-extractors/core/generic\.js|extension/event-extractors/helpers/[^/]+\.js|extension-test/event-extractors/extraction\.test\.js|dev/requirements/extractor/fallback/fallback-coverage\.(baseline\.GENERATED\.json|GENERATED\.md))$'
+allowed='^(extension/generic-extractor\.js|extension/event-extractors/helpers/[^/]+\.js|extension-test/event-extractors/extraction\.test\.js|dev/requirements/extractor/fallback/fallback-coverage\.(baseline\.GENERATED\.json|GENERATED\.md))$'
 changed="$( { git diff --name-only "$ref"...HEAD 2>/dev/null; git diff --name-only HEAD 2>/dev/null; \
               git ls-files --others --exclude-standard; } | sort -u | sed '/^$/d' )"
 offenders="$(printf '%s\n' "$changed" | grep -Ev "$allowed" || true)"

.claudinite/local/packs/gcec/tasks/fallback-extractor-improvements/task.md

@@ -1,7 +1,7 @@
 # Fallback-coverage routine
 
 Weekly Claude routine (on a strong model), offline on a fresh clone. Goal: make the **generic**
-fallback extractor — `extension/event-extractors/core/generic.js` and the
+fallback extractor — `extension/generic-extractor.js` and the
 shared `extension/event-extractors/helpers/*.js` — recover **more** of what the
 dedicated per-site sources get, measured by the fallback-coverage gate. **Most runs
 correctly change nothing** — the gate already banks every prior win, so a new
@@ -53,7 +53,7 @@ work, because a run that trips one there is marked **failed**:
   JSON-LD, OG/meta, microdata, `<time datetime>`). Lifting one `custom/<site>.js`'s
   logic or special-casing a host is out of scope: if the only beneficiary is a
   single case and the rule wouldn't fire elsewhere, it isn't generic. **The only
-  files you may leave changed are `core/generic.js`, `helpers/*.js`, your
+  files you may leave changed are `generic-extractor.js`, `helpers/*.js`, your
   covering test in `extension-test/event-extractors/extraction.test.js`, and the
   regenerated fallback-coverage GENERATED artifacts** — anything else fails scope.
 - **A real, non-regressing win** (postcondition: *win*, via the suite's gate). Every

dev/build/gen-load-order.js

@@ -5,20 +5,26 @@
 //     exercised by the tests).
 // Run it with `npm run index` after adding/removing a source or extractor file.
 //
-// Adding a source is a single-new-file change: drop
+// Adding a per-site extractor is a single-new-file change: drop
 // event-extractors/custom/<site>.js and rerun this; the list is regenerated
 // mechanically rather than kept in sync by hand. A CI test (the load-order
 // drift guard) asserts the committed file matches what this generator would
 // produce, so it can never silently drift.
 //
 // Ordering rule (the only ordering that matters): the registry and shared
 // helpers load FIRST (they build globalThis.GCal), the per-site sources
-// (custom/*.js) load next — sorted, for a stable conflict-free list — and the
-// tail is pinned: the core generic extractor and the hosts it alone serves
-// (core/*.js), then the orchestrator (assemble-events.js, whose completion value
-// is the extraction result).
+// (event-extractors/custom/*.js) load next — sorted, for a stable conflict-free
+// list — and the tail is pinned: the generic extractor and the hosts it alone
+// serves, then the orchestrator (event-extractors/assemble-events.js, whose
+// completion value is the extraction result).
 //
-// core/generic-sites.js is pinned AFTER every custom/<site>.js on purpose: it
+// The two generic files live OUTSIDE event-extractors/ (at the extension root):
+// event-extractors/ is the extensibility point for per-site extractors, and the
+// generic extractor is a different thing — the base layer every one of them
+// overrides. They're pinned by name here rather than globbed, since they're the
+// only injected files outside that folder.
+//
+// generic-sites.js is pinned AFTER every custom/<site>.js on purpose: it
 // registers a bare matcher per host served by the generic extractor alone, and
 // assemble-events.js takes the FIRST source whose matches() accepts the host, so
 // a dedicated source added for one of those hosts always wins.
@@ -38,10 +44,12 @@ const EXT = path.join(ROOT, "extension");
 const DIR = "event-extractors";
 const OUTPUT = "event-extractors/load-order.generated.json";
 
-const PINNED_FIRST = ["registry.js"]; // followed by helpers/*, added below
-// The core layer, then the orchestrator. See the ordering rule above for why
-// core/generic-sites.js is pinned behind the per-site sources.
-const PINNED_LAST = ["core/generic.js", "core/generic-sites.js", "assemble-events.js"];
+// All paths below are extension-root-relative, so the two generic files carry no
+// directory prefix while everything else sits under event-extractors/.
+const PINNED_FIRST = [`${DIR}/registry.js`]; // followed by helpers/*, added below
+// The generic layer, then the orchestrator. See the ordering rule above for why
+// generic-sites.js is pinned behind the per-site sources.
+const PINNED_LAST = ["generic-extractor.js", "generic-sites.js", `${DIR}/assemble-events.js`];
 
 const isJs = (f) => f.endsWith(".js");
 
@@ -50,35 +58,29 @@ function computeLoadOrder() {
     .readdirSync(path.join(EXT, DIR, "helpers"))
     .filter(isJs)
     .sort()
-    .map((f) => `helpers/${f}`);
+    .map((f) => `${DIR}/helpers/${f}`);
 
   const sources = fs
     .readdirSync(path.join(EXT, DIR, "custom"))
     .filter(isJs)
-    .map((f) => `custom/${f}`);
+    .map((f) => `${DIR}/custom/${f}`);
 
   const pinned = new Set([...PINNED_FIRST, ...PINNED_LAST]);
   const topLevelMiddle = fs
     .readdirSync(path.join(EXT, DIR))
-    .filter((f) => isJs(f) && !pinned.has(f));
+    .filter(isJs)
+    .map((f) => `${DIR}/${f}`)
+    .filter((f) => !pinned.has(f));
 
   const middle = [...topLevelMiddle, ...sources].sort();
 
   for (const f of [...PINNED_FIRST, ...PINNED_LAST]) {
-    if (!fs.existsSync(path.join(EXT, DIR, f))) {
-      throw new Error(`expected ${DIR}/${f} to exist`);
-    }
-  }
-
-  // core/ is pinned by name rather than globbed, so a file added there would
-  // otherwise drop out of the load list silently. Fail loudly instead.
-  for (const f of fs.readdirSync(path.join(EXT, DIR, "core")).filter(isJs)) {
-    if (!PINNED_LAST.includes(`core/${f}`)) {
-      throw new Error(`${DIR}/core/${f} is not in PINNED_LAST — add it (order matters, see the ordering rule)`);
+    if (!fs.existsSync(path.join(EXT, f))) {
+      throw new Error(`expected ${f} to exist`);
     }
   }
 
-  return [...PINNED_FIRST, ...helpers, ...middle, ...PINNED_LAST].map((f) => `${DIR}/${f}`);
+  return [...PINNED_FIRST, ...helpers, ...middle, ...PINNED_LAST];
 }
 
 // JSON, one path per line, trailing newline — a clean, reviewable diff.

dev/build/release/shipping-files.js

@@ -20,7 +20,9 @@ const SHIPPING_PATHS = [
   "config.js", // tunable product decisions; imported by the popup modules at runtime
   "fallback-policy.js", // host classifier for the generic fallback; imported by the popup at runtime
   "fallback-lists.json", // the allow/deny/supported host lists; fetched by the worker, imported by config.js
-  "event-extractors", // the extractors (custom sources + helpers + orchestrator) + generated load list (popup fetches/injects)
+  "generic-extractor.js", // the core generic extractor — injected on every page (outside event-extractors/ by design)
+  "generic-sites.js", // the hosts it alone serves; injected alongside it
+  "event-extractors", // the per-site extractors + helpers + registry/orchestrator + generated load list (popup fetches/injects)
   "events-popup", // popup (html/css/js + views) and the calendar-URL builder
   "icon", // the toolbar-icon service worker and its images/ PNGs the manifest references
 ];

dev/procedures/fileDescriptions.md

@@ -17,8 +17,8 @@ extension does.
 | `extension/icon/toolbar-icon.js` | Background service worker: registers `chrome.declarativeContent` rules (from `fallback-lists.json`) so the browser colors the toolbar icon by host pattern — green on supported hosts, gray on denylisted ones, blue elsewhere — without the extension reading any tab URL |
 | `extension/event-extractors/registry.js` | Bootstraps `GCal`, the `GCal.sources` registry, and `isSupportedHost` |
 | `extension/event-extractors/helpers/` | Shared utilities any extractor may use, split by concern: DOM, text (rich-text/`htmlToText`/`parts`), dates, timezones, timezone-names, merge, and `embedded-events` (the `GCal.embeddedEvents` schema.org JSON-LD reader) |
-| `extension/event-extractors/core/generic.js` | `GCal.genericExtractor`: THE core generic extractor, run on every page — best-effort event (embedded JSON-LD + generic heuristics over meta tags, microdata, `<time>`, and visible text). It is the base layer every per-site source overrides, the only extractor on an unsupported host, and the whole of the support for a `generic-sites.js` host |
-| `extension/event-extractors/core/generic-sites.js` | The hosts fully covered by `core/generic.js` alone — each registers a `matches`-only source, so the host is supported with no per-site file. Loaded after `custom/` so a dedicated source always wins the match |
+| `extension/generic-extractor.js` | `GCal.genericExtractor`: THE core generic extractor, run on every page — best-effort event (embedded JSON-LD + generic heuristics over meta tags, microdata, `<time>`, and visible text). It is the base layer every per-site source overrides, the only extractor on an unsupported host, and the whole of the support for a `generic-sites.js` host |
+| `extension/generic-sites.js` | The hosts fully covered by `generic-extractor.js` alone — each registers a `matches`-only source, so the host is supported with no per-site file. Loaded after `custom/` so a dedicated source always wins the match |
 | `extension/event-extractors/custom/meetup.js`, `eventbrite.js`, `edinburghfringe.js`, `telavivcinematheque.js`, `ticketmaster.js`, … | One per site the generic extractor gets wrong: hardcoded selectors + inline host matcher, stating ONLY the fields it overrides (or its own enumerated `events`) |
 | `extension/events-popup/build-calendar-url.js` | Builds the pre-filled Google Calendar template URL (incl. markdown→HTML for details) |
 | `extension/event-extractors/assemble-events.js` | Orchestrator `GCal.extract()`: runs the core generic extractor for the base events and merges the matched site source's overrides over them; normalizes/sorts events and reports `supported` + `fallback` |

dev/procedures/highLevelDesign.md

@@ -7,7 +7,7 @@ in [requirements.md §12–§16](../requirements/requirements.md); the per-file
 
 `toolbar-icon.js` colors the toolbar icon by host (a source matches or it doesn't). On
 click, `popup.js` injects `event-extractors/` and runs `assemble-events.js`, which runs
-`core/generic.js` and merges the matching `custom/<site>.js` over it;
+`generic-extractor.js` and merges the matching `custom/<site>.js` over it;
 `build-calendar-url.js` builds the URL `events-view.js` renders.
 
 The top-level **architecture rules of the road** live in the gcec local pack's
@@ -31,12 +31,15 @@ non-time field are folded into one multi-instance event by the assembler.
 
 Extraction is **two layers, not two paths**:
 
-1. **The core generic extractor** (`extension/event-extractors/core/generic.js`)
+1. **The core generic extractor** (`extension/generic-extractor.js`)
    runs on **every** page, whatever its host, and produces the base events from
    the page's own self-description: embedded schema.org JSON-LD plus generic
    heuristics (microdata, Open Graph / meta tags, `<time datetime>`,
    `<h1>`/`<address>`, venue/location-named elements, and finally a date/time
    scan over the visible text). Nothing in it may know about a specific site.
+   It sits at the extension root, **outside** `event-extractors/`: that folder is
+   the extensibility point for per-site extractors, and this is a fundamentally
+   different thing — the one fixed reader they all layer over.
 
 2. **A per-site source** (`extension/event-extractors/custom/<site>.js`) is a
    layer of **overrides** on top of that base. Its `extract()` returns only the
@@ -47,7 +50,7 @@ Extraction is **two layers, not two paths**:
    Each lives in its own file with a comment describing the HTML it expects.
 
 So a site whose pages describe themselves completely needs **no source file at
-all**: its host goes in `extension/event-extractors/core/generic-sites.js` and is
+all**: its host goes in `extension/generic-sites.js` and is
 fully supported by the base alone — same green icon, same popup, no
 "Suggest Correction" prompt. When adding support for a platform, check the
 generic extractor first and only write a `custom/<site>.js` for what it gets

dev/requirements/extractor/cases/extractor-support.11.16.case.js

@@ -5,8 +5,8 @@
 module.exports = {
   tbd: true,
   description:
-    "facebook.com — fully covered by the core generic extractor (core/generic-sites.js); " +
+    "facebook.com — fully covered by the core generic extractor (generic-sites.js); " +
     "unit-tested only, bot-blocked from caching",
   host: "facebook.com",
-  source: "extension/event-extractors/core/generic-sites.js",
+  source: "extension/generic-sites.js",
 };

dev/requirements/extractor/cases/extractor-support.11.6.case.js

@@ -1,8 +1,8 @@
 "use strict";
 
 module.exports = {
-  description: "bandsintown.com — fully covered by the core generic extractor (core/generic-sites.js)",
+  description: "bandsintown.com — fully covered by the core generic extractor (generic-sites.js)",
   host: "bandsintown.com",
-  source: "extension/event-extractors/core/generic-sites.js",
+  source: "extension/generic-sites.js",
   page: "bandsintown-berry-sakharof",
 };

dev/requirements/extractor/cases/extractor-support.11.8.case.js

@@ -1,8 +1,8 @@
 "use strict";
 
 module.exports = {
-  description: "dash.datadoghq.com — fully covered by the core generic extractor (core/generic-sites.js)",
+  description: "dash.datadoghq.com — fully covered by the core generic extractor (generic-sites.js)",
   host: "dash.datadoghq.com",
-  source: "extension/event-extractors/core/generic-sites.js",
+  source: "extension/generic-sites.js",
   page: "dash-datadoghq",
 };