feedkeeper 0.35.7 → 0.35.8

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "feedkeeper",
-  "version": "0.35.7",
+  "version": "0.35.8",
   "description": "A lightweight, self-hosted RSS reader and remote MCP server for your news, blogs, and AI workflows.",
   "keywords": [
     "rss",

package/server/dist/feeds/content/index.js

@@ -3,6 +3,7 @@ import { EMBED_RULES, resolveEmbeds } from "./embeds/index.js";
 import { embedHeadings } from "./embeds/headings.js";
 import { boilerplate } from "./noise/boilerplate.js";
 import { loadingPlaceholders } from "./noise/loading.js";
+import { pageChrome } from "./noise/pageChrome.js";
 import { trackingPixels } from "./noise/pixels.js";
 import { permalinkGlyphs } from "./noise/permalinkGlyphs.js";
 import { sidebars } from "./noise/sidebars.js";
@@ -12,7 +13,7 @@ import { teaserLinks } from "./noise/teaserLinks.js";
  * tags that are removed next), then the clutter around the story. The tests check each rule against its fixtures in
  * `test/fixtures/content/<rule name>` and that running it twice changes nothing. See `docs/content-rules.md`.
  */
-export const PAGE_RULES = [lazyImages, ...EMBED_RULES, embedHeadings, sidebars, loadingPlaceholders, trackingPixels, boilerplate, teaserLinks, permalinkGlyphs];
+export const PAGE_RULES = [lazyImages, ...EMBED_RULES, embedHeadings, pageChrome, sidebars, loadingPlaceholders, trackingPixels, boilerplate, teaserLinks, permalinkGlyphs];
 export { resolveEmbeds };
 /** What a fetched page goes through before its scripts, frames and `noscript` blocks are removed. */
 export function cleanPage(root, stats = {}) {

package/server/dist/feeds/content/noise/pageChrome.js

@@ -0,0 +1,38 @@
+import { count, flat, pruneEmpty } from "../shared.js";
+/** A paragraph this long is prose, not a menu entry: an element that holds one is never page chrome, whatever it is called. */
+const PROSE_CHARACTERS = 200;
+/** What only a site's own navigation contains. */
+const NAVIGATION = "nav, [role='navigation'], [role='menubar'], menu";
+/** A page header holds the site's navigation, or even the page's main area; an article's own header holds a headline and a byline. */
+const PAGE_HEADER_PARTS = `${NAVIGATION}, main`;
+function holdsProse(element) {
+    return [...element.querySelectorAll("p")].some(paragraph => flat(paragraph.textContent).length >= PROSE_CHARACTERS);
+}
+/**
+ * Some publishers put their whole page header into the text of an article: a menu of magazines with empty entries
+ * (their icons are drawn by a stylesheet), service links, the logo and lists of the publisher's other pages. A reader
+ * shows all of it before the first line of the story. A header that contains navigation, and any navigation of its own, are
+ * the site and not the story: they go, unless prose sits inside, which means the markup is broken and wraps the article.
+ */
+export const pageChrome = {
+    name: "page-chrome",
+    description: "A page header that holds the site's navigation, and any navigation element, are removed from the text of an article.",
+    apply(root, stats) {
+        let changed = false;
+        const remove = (element) => {
+            if (!element.parentNode || holdsProse(element))
+                return;
+            const parent = element.parentNode;
+            element.remove();
+            pruneEmpty(parent);
+            count(stats, "page-chrome");
+            changed = true;
+        };
+        for (const header of [...root.querySelectorAll("header")].filter(candidate => candidate.querySelector(PAGE_HEADER_PARTS)))
+            remove(header);
+        for (const navigation of [...root.querySelectorAll(NAVIGATION)])
+            remove(navigation);
+        return changed;
+    },
+};
+//# sourceMappingURL=pageChrome.js.map

package/server/dist/feeds/contentNoise.js

@@ -7,6 +7,7 @@ import { SIDEBAR_HEADINGS } from "./content/patterns/sidebars/index.js";
 import { resolveEmbeds } from "./content/embeds/index.js";
 import { loadingPlaceholders } from "./content/noise/loading.js";
 import { trackingPixels } from "./content/noise/pixels.js";
+import { pageChrome } from "./content/noise/pageChrome.js";
 import { permalinkGlyphs } from "./content/noise/permalinkGlyphs.js";
 import { teaserLinks } from "./content/noise/teaserLinks.js";
 import { markVideoTeasers } from "./content/noise/videoTeasers.js";
@@ -102,6 +103,8 @@ function removeReaderNoise(root, title, stats, author) {
         changed = true;
     if (trackingPixels.apply(root, stats))
         changed = true;
+    if (pageChrome.apply(root, stats))
+        changed = true;
     for (const selector of BREADCRUMB_MARKERS) {
         try {
             root.querySelectorAll(selector).forEach(element => drop(element, "breadcrumb"));

package/server/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@feedkeeper/server",
-  "version": "0.35.7",
+  "version": "0.35.8",
   "private": true,
   "type": "module"
 }