feedkeeper 0.35.0 → 0.35.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "feedkeeper",
-  "version": "0.35.0",
+  "version": "0.35.2",
   "description": "A lightweight, self-hosted RSS reader and remote MCP server for your news, blogs, and AI workflows.",
   "keywords": [
     "rss",

package/server/dist/backup-cli.js

@@ -31,7 +31,11 @@ async function main() {
         process.umask(0o077);
         const { db } = await import("./db/index.js");
         try {
-            await db.backup(filePath);
+            await db.backup(filePath, {
+                progress({ remainingPages }) {
+                    return remainingPages;
+                },
+            });
             const saved = new Database(filePath, { readonly: true, fileMustExist: true });
             try {
                 assertHealthy(saved);
@@ -72,7 +76,11 @@ async function main() {
         assertHealthy(source);
         mkdirSync(dirname(databasePath), { recursive: true, mode: 0o700 });
         process.umask(0o077);
-        await source.backup(databasePath);
+        await source.backup(databasePath, {
+            progress({ remainingPages }) {
+                return remainingPages;
+            },
+        });
     }
     finally {
         source.close();

package/server/dist/feeds/articleContent.js

@@ -1,6 +1,7 @@
 import { createHash } from "node:crypto";
 import { JSDOM } from "jsdom";
 import { decodeEntities } from "./text.js";
+import { isVideoTeaser } from "./content/noise/videoTeasers.js";
 import { cleanReaderHtml, cleanSnippet, derivedText } from "./contentNoise.js";
 /**
  * Plain publisher content, shared by readers, search consumers and MCP. Deliberately not cleaned: length decisions
@@ -35,8 +36,10 @@ export function readerHtml(html, title, author) {
 }
 export function contentMetadata(item) {
     const source = item.full_content_html ? "reader" : item.content_html ? "feed" : "snippet";
-    return { revision: item.content_revision, source, status: item.extraction_status,
-        attemptedAt: item.extraction_attempted_at, retryAt: item.extraction_retry_at, isEmptyGallery: Boolean(item.is_empty_gallery) };
+    // A video's page has no text to fetch: its card is the whole content, so no client says the article is missing.
+    const videoCard = !item.full_content_html && isVideoTeaser(item.content_html, item.title);
+    return { revision: item.content_revision, source, status: videoCard ? "feed_only" : item.extraction_status,
+        attemptedAt: item.extraction_attempted_at, retryAt: videoCard ? null : item.extraction_retry_at, isEmptyGallery: Boolean(item.is_empty_gallery), ...(videoCard ? { isVideo: true } : {}) };
 }
 export function sourceText(item) {
     return `${decodeEntities(item.title) ?? ""}\n${derivedText(item.full_content_html || item.content_html, item.title) || cleanSnippet(decodeEntities(item.content_snippet), item.title) || ""}`.trim();

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

@@ -0,0 +1,76 @@
+import { JSDOM } from "jsdom";
+import { flat, pruneEmpty } from "../shared.js";
+/** The address of a page that is a video: `/video/…`, `/videos/…`, `/mediathek/…`, `/watch`, `video-1234.html`. */
+const VIDEO_PAGE = /(^|\/)(videos?|mediathek|watch|clips?)(\/|$|[-.])|\/(?:[\w-]+-)?video[\w-]*\.html?$/i;
+function isVideoAddress(href) {
+    if (!href)
+        return false;
+    try {
+        return VIDEO_PAGE.test(new URL(href, "https://example.org").pathname);
+    }
+    catch {
+        return false;
+    }
+}
+/** The posters in a text: a picture that is a link to a video page, and the block that holds it. */
+function posters(root) {
+    const found = [];
+    for (const link of root.querySelectorAll("a[href]")) {
+        if (!link.querySelector("img") || !isVideoAddress(link.getAttribute("href")))
+            continue;
+        const block = link.closest("p, div, li") ?? link.parentElement;
+        if (block)
+            found.push({ link, block });
+    }
+    return found;
+}
+/** Whether what stands beside the poster says nothing but the title. */
+function onlyTheTitle(block, title) {
+    const copy = block.cloneNode(true);
+    copy.querySelectorAll("a, img, picture, figure").forEach(node => node.remove());
+    const text = flat(copy.textContent).replace(/[[\]()\s]/g, "");
+    return text === "" || text.toLowerCase() === (title ?? "").replace(/[[\]()\s]/g, "").toLowerCase();
+}
+/**
+ * A feed entry for a video is often just the poster as a link to the video page, the title again and "[mehr]". A client shows that as a
+ * card for a video, like an embedded player, instead of an article with nothing to read: the poster stays, and under it a marked link to the
+ * video page follows (`data-media="video"`), which clients turn into the card.
+ */
+export function markVideoTeasers(root, title, stats) {
+    let changed = false;
+    for (const { link, block } of posters(root)) {
+        if (!block.isConnected && !root.contains(block))
+            continue;
+        const href = link.getAttribute("href");
+        const next = block.nextElementSibling;
+        if (next?.querySelector(`a[data-media="video"][href="${href.replace(/"/g, """)}"]`))
+            continue;
+        if (!onlyTheTitle(block, title))
+            continue;
+        // The poster stays alone in its block; the words beside it move into the card.
+        [...block.childNodes].forEach(node => {
+            if (node.nodeType === 3 || node.nodeName === "BR")
+                node.parentNode?.removeChild(node);
+        });
+        const document = block.ownerDocument;
+        const paragraph = document.createElement("p");
+        const card = document.createElement("a");
+        card.setAttribute("href", href);
+        card.setAttribute("data-media", "video");
+        card.textContent = flat(title) || "Video";
+        paragraph.append(card);
+        block.after(paragraph);
+        pruneEmpty(block.parentNode);
+        stats["video-teasers"] = (stats["video-teasers"] ?? 0) + 1;
+        changed = true;
+    }
+    return changed;
+}
+/** Whether a feed entry's text is such a video teaser; cheap to ask, as most entries fail the first test. */
+export function isVideoTeaser(html, title) {
+    if (!html || html.length > 2000 || !/<img\b/i.test(html) || !/video|mediathek|watch|clip/i.test(html))
+        return false;
+    const fragment = JSDOM.fragment(html);
+    return posters(fragment).some(({ block }) => onlyTheTitle(block, title));
+}
+//# sourceMappingURL=videoTeasers.js.map

package/server/dist/feeds/contentNoise.js

@@ -2,6 +2,7 @@ import { JSDOM } from "jsdom";
 import { CONSENT_ANCHORS, CONSENT_COMPANIONS } from "./content/patterns/consent/index.js";
 import { resolveEmbeds } from "./content/embeds/index.js";
 import { teaserLinks } from "./content/noise/teaserLinks.js";
+import { markVideoTeasers } from "./content/noise/videoTeasers.js";
 import { flat, pruneEmpty } from "./content/shared.js";
 /**
  * Publisher pages carry page chrome inside the article body: breadcrumbs, dateline repeats, player leftovers,
@@ -212,7 +213,8 @@ function removeReaderNoise(root, title, stats, author) {
                     continue;
             }
         }
-        if (headline && normalized(text) === headline && (position < 4 || (heading && position < 14))) {
+        // A block that holds a picture is more than the repeated headline: the picture stays.
+        if (headline && normalized(text) === headline && (position < 4 || (heading && position < 14)) && !block.querySelector("img, picture, video")) {
             drop(block, "title-repeat");
             continue;
         }
@@ -320,6 +322,9 @@ function removeReaderNoise(root, title, stats, author) {
                 drop(next, "consent-placeholder");
         }
     });
+    // Last, so that nothing above takes the card for a repeated headline.
+    if (markVideoTeasers(root, title, stats))
+        changed = true;
     return changed;
 }
 const PRESENTATIONAL_TAGS = new Set(["FONT", "BIG", "SMALL", "CENTER", "TT", "U", "STRIKE", "S", "BASEFONT", "BLINK", "NOBR", "SPAN", "MARK", "PICTURE"]);

package/server/dist/sync/changeLog.js

@@ -1,4 +1,5 @@
 import { canAccessItem } from "../feeds/repository.js";
+import { isVideoTeaser } from "../feeds/content/noise/videoTeasers.js";
 import { db } from "../db/index.js";
 import { listFoldersForUser, listMutedKeywords, listSubscriptionsForUser, } from "../feeds/repository.js";
 import { getEdition, getEditionState } from "../native/editions.js";
@@ -63,20 +64,25 @@ function loadItemContents(userId, itemIds) {
                    ELSE 'snippet' END AS source,
               i.extraction_status,
               i.extraction_attempted_at,
-              i.extraction_retry_at
+              i.extraction_retry_at,
+              i.title,
+              CASE WHEN i.full_content_html IS NULL AND length(i.content_html) <= 2000 THEN i.content_html END AS short_html
        FROM items i
        WHERE i.id IN (SELECT CAST(value AS INTEGER) FROM json_each(?))`)
         .all(JSON.stringify(itemIds));
     for (const row of rows) {
         if (!canAccessItem(userId, row.id))
             continue;
+        // A video's page has no text to fetch: its card is the whole content (see `contentMetadata`).
+        const isVideo = isVideoTeaser(row.short_html, row.title);
         result.set(row.id, {
             revision: row.content_revision,
             source: row.source,
-            status: row.extraction_status,
+            status: isVideo ? "feed_only" : row.extraction_status,
             attemptedAt: row.extraction_attempted_at,
-            retryAt: row.extraction_retry_at,
+            retryAt: isVideo ? null : row.extraction_retry_at,
             isEmptyGallery: Boolean(row.is_empty_gallery),
+            ...(isVideo ? { isVideo: true } : {}),
         });
     }
     return result;

package/server/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@feedkeeper/server",
-  "version": "0.35.0",
+  "version": "0.35.2",
   "private": true,
   "type": "module"
 }