feedkeeper 0.28.1 → 0.28.3

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "feedkeeper",
-  "version": "0.28.1",
+  "version": "0.28.3",
   "description": "A lightweight, self-hosted RSS reader and remote MCP server for your news, blogs, and AI workflows.",
   "keywords": [
     "rss",

package/server/dist/feeds/contentNoise.js

@@ -143,6 +143,32 @@ function removeReaderNoise(root, title, stats, author) {
     const headline = normalized(title ?? "");
     const seen = new Set();
     const authorKeys = (author ?? "").split(",").map(name => normalized(name)).filter(Boolean);
+    const cleanByline = (block, text) => {
+        if (!BYLINE.test(text))
+            return;
+        if (authorKeys.length > 0 && authorKeys.every(name => normalized(text).includes(name))) {
+            drop(block, "byline");
+            return;
+        }
+        if (stripBylineDate(block)) {
+            stats["byline-date"] = (stats["byline-date"] ?? 0) + 1;
+            changed = true;
+        }
+    };
+    // Some publishers use an inline wrapper as a standalone byline. Never treat spans inside prose as metadata.
+    for (const span of root.querySelectorAll("span")) {
+        if (!root.contains(span) || span.closest(BLOCKS) || span.querySelector(`${BLOCKS}, div`))
+            continue;
+        const text = flat(span.textContent);
+        if (text.length > 60 || !BYLINE.test(text) || charsBefore(root, span) > 800)
+            continue;
+        if ([...(span.parentNode?.childNodes ?? [])].some(node => node !== span && node.nodeType === 3 && flat(node.textContent)))
+            continue;
+        if ([...(span.parentElement?.children ?? [])].some(sibling => sibling !== span &&
+            !sibling.matches(`${BLOCKS}, div, article, section, header, footer, figure, ul, ol, table, aside`) && flat(sibling.textContent)))
+            continue;
+        cleanByline(span, text);
+    }
     let leading = 0;
     let reachedText = false;
     let pastStart = false;
@@ -183,14 +209,9 @@ function removeReaderNoise(root, title, stats, author) {
                     continue;
                 }
                 // When the feed names the author, the reader shows the name in its head: a byline that only says that goes.
-                if (BYLINE.test(text) && authorKeys.length > 0 && authorKeys.every(name => normalized(text).includes(name))) {
-                    drop(block, "byline");
+                cleanByline(block, text);
+                if (!root.contains(block))
                     continue;
-                }
-                if (BYLINE.test(text) && stripBylineDate(block)) {
-                    stats["byline-date"] = (stats["byline-date"] ?? 0) + 1;
-                    changed = true;
-                }
             }
         }
         if (headline && normalized(text) === headline && (position < 4 || (heading && position < 14))) {

package/server/dist/feeds/discovery.js

@@ -11,7 +11,7 @@ const FEED_MIME_TYPES = [
     "application/xml",
 ];
 const MAX_PROBES = 15;
-const COMMON_PATHS = ["/feed", "/rss.xml", "/atom.xml", "/feed.xml", "/index.xml", "/rss", "/rss.php?feed=RSS2.0", "/rss.php?feed=ATOM1.0"];
+const COMMON_PATHS = ["/feed/", "/feed", "/rss.xml", "/atom.xml", "/feed.xml", "/index.xml", "/rss/", "/rss", "/rss.php?feed=RSS2.0", "/rss.php?feed=ATOM1.0"];
 // Feed-looking URLs anywhere in a page, including JSON embedded in attributes.
 const FEED_URL_IN_TEXT = /https?:\/\/[^\s"'<>\\]+?(?:\.rss|\.atom|\/feed\/?|\/(?:rss|atom|feed)\.xml)(?=["'\s<>\\]|$)/gi;
 function sameUrl(a, b) {
@@ -28,18 +28,23 @@ function sameUrl(a, b) {
     return normalize(a) === normalize(b);
 }
 async function parseFeedAt(url) {
-    const fetched = await fetchFeed(url);
-    const finalUrl = fetched.finalUrl || url;
     try {
-        const parsed = await rssParser.parseString(fetched.body);
-        if (parsed.items && (parsed.title || parsed.description || parsed.items.length > 0)) {
-            return { feed: { url: finalUrl, title: plainTitle(parsed.title) ?? null, type: "application/rss+xml" }, html: null, finalUrl };
+        const fetched = await fetchFeed(url);
+        const finalUrl = fetched.finalUrl || url;
+        try {
+            const parsed = await rssParser.parseString(fetched.body);
+            if (parsed.items && (parsed.title || parsed.description || parsed.items.length > 0)) {
+                return { feed: { url: finalUrl, title: plainTitle(parsed.title) ?? null, type: "application/rss+xml" }, html: null, finalUrl };
+            }
+        }
+        catch {
+            // Not a feed; the caller may look for feed links in the HTML instead.
         }
+        return { feed: null, html: fetched.body, finalUrl };
     }
     catch {
-        // Not a feed; the caller may look for feed links in the HTML instead.
+        return { feed: null, html: null, finalUrl: url };
     }
-    return { feed: null, html: fetched.body, finalUrl };
 }
 /**
  * Keep only candidates that really are feeds. Sites sometimes announce their own
@@ -75,22 +80,29 @@ export async function discoverFeeds(inputUrl) {
     const page = await parseFeedAt(targetUrl);
     if (page.feed)
         return [page.feed];
-    if (!page.html)
-        return [];
+    // 1b. If the URL did not end in a slash, try with a trailing slash (e.g. /feed -> /feed/)
+    if (!targetUrl.endsWith("/")) {
+        const slashCandidate = await parseFeedAt(targetUrl + "/");
+        if (slashCandidate.feed)
+            return [slashCandidate.feed];
+    }
     // 2. Feeds announced or linked on the page.
-    const fromPage = await keepRealFeeds(extractFeedsFromHtml(page.html, page.finalUrl), page.finalUrl);
-    if (fromPage.length > 0)
-        return fromPage;
+    if (page.html) {
+        const fromPage = await keepRealFeeds(extractFeedsFromHtml(page.html, page.finalUrl), page.finalUrl);
+        if (fromPage.length > 0)
+            return fromPage;
+    }
     // 3. Common feed paths; some lead to an overview page that lists the site's feeds.
+    const probeBase = page.finalUrl || targetUrl;
     for (const path of COMMON_PATHS) {
         try {
-            const candidateUrl = new URL(path, page.finalUrl).toString();
+            const candidateUrl = new URL(path, probeBase).toString();
             if (sameUrl(candidateUrl, targetUrl))
                 continue;
             const candidate = await parseFeedAt(candidateUrl);
             if (candidate.feed)
                 return [candidate.feed];
-            if (candidate.html && !sameUrl(candidate.finalUrl, page.finalUrl)) {
+            if (candidate.html && !sameUrl(candidate.finalUrl, probeBase)) {
                 const listed = await keepRealFeeds(extractFeedsFromHtml(candidate.html, candidate.finalUrl), candidate.finalUrl);
                 if (listed.length > 0)
                     return listed;

package/server/dist/feeds/service.js

@@ -64,7 +64,10 @@ export async function subscribeToFeed(userId, inputUrl, label = null, folderId =
     // (e.g. a homepage that announces itself as RSS); don't keep the subscription.
     if (pollResult.error && !pollResult.deferred && !findFeedById(feed.id)?.last_success_at) {
         unsubscribeRepo(userId, feed.id);
-        throw new NoFeedsFoundError();
+        if (pollResult.error === "feed_invalid_or_unavailable") {
+            throw new NoFeedsFoundError();
+        }
+        throw new FeedError(pollResult.error);
     }
     const subscription = findSubscriptionForUser(userId, feed.id);
     if (!subscription)

package/server/dist/native/editionContent.js

@@ -0,0 +1,32 @@
+import { JSDOM } from "jsdom";
+import { decodeEntities } from "../feeds/text.js";
+const PROMPTS = new Set([
+    "mehr", "weiterlesen", "mehr lesen", "read more", "more", "continue reading",
+    "video", "zum video", "video ansehen", "video abspielen", "watch video", "watch the video", "play video",
+    "audio", "zum audio", "audio anhören", "listen", "listen now", "listen to audio",
+    "続きを読む", "もっと読む", "動画を見る", "動画を再生", "音声を聞く",
+]);
+function normalize(text) {
+    return text.toLowerCase().replace(/[^\p{L}\p{N}]+/gu, " ").trim();
+}
+/** No length threshold: a short news sentence is useful, a media player or read-more link alone is not. */
+export function hasEditionText(item) {
+    const title = normalize(decodeEntities(item.title) ?? "");
+    return [item.content_snippet, item.content_html, item.full_content_html].some(source => {
+        if (!source)
+            return false;
+        const fragment = JSDOM.fragment(source);
+        fragment.querySelectorAll("script, style, noscript, nav, footer, video, audio, iframe, object, embed").forEach(node => node.remove());
+        fragment.querySelectorAll("a, p, span, h1, h2, h3, h4").forEach(node => {
+            const text = normalize(node.textContent ?? "");
+            if (text === title || PROMPTS.has(text))
+                node.remove();
+        });
+        fragment.querySelectorAll("p, div, br, li, h1, h2, h3, h4, blockquote, figcaption").forEach(node => node.append("\n"));
+        return (fragment.textContent ?? "").split(/\n/).some(line => {
+            const text = normalize(line);
+            return text !== "" && text !== title && !PROMPTS.has(text);
+        });
+    });
+}
+//# sourceMappingURL=editionContent.js.map

package/server/dist/native/editions.js

@@ -7,6 +7,7 @@ import { editionPeriod, selectEdition } from "./editionSelection.js";
 import { getNativePreferences } from "./preferences.js";
 import { checkEditionLayout, editionLayoutSchema } from "./editionLayout.js";
 import { NativeError, nativeRequest } from "./requests.js";
+import { hasEditionText } from "./editionContent.js";
 export const overviewSchema = z.array(z.object({
     heading: z.string().trim().min(1).max(160),
     segments: z.array(z.object({
@@ -109,10 +110,11 @@ export function publishCuratedEdition(userId, input, now = new Date()) {
             if (!canAccessItem(userId, id))
                 throw new NativeError("item_not_found", 404);
             const item = db.prepare(`SELECT s.folder_id, s.id IS NOT NULL AS subscribed, i.is_empty_gallery,
+          i.title, i.content_snippet, i.content_html, i.full_content_html,
           EXISTS (SELECT 1 FROM user_muted_keywords m WHERE m.user_id = ? AND
             (instr(lower(coalesce(i.title, '')), lower(m.keyword)) > 0 OR instr(lower(coalesce(i.content_snippet, '')), lower(m.keyword)) > 0)) AS muted
          FROM items i LEFT JOIN subscriptions s ON s.feed_id = i.feed_id AND s.user_id = ? WHERE i.id = ?`).get(userId, userId, id);
-            if (item.is_empty_gallery || item.muted || (item.subscribed && (item.folder_id === null ? !prefs.showUnfiled : prefs.hiddenFolderIds.includes(item.folder_id)))) {
+            if (item.is_empty_gallery || !hasEditionText(item) || item.muted || (item.subscribed && (item.folder_id === null ? !prefs.showUnfiled : prefs.hiddenFolderIds.includes(item.folder_id)))) {
                 throw new NativeError("item_excluded");
             }
         }
@@ -215,7 +217,7 @@ export function editionCandidates(userId, now = new Date(), limit = 200) {
     const previous = new Set(previousItemIds);
     // Fetch bounded candidates per source, so a busy feed cannot hide the others.
     const rows = db.prepare(`SELECT r.id, r.feed_id, r.folder_id, r.title, r.link, r.content_snippet, r.image_url,
-        r.source_name, r.folder_name, r.published, r.source_rank,
+        r.source_name, r.folder_name, r.published, r.source_rank, full_item.content_html, full_item.full_content_html,
         length(coalesce(full_item.full_content_html, full_item.content_html, r.content_snippet, '')) AS content_chars
       FROM (
         SELECT i.id, i.feed_id, s.folder_id, i.title, i.link, i.content_snippet, i.image_url,
@@ -233,7 +235,7 @@ export function editionCandidates(userId, now = new Date(), limit = 200) {
       JOIN items full_item ON full_item.id = r.id
       WHERE r.source_rank <= 40 ORDER BY r.source_rank, r.published DESC, r.id DESC LIMIT 4000`)
         .all(userId, userId, userId, new Date(now.getTime() - 7 * 86_400_000).toISOString(), now.toISOString());
-    const scored = rows.filter((row) => row.folder_id === null ? prefs.showUnfiled : !prefs.hiddenFolderIds.includes(row.folder_id))
+    const scored = rows.filter((row) => (row.folder_id === null ? prefs.showUnfiled : !prefs.hiddenFolderIds.includes(row.folder_id)) && hasEditionText(row))
         .map((row) => ({
         id: row.id, sourceName: row.source_name, folderName: row.folder_name, subscriptionId: row.feed_id, folderId: row.folder_id,
         title: row.title, url: row.link, publishedAt: row.published, snippet: decodeEntities(row.content_snippet) ?? null, imageUrl: row.image_url,

package/server/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@feedkeeper/server",
-  "version": "0.28.1",
+  "version": "0.28.3",
   "private": true,
   "type": "module"
 }