release: 0.69.1 (bug/feed-scroll-thumbnails-about)

VERSION

@@ -1 +1 @@
-0.69.0
+0.69.1

backend/app/sync/subscriptions.py

@@ -6,6 +6,7 @@ from sqlalchemy import select
 from sqlalchemy.orm import Session
 
 from app.models import Channel, Subscription, User
+from app.utils import normalize_description
 from app.youtube.client import YouTubeClient, best_thumbnail
 
 log = logging.getLogger("siftlode.sync")
@@ -55,7 +56,7 @@ def apply_channel_details(db: Session, items: list[dict]) -> None:
 
         channel.title = snippet.get("title") or channel.title
         channel.handle = snippet.get("customUrl")
-        channel.description = snippet.get("description")
+        channel.description = normalize_description(snippet.get("description"))
         channel.thumbnail_url = (
             best_thumbnail(snippet.get("thumbnails")) or channel.thumbnail_url
         )
@@ -104,7 +105,7 @@ def import_subscriptions(db: Session, user: User) -> dict:
                 channel = Channel(
                     id=cid,
                     title=sub.get("title"),
-                    description=sub.get("description"),
+                    description=normalize_description(sub.get("description")),
                     thumbnail_url=sub.get("thumbnail_url"),
                 )
                 db.add(channel)

backend/app/utils.py

@@ -36,3 +36,18 @@ def to_tsquery_str(q: str) -> str | None:
         return None
     tokens[-1] = tokens[-1] + ":*"
     return " & ".join(tokens)
+
+
+def normalize_description(text: str | None) -> str | None:
+    """Strip whitespace-only lines from a CHANNEL description before storing. YouTube channel
+    descriptions pad their contact/link blocks with runs of blank lines; keeping them wastes storage,
+    bloats the About popover, and adds nothing to the search document. Drops purely-empty lines and
+    right-strips each kept line; `str.splitlines()` also folds CR / CRLF and Unicode line separators
+    to "\\n" (a benign line-ending normalization). None stays None; all-blank collapses to "".
+
+    Deliberately CHANNEL-only — video descriptions are left with their original formatting: they are
+    user-facing prose where paragraph breaks carry meaning, unlike a channel's metadata block."""
+    if text is None:
+        return None
+    kept = [line.rstrip() for line in text.splitlines() if line.strip()]
+    return "\n".join(kept)

backend/tests/test_utils.py

@@ -1,7 +1,7 @@
 """iso(): the single date/datetime→JSON serializer the routes share (R6 S1a)."""
 from datetime import date, datetime, timezone
 
-from app.utils import iso
+from app.utils import iso, normalize_description
 
 
 def test_none_maps_to_none():
@@ -28,3 +28,30 @@ def test_matches_datetime_isoformat_contract():
     # hand-roll `x.isoformat() if x else None`.
     dt = datetime(2020, 1, 2, 3, 4, 5, 678901, tzinfo=timezone.utc)
     assert iso(dt) == dt.isoformat()
+
+
+def test_normalize_description_none_stays_none():
+    assert normalize_description(None) is None
+
+
+def test_normalize_description_drops_blank_lines():
+    # A YouTube contact block padded with runs of blank lines — the waste we strip before storing.
+    raw = "Contact us\n\n\nEmail: a@b.com\n \nInstagram: @x\n\n\n"
+    assert normalize_description(raw) == "Contact us\nEmail: a@b.com\nInstagram: @x"
+
+
+def test_normalize_description_right_strips_and_trims():
+    assert normalize_description("  hi there  \n\n  bye  ") == "  hi there\n  bye"
+
+
+def test_normalize_description_all_blank_collapses_to_empty():
+    assert normalize_description("\n \n\t\n") == ""
+
+
+def test_normalize_description_single_line_unchanged():
+    assert normalize_description("just one line") == "just one line"
+
+
+def test_normalize_description_folds_crlf_line_endings():
+    # splitlines() folds CR/CRLF to \n — a documented, benign line-ending normalization.
+    assert normalize_description("a\r\nb\r\n\r\nc") == "a\nb\nc"

frontend/src/components/AboutCell.tsx

@@ -28,6 +28,10 @@ export default function AboutCell({ text, channelId }: { text: string | null; ch
     offset: 6,
     // Pop instantly like the old cell did; the close grace lets the cursor reach the card.
     hoverDelay: { open: 0, close: 160 },
+    // Cap the panel to the space between the trigger and the viewport edge (publishes --fit-max-h) so a
+    // long About never runs its top off-screen — at high browser zoom the short viewport made a fixed
+    // max-height overflow above the top edge, where the scroll container can't reach.
+    fitHeight: true,
   });
 
   const { data: ch } = useQuery({
@@ -51,7 +55,10 @@ export default function AboutCell({ text, channelId }: { text: string | null; ch
       </span>
       <PopoverSurface
         popover={pop}
-        className="glass w-[480px] max-h-112 overflow-y-auto px-3.5 py-3 rounded-lg animate-[fadeIn_0.12s_ease]"
+        className="glass w-[480px] overflow-y-auto px-3.5 py-3 rounded-lg animate-[fadeIn_0.12s_ease]"
+        // min(design cap, available space): keeps the 28rem cap on a tall viewport, shrinks to fit on a
+        // short one (high zoom). --fit-max-h is set by the fitHeight `size` middleware once positioned.
+        style={{ maxHeight: "min(28rem, var(--fit-max-h, 28rem))" }}
       >
         {/* Full About once the detail loads; until then the list's truncated description. */}
         {ch ? (

frontend/src/components/AnchoredPopover.tsx

@@ -131,7 +131,14 @@ export function useAnchoredPopover(opts: AnchoredPopoverOptions) {
             size({
               padding: 8,
               apply({ availableHeight, elements }) {
+                // Cap the positioned element, AND publish the available space as a custom property so
+                // the INNER styled/scroll wrapper (PopoverSurface renders `className` on a child, not on
+                // the floating element) can cap itself too — e.g. `maxHeight: min(28rem, var(--fit-max-h))`
+                // keeps a caller's design cap while still shrinking to fit at high zoom / short viewports.
+                // Capping only the outer element would leave the inner overflow-y-auto box at its fixed
+                // height, spilling its top off-screen (the whole point of fitHeight).
                 elements.floating.style.maxHeight = `${availableHeight}px`;
+                elements.floating.style.setProperty("--fit-max-h", `${availableHeight}px`);
               },
             }),
           ]

frontend/src/components/DownloadCenter.tsx

@@ -84,10 +84,13 @@ function StatusPill({ job }: { job: DownloadJob }) {
 }
 
 function Thumb({ job }: { job: DownloadJob }) {
-  // Prefer the source thumbnail; fall back to our generated poster for thumbnail-less sources.
-  const primary = job.asset?.thumbnail_url || job.poster_url;
-  // A remote i.ytimg.com thumbnail 404s once the source video is deleted/privated, leaving a
-  // broken image. On load failure fall back to our always-present self-hosted poster.
+  // Prefer OUR self-hosted poster: it's always present on disk, never expires, and is same-origin, so
+  // a source-CDN signature expiry (Facebook's fbcdn URLs expire within days) or an ad-blocker/Shields
+  // rule on the source domain can't break it. The source's remote thumbnail_url is only a fallback for
+  // the rare poster-less asset (a thumbnail-less source whose ffmpeg frame-cut also failed).
+  const remote = job.asset?.thumbnail_url ?? null;
+  const primary = job.poster_url || remote;
+  // On load failure (a genuinely missing poster file) fall back to the remote source thumbnail.
   const [src, setSrc] = useState(primary);
   useEffect(() => setSrc(primary), [primary]);
   return (
@@ -99,7 +102,7 @@ function Thumb({ job }: { job: DownloadJob }) {
           loading="lazy"
           className="w-full h-full object-cover"
           onError={() => {
-            if (job.poster_url && src !== job.poster_url) setSrc(job.poster_url);
+            if (remote && src !== remote) setSrc(remote);
           }}
         />
       ) : null}

frontend/src/components/VirtualGrid.tsx

@@ -8,6 +8,7 @@ import {
   type ReactNode,
 } from "react";
 import { useVirtualizer } from "@tanstack/react-virtual";
+import { shouldPrefetchNext } from "./readAhead";
 
 // The virtualization engine, lifted out of VirtualFeed so any long client- or server-side list can
 // reuse it (the feed's video grid, the channel manager's card/row views). It owns ONLY the layout
@@ -23,9 +24,6 @@ import { useVirtualizer } from "@tanstack/react-virtual";
 // (grid-cols-[repeat(auto-fill,minmax(260px,1fr))] with a gap-4 / 1rem gutter).
 const GRID_GAP = 16;
 const LIST_GAP = 4;
-// Start fetching the next page this many rows before the end scrolls into view
-// (keeps the old IntersectionObserver's generous prefetch feel).
-const PREFETCH_ROWS = 4;
 
 /** Columns that fit `width`, or 1 for the single-column layout (`minCol == null`). */
 function columnsFor(minCol: number | null, width: number): number {
@@ -140,13 +138,17 @@ export default function VirtualGrid<T>({
 
   const virtualRows = virtualizer.getVirtualItems();
 
-  // Infinite-scroll trigger: once the last rendered row is within PREFETCH_ROWS of the end, pull
-  // the next page. A no-op for finite lists (no fetchNextPage / hasNextPage false).
+  // Infinite-scroll read-ahead: pull the next page once the last rendered row is within the
+  // read-ahead window of the loaded end (see readAhead.ts — the window scales with the rendered row
+  // count, clamped so a tall window doesn't prefetch unboundedly). A no-op for finite lists (no
+  // fetchNextPage / hasNextPage false). Because this effect re-runs when a page settles
+  // (isFetchingNextPage flips false) it CHAINS — refilling the buffer as the user consumes it, so a
+  // fast scroll reaches the loaded end far less often.
   useEffect(() => {
-    if (!fetchNextPage) return;
+    if (!fetchNextPage || !hasNextPage || isFetchingNextPage) return;
     const last = virtualRows[virtualRows.length - 1];
     if (!last) return;
-    if (last.index >= rows.length - 1 - PREFETCH_ROWS && hasNextPage && !isFetchingNextPage) {
+    if (shouldPrefetchNext(last.index, rows.length, virtualRows.length)) {
       fetchNextPage();
     }
   }, [virtualRows, rows.length, hasNextPage, isFetchingNextPage, fetchNextPage]);

frontend/src/components/readAhead.test.ts

@@ -0,0 +1,44 @@
+import { describe, expect, it } from "vitest";
+import {
+  MAX_LOOKAHEAD_ROWS,
+  PREFETCH_ROWS,
+  PREFETCH_VIEWPORTS,
+  readAheadLookahead,
+  shouldPrefetchNext,
+} from "./readAhead";
+
+describe("readAheadLookahead", () => {
+  it("floors at PREFETCH_ROWS when nothing (or little) is rendered", () => {
+    expect(readAheadLookahead(0)).toBe(PREFETCH_ROWS);
+    expect(readAheadLookahead(1)).toBe(PREFETCH_ROWS); // 1*2=2 < floor 4
+  });
+
+  it("scales with the rendered row count for a normal viewport", () => {
+    expect(readAheadLookahead(10)).toBe(10 * PREFETCH_VIEWPORTS); // 20
+    expect(readAheadLookahead(15)).toBe(30);
+  });
+
+  it("caps at MAX_LOOKAHEAD_ROWS so a tall window doesn't prefetch unboundedly", () => {
+    // 30 rendered rows would be 60 uncapped — the cap is what bounds per-mount feed prefetch.
+    expect(readAheadLookahead(30)).toBe(MAX_LOOKAHEAD_ROWS);
+    expect(readAheadLookahead(1000)).toBe(MAX_LOOKAHEAD_ROWS);
+  });
+});
+
+describe("shouldPrefetchNext", () => {
+  it("fetches once the last rendered row enters the read-ahead window", () => {
+    // 100 rows loaded, viewport renders 10 → lookahead 20. Fetch when lastIndex >= 100-1-20 = 79.
+    expect(shouldPrefetchNext(78, 100, 10)).toBe(false);
+    expect(shouldPrefetchNext(79, 100, 10)).toBe(true);
+    expect(shouldPrefetchNext(99, 100, 10)).toBe(true);
+  });
+
+  it("fetches eagerly while fewer rows are loaded than the lookahead (early pages)", () => {
+    // Only 15 rows loaded, lookahead 20 → the whole list is inside the window.
+    expect(shouldPrefetchNext(8, 15, 10)).toBe(true);
+  });
+
+  it("does not fetch near the top of a long loaded list", () => {
+    expect(shouldPrefetchNext(12, 500, 10)).toBe(false);
+  });
+});

frontend/src/components/readAhead.ts

@@ -0,0 +1,30 @@
+// Infinite-scroll read-ahead policy for VirtualGrid, extracted as a pure function so the decision is
+// unit-testable without a layout engine.
+//
+// The next page is pulled once the last rendered row comes within `lookahead` rows of the loaded end,
+// where lookahead scales with the number of rows the viewport currently renders (so it adapts to
+// window height / column count / row height) but is clamped to a floor and a ceiling:
+//   - floor PREFETCH_ROWS: the finite-list / first-paint case where nothing is rendered yet.
+//   - ceiling MAX_LOOKAHEAD_ROWS: a tall window renders many rows, and an uncapped lookahead would
+//     eagerly prefetch several pages on every mount / filter change — wasted work on the app's
+//     heaviest query (the feed). The cap bounds that while still buffering a couple of screens ahead.
+export const PREFETCH_ROWS = 4;
+export const PREFETCH_VIEWPORTS = 2;
+export const MAX_LOOKAHEAD_ROWS = 40;
+
+/** How many rows ahead of the last rendered row to keep buffered, for a viewport rendering
+ *  `renderedRows` rows. Clamped to [PREFETCH_ROWS, MAX_LOOKAHEAD_ROWS]. */
+export function readAheadLookahead(renderedRows: number): number {
+  return Math.min(MAX_LOOKAHEAD_ROWS, Math.max(PREFETCH_ROWS, renderedRows * PREFETCH_VIEWPORTS));
+}
+
+/** Whether to fetch the next page: true once the last rendered row (`lastIndex`) is within the
+ *  read-ahead window of the loaded end (`rowCount` total loaded rows). Because the caller re-runs
+ *  this as each page settles, it chains — refilling the buffer as the user consumes it. */
+export function shouldPrefetchNext(
+  lastIndex: number,
+  rowCount: number,
+  renderedRows: number,
+): boolean {
+  return lastIndex >= rowCount - 1 - readAheadLookahead(renderedRows);
+}

frontend/src/lib/releaseNotes.ts

@@ -14,6 +14,18 @@ export interface ReleaseEntry {
 }
 
 export const RELEASE_NOTES: ReleaseEntry[] = [
+  {
+    version: "0.69.1",
+    date: "2026-08-19",
+    summary:
+      "Smoother fast-scrolling in the feed, reliable download thumbnails, and a tidier channel About popover.",
+    fixes: [
+      "Download thumbnails are reliable again. The Library now shows the copy saved on the server instead of the source's own link — which for Facebook videos expires within days (and can be blocked by a browser tracker shield), leaving a blank thumbnail even though the image was on disk the whole time.",
+      "Fast-scrolling the feed no longer stalls. A quick flick of the scroll wheel used to outrun the loader and leave you waiting several seconds on “Load more”; the feed now keeps a couple of screens of videos loaded ahead, so the content is usually already there when you reach it.",
+      "The channel About popover stays on screen at higher browser zoom. A long description could push the top of the box above the edge of the window where you couldn't read it; it now shrinks to fit the available height and scrolls inside.",
+      "Channel descriptions are tidier: the runs of blank lines YouTube pads its contact/link blocks with are stripped, so the About box is more compact.",
+    ],
+  },
   {
     version: "0.69.0",
     date: "2026-08-19",