Audible: HTML refresh, multi-narrator & works dedup

Switch nightly discovery refresh to scrape Audible's curated HTML storefronts (popular, new releases, category pages) while keeping real-time user paths on the JSON catalog API. Add robust HTML resilience knobs (increased retries, capped jittered backoff, AdaptivePacer changes and per-batch cooldowns) to avoid failing nightly jobs during 503 storms. Implement multi-narrator capture via a new extractAllNarrators helper and update parsers to preserve all narrator anchors. Introduce two-pass dedup: in-memory deduplicateAndCollectGroups + collapseByExistingWorks that consults the works table, export metadataScore for consistent representative selection, and persist dedup groups (fire-and-forget). Wire collapseByExistingWorks into search/author/series routes and make defensive dedup in the refresh processor. Add HTML parsing helpers, runtime/lang-aware parsing, jitteredBackoff cap, and tests for the new behaviors.

src/lib/utils/deduplicate-audiobooks.ts

@@ -109,7 +109,12 @@ export function areDurationsCompatible(a?: number, b?: number): boolean {
 // Metadata scoring (for picking best representative)
 // ---------------------------------------------------------------------------
 
-function metadataScore(book: AudibleAudiobook): number {
+/**
+ * Score a book by how much metadata it carries. Used as the tie-breaker when
+ * collapsing duplicates — the entry with the richest metadata wins. Exported
+ * so the works-table collapse pass can apply the same ranking.
+ */
+export function metadataScore(book: AudibleAudiobook): number {
   let score = 0;
   if (book.coverArtUrl) score++;
   if (book.rating != null) score++;

src/lib/utils/extract-narrator.ts

@@ -0,0 +1,37 @@
+/**
+ * Component: Narrator Extraction Utility
+ * Documentation: documentation/integrations/audible.md
+ *
+ * Shared helper for Audible HTML scrapers. Audible product listings render
+ * each narrator as a separate `<a href="?searchNarrator=...">` link; using
+ * `.first()` on that selector silently drops co-narrators and breaks dedup
+ * for multi-narrator productions (e.g. full-cast audiobooks). This helper
+ * captures every narrator link and joins them, falling back to the
+ * `.narratorLabel` span when no anchor links are present.
+ */
+
+import type * as cheerio from 'cheerio';
+import type { AnyNode } from 'domhandler';
+
+/**
+ * Extract a comma-joined narrator string from an Audible product list item.
+ *
+ * Order is not semantically significant — downstream `normalizeNarrator()`
+ * sorts before comparison — but document-order preserves a stable, legible
+ * value for caching and logging.
+ */
+export function extractAllNarrators(
+  $: cheerio.CheerioAPI,
+  $el: cheerio.Cheerio<AnyNode>,
+): string {
+  const links = $el.find('a[href*="searchNarrator="]');
+  if (links.length > 0) {
+    const names: string[] = [];
+    links.each((_, link) => {
+      const name = $(link).text().trim();
+      if (name) names.push(name);
+    });
+    if (names.length > 0) return names.join(', ');
+  }
+  return $el.find('.narratorLabel').text().trim();
+}

src/lib/utils/scrape-resilience.ts

@@ -38,12 +38,18 @@ export function getBrowserHeaders(userAgent: string): Record<string, string> {
 }
 
 /**
- * Jittered exponential backoff: 2^attempt * baseMs * random(0.5, 1.5)
+ * Jittered exponential backoff: 2^attempt * baseMs * random(0.5, 1.5),
+ * optionally capped so high attempt counts don't produce absurd waits.
  * Avoids predictable retry timing that is trivially fingerprinted.
  */
-export function jitteredBackoff(attempt: number, baseMs: number = 1000): number {
+export function jitteredBackoff(
+  attempt: number,
+  baseMs: number = 1000,
+  maxBackoffMs: number = Number.POSITIVE_INFINITY,
+): number {
   const jitter = 0.5 + Math.random(); // 0.5 – 1.5
-  return Math.round(Math.pow(2, attempt) * baseMs * jitter);
+  const raw = Math.pow(2, attempt) * baseMs * jitter;
+  return Math.round(Math.min(raw, maxBackoffMs));
 }
 
 /** Random integer in [minMs, maxMs] */