Capture YouTube titles from Discord's embed cache

poc/public/media.css

@@ -885,3 +885,9 @@
     font: 400 11px/1.5 ui-monospace, Menlo, monospace;
     color: var(--color-neutral-500);
 }
+
+.vk-media-viewer-title {
+    font: 600 13px/1.5 ui-monospace, Menlo, monospace;
+    color: var(--color-text);
+    margin-bottom: 6px;
+}

poc/public/media.js

@@ -2669,6 +2669,18 @@
         }
 
         meta.replaceChildren();
+
+        // The clip title matters most for finding footage; show it big when we have one.
+        const heading =
+            item.title || (item.titleError && (item.provider ?? "").toLowerCase().includes("youtube"))
+                ? item.title
+                    ? `${item.title}${item.titleAuthor ? ` — ${item.titleAuthor}` : ""}`
+                    : "Video unavailable (removed or private)"
+                : "";
+        if (heading) {
+            meta.append(el("div", "vk-media-viewer-title", heading));
+        }
+
         const values = metaValues(item);
         for (const [key, label] of META_FIELDS) {
             if (!values[key]) continue;

poc/src/media.ts

@@ -136,6 +136,8 @@ export interface MediaItemResult {
     remotePreviewUrl: string | null;
     /** Why no preview exists, when a fetch failed (for example "http 404"). */
     previewError: string | null;
+    /** Why no title was resolved (provider oEmbed and Discord's embed both failed). */
+    titleError: string | null;
     /** Local animated GIF for hover playback, when the original was cached. */
     animUrl: string | null;
     localUrl: string | null;
@@ -217,6 +219,7 @@ export class MediaService {
             maxRetries: tunables.scrape.maxRetries,
             previewMaxAttempts: tunables.backfill.maxAttempts,
             suspectRetryMs: 50_000,
+            embedMeta: async (scrapeId, media) => this.embedMetaFor(scrapeId, media),
             refreshUrl: async (scrapeId, media) => {
                 const meta = this.engine.get(scrapeId);
                 if (!meta) return null;
@@ -683,6 +686,67 @@ export class MediaService {
         this.engine.stop(id);
     }
 
+    /**
+     * Title and channel from the source message's Discord embed. Discord keeps the embed data
+     * after a video is deleted, so this is the only title source for gone YouTube videos.
+     */
+    private async embedMetaFor(
+        scrapeId: string,
+        media: { id: number; url: string; playUrl: string | null }
+    ): Promise<{ title: string; author: string | null } | null> {
+        const meta = this.engine.get(scrapeId);
+        if (!meta) return null;
+
+        try {
+            const store = new ScrapeStore(scrapeDir(this.rootDir, scrapeId));
+            let messageId: string | null = null;
+            try {
+                messageId = store.sourceMessageId(media.id);
+            } finally {
+                store.close();
+            }
+            if (!messageId) return null;
+
+            const message = await this.fetchSourceMessage(meta, messageId);
+            if (!message) return null;
+
+            const path = (url: string) => url.split("?")[0];
+            const embeds: any[] = message.embeds ?? [];
+            const embed = embeds.find(
+                candidate =>
+                    (candidate?.url && path(candidate.url) === path(media.url)) ||
+                    (media.playUrl && candidate?.videoUrl && path(candidate.videoUrl) === path(media.playUrl)) ||
+                    (candidate?.video?.url && media.playUrl && path(candidate.video.url) === path(media.playUrl))
+            );
+
+            const title = embed?.title ?? null;
+            if (!title) return null;
+            const author = embed?.authorName ?? embed?.author?.name ?? null;
+            return { title, author };
+        } catch (error) {
+            this.deps.log?.(`embed meta failed for ${scrapeId}:${media.id}: ${String(error)}`);
+            return null;
+        }
+    }
+
+    /** The source message through slopbot REST (guild scrapes) or the puppet bridge. */
+    private async fetchSourceMessage(meta: ScrapeMeta, messageId: string): Promise<any | null> {
+        if (this.bot && meta.guildId && meta.backend === "slopbot") {
+            const channel = await this.bot.channels.fetch(meta.channelId);
+            if (channel?.isTextBased()) return await channel.messages.fetch(messageId);
+            return null;
+        }
+
+        const core = this.deps.getCore();
+        if (!core || !this.deps.isCoreConnected()) return null;
+
+        const result = (await core.fetchMessages(meta.channelId, 1, {
+            before: (BigInt(messageId) + 1n).toString()
+        })) as { messages?: Array<Record<string, any>> };
+
+        return result?.messages?.find(candidate => candidate?.id === messageId) ?? result?.messages?.[0] ?? null;
+    }
+
     /** Continue a paused scrape from its cursor with the same backend family. */
     async resumeScrape(scrapeId: string): Promise<ScrapeMeta | null> {
         const meta = this.engine.get(scrapeId);
@@ -1224,6 +1288,7 @@ export class MediaService {
                 playUrl: row.play_url,
                 remotePreviewUrl: row.preview_url,
                 previewError: row.preview_error,
+                titleError: row.title_error,
                 animUrl: row.anim_hash ? `/api/media/${meta.id}/${row.id}/anim` : null,
                 localUrl:
                     row.file_path && !meta.downloadDir