@raviolelabs/engram-mcp 0.2.0

package/dist/memory/modules/youtube/tools.js

@@ -0,0 +1,182 @@
+// src/memory/modules/youtube/tools.ts
+import { createLogger } from '../../../logger.js';
+import { fetchTranscript } from './transcript-fetcher.js';
+import { buildYoutubeItem } from './ingest.js';
+import { sourceRegistry } from '../../core/source-registry.js';
+const log = createLogger('youtube:tools');
+export function buildYoutubeTools(store, config) {
+    const embeddingModel = `${config.embeddings.provider}/${config.embeddings.model}`;
+    return [
+        {
+            name: 'add_youtube_url',
+            description: 'Fetch the transcript of a YouTube video and add it to memory. Works for any public video with captions (auto-generated or human). The video title from the page is used automatically as the title field. You (the calling LLM) should provide additional context like tags when calling suggest_properties afterwards.',
+            inputSchema: {
+                type: 'object',
+                properties: {
+                    url: {
+                        type: 'string',
+                        description: 'YouTube watch URL or video id.',
+                    },
+                },
+                required: ['url'],
+            },
+            handler: async (args) => {
+                const url = args.url;
+                const transcript = await fetchTranscript(url, config.youtube);
+                const item = buildYoutubeItem({ transcript, embeddingModel });
+                await store.insert(item);
+                log.info(`Ingested YouTube ${transcript.title} (${transcript.segments.length} segments) as ${item.id}`);
+                return {
+                    id: item.id,
+                    title: transcript.title,
+                    channel: transcript.channel,
+                    segments: transcript.segments.length,
+                };
+            },
+        },
+        {
+            name: 'search_youtube',
+            description: 'Search ingested YouTube transcripts by semantic similarity. Returns video metadata + transcript snippets. Use this to find previously ingested YouTube videos by content.',
+            inputSchema: {
+                type: 'object',
+                properties: {
+                    query: { type: 'string', description: 'Natural language query to search YouTube transcripts.' },
+                    limit: { type: 'number', default: 10, description: 'Maximum number of results to return.' },
+                },
+                required: ['query'],
+            },
+            handler: async (args) => {
+                const hits = await store.search('youtube', args.query, args.limit ?? 10);
+                return hits.map((h) => ({
+                    id: h.memory.id,
+                    score: h.score,
+                    snippet: h.snippet,
+                    title: h.memory.properties.title,
+                    channel: h.memory.properties.author,
+                    source_url: h.memory.properties.source_url,
+                }));
+            },
+        },
+        // ── watch_youtube_channel ───────────────────────────────────────────────
+        {
+            name: 'watch_youtube_channel',
+            description: 'Subscribe to a YouTube channel. EngramMCP will poll the channel RSS feed periodically ' +
+                '(default 6h, configurable via youtube.channelPollIntervalMs in ~/.engram/config.json) ' +
+                'and automatically ingest new videos as YouTube memories. ' +
+                'Call with the channel URL (e.g. https://www.youtube.com/@handle) or channel ID (UCxxx). ' +
+                'Use this when a user says "watch this channel" or "keep track of new videos from X". ' +
+                'Providing a descriptive channelName improves search retrieval.',
+            inputSchema: {
+                type: 'object',
+                properties: {
+                    channelId: {
+                        type: 'string',
+                        description: 'YouTube channel ID (starts with UC…) OR full channel URL (e.g. https://www.youtube.com/@handle). Required.',
+                    },
+                    channelName: {
+                        type: 'string',
+                        description: 'Human-readable channel name. Used as a tag on ingested memories.',
+                    },
+                },
+                required: ['channelId'],
+            },
+            handler: async (args) => {
+                const { channelId: rawChannelId, channelName: rawChannelName } = args;
+                const { resolveChannelId } = await import('./watcher.js');
+                const channelId = await resolveChannelId(rawChannelId);
+                const existing = sourceRegistry.listEnabled('youtube').find((s) => s.external_id === channelId);
+                if (existing) {
+                    return {
+                        content: [
+                            { type: 'text', text: `Channel ${channelId} is already being watched.` },
+                        ],
+                    };
+                }
+                sourceRegistry.add({
+                    module_id: 'youtube',
+                    external_id: channelId,
+                    display_name: rawChannelName ?? channelId,
+                    config: { channelId, channelName: rawChannelName ?? channelId },
+                });
+                return {
+                    content: [
+                        {
+                            type: 'text',
+                            text: `Now watching channel ${rawChannelName ?? channelId} (${channelId}). New videos will be ingested automatically.`,
+                        },
+                    ],
+                };
+            },
+        },
+        // ── unwatch_youtube_channel ─────────────────────────────────────────────
+        {
+            name: 'unwatch_youtube_channel',
+            description: 'Stop watching a YouTube channel. Future videos will not be ingested. Existing memories are kept.',
+            inputSchema: {
+                type: 'object',
+                properties: {
+                    channelId: { type: 'string', description: 'Channel ID to unwatch (UCxxx).' },
+                },
+                required: ['channelId'],
+            },
+            handler: async (args) => {
+                const { channelId } = args;
+                const sources = sourceRegistry.list('youtube');
+                const found = sources.find((s) => s.external_id === channelId);
+                if (found) {
+                    sourceRegistry.remove(found.id);
+                    return {
+                        content: [
+                            { type: 'text', text: `Stopped watching channel ${channelId}.` },
+                        ],
+                    };
+                }
+                return {
+                    content: [
+                        {
+                            type: 'text',
+                            text: `Channel ${channelId} was not in the watch list.`,
+                        },
+                    ],
+                };
+            },
+        },
+        // ── import_watch_later ──────────────────────────────────────────────────
+        {
+            name: 'import_watch_later',
+            description: 'Bulk-import a YouTube playlist (paste the playlist URL). Fetches video IDs from the playlist page HTML and ingests each video as a YouTube memory. ' +
+                'Note: Watch Later (WL) is a private playlist — the user must make it temporarily public or use a different playlist URL. ' +
+                'Use this when a user says "import my watch later" or "ingest this playlist".',
+            inputSchema: {
+                type: 'object',
+                properties: {
+                    playlistUrl: {
+                        type: 'string',
+                        description: 'Full YouTube playlist URL, e.g. https://www.youtube.com/playlist?list=PLxxxxxx',
+                    },
+                    limit: {
+                        type: 'number',
+                        description: 'Maximum number of videos to import (default 50, max 200). Older videos are skipped first.',
+                    },
+                },
+                required: ['playlistUrl'],
+            },
+            handler: async (args) => {
+                const { playlistUrl, limit: rawLimit } = args;
+                const { importPlaylist } = await import('./watcher.js');
+                const result = await importPlaylist(playlistUrl, store, config.embeddings, config.youtube, rawLimit ?? 50);
+                return {
+                    content: [
+                        {
+                            type: 'text',
+                            text: result.imported > 0
+                                ? `Imported ${result.imported} videos from the playlist. ${result.skipped} skipped (already ingested or failed).`
+                                : `No new videos found in the playlist. Either the playlist is empty, private, or all videos were already ingested.`,
+                        },
+                    ],
+                };
+            },
+        },
+    ];
+}
+//# sourceMappingURL=tools.js.map

package/dist/memory/modules/youtube/tools.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"tools.js","sourceRoot":"","sources":["../../../../src/memory/modules/youtube/tools.ts"],"names":[],"mappings":"AAAA,sCAAsC;AACtC,OAAO,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAIlD,OAAO,EAAE,eAAe,EAAE,MAAM,yBAAyB,CAAC;AAC1D,OAAO,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAC/C,OAAO,EAAE,cAAc,EAAE,MAAM,+BAA+B,CAAC;AAE/D,MAAM,GAAG,GAAG,YAAY,CAAC,eAAe,CAAC,CAAC;AAE1C,MAAM,UAAU,iBAAiB,CAAC,KAAkB,EAAE,MAAoB;IACxE,MAAM,cAAc,GAAG,GAAG,MAAM,CAAC,UAAU,CAAC,QAAQ,IAAI,MAAM,CAAC,UAAU,CAAC,KAAK,EAAE,CAAC;IAElF,OAAO;QACL;YACE,IAAI,EAAE,iBAAiB;YACvB,WAAW,EACT,yTAAyT;YAC3T,WAAW,EAAE;gBACX,IAAI,EAAE,QAAQ;gBACd,UAAU,EAAE;oBACV,GAAG,EAAE;wBACH,IAAI,EAAE,QAAQ;wBACd,WAAW,EAAE,gCAAgC;qBAC9C;iBACF;gBACD,QAAQ,EAAE,CAAC,KAAK,CAAC;aAClB;YACD,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,EAAE;gBACtB,MAAM,GAAG,GAAG,IAAI,CAAC,GAAa,CAAC;gBAC/B,MAAM,UAAU,GAAG,MAAM,eAAe,CAAC,GAAG,EAAE,MAAM,CAAC,OAAO,CAAC,CAAC;gBAC9D,MAAM,IAAI,GAAG,gBAAgB,CAAC,EAAE,UAAU,EAAE,cAAc,EAAE,CAAC,CAAC;gBAC9D,MAAM,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;gBACzB,GAAG,CAAC,IAAI,CACN,oBAAoB,UAAU,CAAC,KAAK,KAAK,UAAU,CAAC,QAAQ,CAAC,MAAM,iBAAiB,IAAI,CAAC,EAAE,EAAE,CAC9F,CAAC;gBACF,OAAO;oBACL,EAAE,EAAE,IAAI,CAAC,EAAE;oBACX,KAAK,EAAE,UAAU,CAAC,KAAK;oBACvB,OAAO,EAAE,UAAU,CAAC,OAAO;oBAC3B,QAAQ,EAAE,UAAU,CAAC,QAAQ,CAAC,MAAM;iBACrC,CAAC;YACJ,CAAC;SACF;QACD;YACE,IAAI,EAAE,gBAAgB;YACtB,WAAW,EACT,2KAA2K;YAC7K,WAAW,EAAE;gBACX,IAAI,EAAE,QAAQ;gBACd,UAAU,EAAE;oBACV,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,uDAAuD,EAAE;oBAC/F,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,EAAE,EAAE,EAAE,WAAW,EAAE,sCAAsC,EAAE;iBAC5F;gBACD,QAAQ,EAAE,CAAC,OAAO,CAAC;aACpB;YACD,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,EAAE;gBACtB,MAAM,IAAI,GAAG,MAAM,KAAK,CAAC,MAAM,CAC7B,SAAS,EACT,IAAI,CAAC,KAAe,EACnB,IAAI,CAAC,KAAgB,IAAI,EAAE,CAC7B,CAAC;gBACF,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;oBACtB,EAAE,EAAE,CAAC,CAAC,MAAM,CAAC,EAAE;oBACf,KAAK,EAAE,CAAC,CAAC,KAAK;oBACd,OAAO,EAAE,CAAC,CAAC,OAAO;oBAClB,KAAK,EAAE,CAAC,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK;oBAChC,OAAO,EAAE,CAAC,CAAC,MAAM,CAAC,UAAU,CAAC,MAAM;oBACnC,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC,UAAU,CAAC,UAAU;iBAC3C,CAAC,CAAC,CAAC;YACN,CAAC;SACF;QAED,2EAA2E;QAC3E;YACE,IAAI,EAAE,uBAAuB;YAC7B,WAAW,EACT,wFAAwF;gBACxF,wFAAwF;gBACxF,2DAA2D;gBAC3D,0FAA0F;gBAC1F,uFAAuF;gBACvF,gEAAgE;YAClE,WAAW,EAAE;gBACX,IAAI,EAAE,QAAQ;gBACd,UAAU,EAAE;oBACV,SAAS,EAAE;wBACT,IAAI,EAAE,QAAQ;wBACd,WAAW,EACT,4GAA4G;qBAC/G;oBACD,WAAW,EAAE;wBACX,IAAI,EAAE,QAAQ;wBACd,WAAW,EAAE,kEAAkE;qBAChF;iBACF;gBACD,QAAQ,EAAE,CAAC,WAAW,CAAC;aACxB;YACD,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,EAAE;gBACtB,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,WAAW,EAAE,cAAc,EAAE,GAAG,IAGhE,CAAC;gBACF,MAAM,EAAE,gBAAgB,EAAE,GAAG,MAAM,MAAM,CAAC,cAAc,CAAC,CAAC;gBAE1D,MAAM,SAAS,GAAG,MAAM,gBAAgB,CAAC,YAAY,CAAC,CAAC;gBACvD,MAAM,QAAQ,GAAG,cAAc,CAAC,WAAW,CAAC,SAAS,CAAC,CAAC,IAAI,CACzD,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,WAAW,KAAK,SAAS,CACnC,CAAC;gBACF,IAAI,QAAQ,EAAE,CAAC;oBACb,OAAO;wBACL,OAAO,EAAE;4BACP,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,WAAW,SAAS,4BAA4B,EAAE;yBACzE;qBACF,CAAC;gBACJ,CAAC;gBAED,cAAc,CAAC,GAAG,CAAC;oBACjB,SAAS,EAAE,SAAS;oBACpB,WAAW,EAAE,SAAS;oBACtB,YAAY,EAAE,cAAc,IAAI,SAAS;oBACzC,MAAM,EAAE,EAAE,SAAS,EAAE,WAAW,EAAE,cAAc,IAAI,SAAS,EAAE;iBAChE,CAAC,CAAC;gBAEH,OAAO;oBACL,OAAO,EAAE;wBACP;4BACE,IAAI,EAAE,MAAM;4BACZ,IAAI,EAAE,wBAAwB,cAAc,IAAI,SAAS,KAAK,SAAS,+CAA+C;yBACvH;qBACF;iBACF,CAAC;YACJ,CAAC;SACF;QAED,2EAA2E;QAC3E;YACE,IAAI,EAAE,yBAAyB;YAC/B,WAAW,EACT,kGAAkG;YACpG,WAAW,EAAE;gBACX,IAAI,EAAE,QAAQ;gBACd,UAAU,EAAE;oBACV,SAAS,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,gCAAgC,EAAE;iBAC7E;gBACD,QAAQ,EAAE,CAAC,WAAW,CAAC;aACxB;YACD,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,EAAE;gBACtB,MAAM,EAAE,SAAS,EAAE,GAAG,IAA6B,CAAC;gBACpD,MAAM,OAAO,GAAG,cAAc,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;gBAC/C,MAAM,KAAK,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,WAAW,KAAK,SAAS,CAAC,CAAC;gBAC/D,IAAI,KAAK,EAAE,CAAC;oBACV,cAAc,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;oBAChC,OAAO;wBACL,OAAO,EAAE;4BACP,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,4BAA4B,SAAS,GAAG,EAAE;yBACjE;qBACF,CAAC;gBACJ,CAAC;gBACD,OAAO;oBACL,OAAO,EAAE;wBACP;4BACE,IAAI,EAAE,MAAM;4BACZ,IAAI,EAAE,WAAW,SAAS,6BAA6B;yBACxD;qBACF;iBACF,CAAC;YACJ,CAAC;SACF;QAED,2EAA2E;QAC3E;YACE,IAAI,EAAE,oBAAoB;YAC1B,WAAW,EACT,qJAAqJ;gBACrJ,2HAA2H;gBAC3H,8EAA8E;YAChF,WAAW,EAAE;gBACX,IAAI,EAAE,QAAQ;gBACd,UAAU,EAAE;oBACV,WAAW,EAAE;wBACX,IAAI,EAAE,QAAQ;wBACd,WAAW,EACT,gFAAgF;qBACnF;oBACD,KAAK,EAAE;wBACL,IAAI,EAAE,QAAQ;wBACd,WAAW,EACT,2FAA2F;qBAC9F;iBACF;gBACD,QAAQ,EAAE,CAAC,aAAa,CAAC;aAC1B;YACD,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,EAAE;gBACtB,MAAM,EAAE,WAAW,EAAE,KAAK,EAAE,QAAQ,EAAE,GAAG,IAA+C,CAAC;gBACzF,MAAM,EAAE,cAAc,EAAE,GAAG,MAAM,MAAM,CAAC,cAAc,CAAC,CAAC;gBACxD,MAAM,MAAM,GAAG,MAAM,cAAc,CACjC,WAAW,EACX,KAAK,EACL,MAAM,CAAC,UAAU,EACjB,MAAM,CAAC,OAAO,EACd,QAAQ,IAAI,EAAE,CACf,CAAC;gBACF,OAAO;oBACL,OAAO,EAAE;wBACP;4BACE,IAAI,EAAE,MAAM;4BACZ,IAAI,EACF,MAAM,CAAC,QAAQ,GAAG,CAAC;gCACjB,CAAC,CAAC,YAAY,MAAM,CAAC,QAAQ,8BAA8B,MAAM,CAAC,OAAO,wCAAwC;gCACjH,CAAC,CAAC,kHAAkH;yBACzH;qBACF;iBACF,CAAC;YACJ,CAAC;SACF;KACF,CAAC;AACJ,CAAC"}

package/dist/memory/modules/youtube/transcript-fetcher.d.ts

@@ -0,0 +1,17 @@
+import type { YoutubeConfig } from '../../../config/schema.js';
+export interface YoutubeTranscriptSegment {
+    start: number;
+    duration: number;
+    text: string;
+}
+export interface YoutubeTranscriptResult {
+    video_id: string;
+    title: string;
+    channel: string;
+    language: string;
+    segments: YoutubeTranscriptSegment[];
+    full_text: string;
+}
+export declare function extractVideoId(url: string): string;
+export declare function fetchTranscript(url: string, config: YoutubeConfig): Promise<YoutubeTranscriptResult>;
+//# sourceMappingURL=transcript-fetcher.d.ts.map

package/dist/memory/modules/youtube/transcript-fetcher.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcript-fetcher.d.ts","sourceRoot":"","sources":["../../../../src/memory/modules/youtube/transcript-fetcher.ts"],"names":[],"mappings":"AAeA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,2BAA2B,CAAC;AAI/D,MAAM,WAAW,wBAAwB;IACvC,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,CAAC;CACd;AAED,MAAM,WAAW,uBAAuB;IACtC,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,EAAE,MAAM,CAAC;IAChB,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,EAAE,wBAAwB,EAAE,CAAC;IACrC,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,wBAAgB,cAAc,CAAC,GAAG,EAAE,MAAM,GAAG,MAAM,CAkBlD;AAED,wBAAsB,eAAe,CACnC,GAAG,EAAE,MAAM,EACX,MAAM,EAAE,aAAa,GACpB,OAAO,CAAC,uBAAuB,CAAC,CAmFlC"}

package/dist/memory/modules/youtube/transcript-fetcher.js

@@ -0,0 +1,178 @@
+// src/memory/modules/youtube/transcript-fetcher.ts
+/**
+ * YouTube transcript fetcher.
+ *
+ * Strategy:
+ *   1. Fetch the video page HTML.
+ *   2. Extract the player config containing `captionTracks`.
+ *   3. Pick the preferred language track (or first available).
+ *   4. Download the timedtext XML and parse into segments.
+ *
+ * Fallback: shell out to `yt-dlp --write-auto-sub --skip-download --sub-format vtt`
+ *           if step 2/3 fails AND `yt-dlp` is on PATH AND config.fallbackToYtdlp is true.
+ */
+import { spawn } from 'child_process';
+import { createLogger } from '../../../logger.js';
+const log = createLogger('youtube:transcript');
+export function extractVideoId(url) {
+    // Handle multiple URL shapes:
+    //   https://www.youtube.com/watch?v=ID
+    //   https://youtu.be/ID
+    //   https://www.youtube.com/embed/ID
+    //   https://www.youtube.com/shorts/ID
+    const patterns = [
+        /[?&]v=([A-Za-z0-9_-]{11})/,
+        /youtu\.be\/([A-Za-z0-9_-]{11})/,
+        /embed\/([A-Za-z0-9_-]{11})/,
+        /shorts\/([A-Za-z0-9_-]{11})/,
+    ];
+    for (const p of patterns) {
+        const m = url.match(p);
+        if (m)
+            return m[1];
+    }
+    if (/^[A-Za-z0-9_-]{11}$/.test(url))
+        return url; // raw id
+    throw new Error(`Could not extract YouTube video id from: ${url}`);
+}
+export async function fetchTranscript(url, config) {
+    const videoId = extractVideoId(url);
+    const watchUrl = `https://www.youtube.com/watch?v=${videoId}`;
+    // Step 1: fetch watch page
+    const resp = await fetch(watchUrl, {
+        headers: {
+            'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/17 Safari/605.1.15',
+            'Accept-Language': `${config.preferLanguage},en;q=0.9`,
+        },
+    });
+    if (!resp.ok)
+        throw new Error(`YouTube watch page returned ${resp.status}`);
+    const html = await resp.text();
+    // Step 2: extract title + channel + captionTracks
+    const titleMatch = html.match(/<title>([^<]+)<\/title>/);
+    const title = titleMatch ? titleMatch[1].replace(/ - YouTube$/, '').trim() : videoId;
+    const channelMatch = html.match(/"ownerChannelName":"([^"]+)"/);
+    const channel = channelMatch ? channelMatch[1] : 'unknown';
+    const captionsMatch = html.match(/"captionTracks":(\[[^\]]+\])/);
+    if (!captionsMatch) {
+        if (config.fallbackToYtdlp) {
+            log.info(`No captionTracks in HTML for ${videoId} — falling back to yt-dlp`);
+            return await fetchViaYtdlp(videoId, watchUrl, title, channel, config);
+        }
+        throw new Error(`No captions available for video ${videoId}`);
+    }
+    const tracks = JSON.parse(captionsMatch[1]);
+    const preferred = tracks.find((t) => t.languageCode === config.preferLanguage && t.kind !== 'asr') ??
+        tracks.find((t) => t.languageCode === config.preferLanguage) ??
+        tracks.find((t) => t.kind !== 'asr') ??
+        tracks[0];
+    if (!preferred)
+        throw new Error('No usable caption track');
+    // Step 3: download timedtext
+    const xmlResp = await fetch(preferred.baseUrl, {
+        headers: {
+            'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/17 Safari/605.1.15',
+            'Referer': 'https://www.youtube.com/',
+        },
+    });
+    if (!xmlResp.ok)
+        throw new Error(`Timedtext fetch failed: ${xmlResp.status}`);
+    const xml = await xmlResp.text();
+    // Step 4: parse <text start="X" dur="Y">content</text>
+    const segments = [];
+    const re = /<text start="([\d.]+)" dur="([\d.]+)"[^>]*>([\s\S]*?)<\/text>/g;
+    let m;
+    while ((m = re.exec(xml)) !== null) {
+        segments.push({
+            start: parseFloat(m[1]),
+            duration: parseFloat(m[2]),
+            text: decodeHtmlEntities(m[3]).replace(/\n/g, ' ').trim(),
+        });
+    }
+    // YouTube's timedtext API may return empty XML even for videos with captions.
+    // If we got 0 segments and yt-dlp fallback is enabled, use it.
+    if (segments.length === 0 && config.fallbackToYtdlp) {
+        log.info(`Timedtext returned 0 segments for ${videoId} — falling back to yt-dlp`);
+        return await fetchViaYtdlp(videoId, watchUrl, title, channel, config);
+    }
+    if (segments.length === 0) {
+        throw new Error(`No transcript segments found for video ${videoId}. Enable fallbackToYtdlp or check if the video has captions.`);
+    }
+    return {
+        video_id: videoId,
+        title,
+        channel,
+        language: preferred.languageCode,
+        segments,
+        full_text: segments.map((s) => s.text).join(' '),
+    };
+}
+function decodeHtmlEntities(s) {
+    return s
+        .replace(/&amp;/g, '&')
+        .replace(/&lt;/g, '<')
+        .replace(/&gt;/g, '>')
+        .replace(/&quot;/g, '"')
+        .replace(/&#39;/g, "'")
+        .replace(/&#x27;/g, "'")
+        .replace(/&nbsp;/g, ' ');
+}
+async function fetchViaYtdlp(videoId, url, title, channel, config) {
+    return new Promise((resolve, reject) => {
+        const args = [
+            '--write-auto-sub',
+            '--write-sub',
+            '--sub-lang',
+            `${config.preferLanguage},en`,
+            '--skip-download',
+            '--sub-format',
+            'vtt',
+            '--output',
+            `/tmp/engram-yt-%(id)s.%(ext)s`,
+            url,
+        ];
+        const child = spawn('yt-dlp', args);
+        let stderr = '';
+        child.stderr.on('data', (b) => (stderr += b.toString()));
+        child.on('error', (e) => reject(new Error(`yt-dlp not available: ${e.message}. Install via 'brew install yt-dlp'.`)));
+        child.on('exit', async (code) => {
+            if (code !== 0) {
+                reject(new Error(`yt-dlp exit ${code}: ${stderr.slice(-300)}`));
+                return;
+            }
+            try {
+                const fs = await import('fs');
+                // Find the produced VTT file
+                const candidates = fs.readdirSync('/tmp').filter((n) => n.startsWith(`engram-yt-${videoId}`));
+                const vttFile = candidates.find((n) => n.endsWith('.vtt'));
+                if (!vttFile)
+                    throw new Error('yt-dlp produced no .vtt');
+                const vtt = fs.readFileSync(`/tmp/${vttFile}`, 'utf-8');
+                const segments = parseVtt(vtt);
+                resolve({
+                    video_id: videoId,
+                    title,
+                    channel,
+                    language: config.preferLanguage,
+                    segments,
+                    full_text: segments.map((s) => s.text).join(' '),
+                });
+            }
+            catch (e) {
+                reject(e);
+            }
+        });
+    });
+}
+function parseVtt(vtt) {
+    const segments = [];
+    const re = /(\d+):(\d{2}):(\d{2})[.,](\d+) --> (\d+):(\d{2}):(\d{2})[.,](\d+)\n([^\n]+)/g;
+    let m;
+    while ((m = re.exec(vtt)) !== null) {
+        const start = Number(m[1]) * 3600 + Number(m[2]) * 60 + Number(m[3]) + Number(m[4]) / 1000;
+        const end = Number(m[5]) * 3600 + Number(m[6]) * 60 + Number(m[7]) + Number(m[8]) / 1000;
+        segments.push({ start, duration: end - start, text: m[9].trim() });
+    }
+    return segments;
+}
+//# sourceMappingURL=transcript-fetcher.js.map

package/dist/memory/modules/youtube/transcript-fetcher.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcript-fetcher.js","sourceRoot":"","sources":["../../../../src/memory/modules/youtube/transcript-fetcher.ts"],"names":[],"mappings":"AAAA,mDAAmD;AACnD;;;;;;;;;;;GAWG;AACH,OAAO,EAAE,KAAK,EAAE,MAAM,eAAe,CAAC;AACtC,OAAO,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAGlD,MAAM,GAAG,GAAG,YAAY,CAAC,oBAAoB,CAAC,CAAC;AAiB/C,MAAM,UAAU,cAAc,CAAC,GAAW;IACxC,8BAA8B;IAC9B,uCAAuC;IACvC,wBAAwB;IACxB,qCAAqC;IACrC,sCAAsC;IACtC,MAAM,QAAQ,GAAG;QACf,2BAA2B;QAC3B,gCAAgC;QAChC,4BAA4B;QAC5B,6BAA6B;KAC9B,CAAC;IACF,KAAK,MAAM,CAAC,IAAI,QAAQ,EAAE,CAAC;QACzB,MAAM,CAAC,GAAG,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QACvB,IAAI,CAAC;YAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;IACrB,CAAC;IACD,IAAI,qBAAqB,CAAC,IAAI,CAAC,GAAG,CAAC;QAAE,OAAO,GAAG,CAAC,CAAC,SAAS;IAC1D,MAAM,IAAI,KAAK,CAAC,4CAA4C,GAAG,EAAE,CAAC,CAAC;AACrE,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,eAAe,CACnC,GAAW,EACX,MAAqB;IAErB,MAAM,OAAO,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC;IACpC,MAAM,QAAQ,GAAG,mCAAmC,OAAO,EAAE,CAAC;IAE9D,2BAA2B;IAC3B,MAAM,IAAI,GAAG,MAAM,KAAK,CAAC,QAAQ,EAAE;QACjC,OAAO,EAAE;YACP,YAAY,EACV,qHAAqH;YACvH,iBAAiB,EAAE,GAAG,MAAM,CAAC,cAAc,WAAW;SACvD;KACF,CAAC,CAAC;IACH,IAAI,CAAC,IAAI,CAAC,EAAE;QAAE,MAAM,IAAI,KAAK,CAAC,+BAA+B,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC;IAC5E,MAAM,IAAI,GAAG,MAAM,IAAI,CAAC,IAAI,EAAE,CAAC;IAE/B,kDAAkD;IAClD,MAAM,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAC;IACzD,MAAM,KAAK,GAAG,UAAU,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,aAAa,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC;IACrF,MAAM,YAAY,GAAG,IAAI,CAAC,KAAK,CAAC,8BAA8B,CAAC,CAAC;IAChE,MAAM,OAAO,GAAG,YAAY,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAE3D,MAAM,aAAa,GAAG,IAAI,CAAC,KAAK,CAAC,8BAA8B,CAAC,CAAC;IACjE,IAAI,CAAC,aAAa,EAAE,CAAC;QACnB,IAAI,MAAM,CAAC,eAAe,EAAE,CAAC;YAC3B,GAAG,CAAC,IAAI,CAAC,gCAAgC,OAAO,2BAA2B,CAAC,CAAC;YAC7E,OAAO,MAAM,aAAa,CAAC,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC;QACxE,CAAC;QACD,MAAM,IAAI,KAAK,CAAC,mCAAmC,OAAO,EAAE,CAAC,CAAC;IAChE,CAAC;IAGD,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,aAAa,CAAC,CAAC,CAAC,CAAe,CAAC;IAC1D,MAAM,SAAS,GACb,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,YAAY,KAAK,MAAM,CAAC,cAAc,IAAI,CAAC,CAAC,IAAI,KAAK,KAAK,CAAC;QAChF,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,YAAY,KAAK,MAAM,CAAC,cAAc,CAAC;QAC5D,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,KAAK,CAAC;QACpC,MAAM,CAAC,CAAC,CAAC,CAAC;IACZ,IAAI,CAAC,SAAS;QAAE,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC;IAE3D,6BAA6B;IAC7B,MAAM,OAAO,GAAG,MAAM,KAAK,CAAC,SAAS,CAAC,OAAO,EAAE;QAC7C,OAAO,EAAE;YACP,YAAY,EACV,qHAAqH;YACvH,SAAS,EAAE,0BAA0B;SACtC;KACF,CAAC,CAAC;IACH,IAAI,CAAC,OAAO,CAAC,EAAE;QAAE,MAAM,IAAI,KAAK,CAAC,2BAA2B,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;IAC9E,MAAM,GAAG,GAAG,MAAM,OAAO,CAAC,IAAI,EAAE,CAAC;IAEjC,uDAAuD;IACvD,MAAM,QAAQ,GAA+B,EAAE,CAAC;IAChD,MAAM,EAAE,GAAG,gEAAgE,CAAC;IAC5E,IAAI,CAAyB,CAAC;IAC9B,OAAO,CAAC,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,KAAK,IAAI,EAAE,CAAC;QACnC,QAAQ,CAAC,IAAI,CAAC;YACZ,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;YACvB,QAAQ,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;YAC1B,IAAI,EAAE,kBAAkB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE;SAC1D,CAAC,CAAC;IACL,CAAC;IAED,8EAA8E;IAC9E,+DAA+D;IAC/D,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC,IAAI,MAAM,CAAC,eAAe,EAAE,CAAC;QACpD,GAAG,CAAC,IAAI,CAAC,qCAAqC,OAAO,2BAA2B,CAAC,CAAC;QAClF,OAAO,MAAM,aAAa,CAAC,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC;IACxE,CAAC;IAED,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CACb,0CAA0C,OAAO,8DAA8D,CAChH,CAAC;IACJ,CAAC;IAED,OAAO;QACL,QAAQ,EAAE,OAAO;QACjB,KAAK;QACL,OAAO;QACP,QAAQ,EAAE,SAAS,CAAC,YAAY;QAChC,QAAQ;QACR,SAAS,EAAE,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC;KACjD,CAAC;AACJ,CAAC;AAED,SAAS,kBAAkB,CAAC,CAAS;IACnC,OAAO,CAAC;SACL,OAAO,CAAC,QAAQ,EAAE,GAAG,CAAC;SACtB,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC;SACrB,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC;SACrB,OAAO,CAAC,SAAS,EAAE,GAAG,CAAC;SACvB,OAAO,CAAC,QAAQ,EAAE,GAAG,CAAC;SACtB,OAAO,CAAC,SAAS,EAAE,GAAG,CAAC;SACvB,OAAO,CAAC,SAAS,EAAE,GAAG,CAAC,CAAC;AAC7B,CAAC;AAED,KAAK,UAAU,aAAa,CAC1B,OAAe,EACf,GAAW,EACX,KAAa,EACb,OAAe,EACf,MAAqB;IAErB,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;QACrC,MAAM,IAAI,GAAG;YACX,kBAAkB;YAClB,aAAa;YACb,YAAY;YACZ,GAAG,MAAM,CAAC,cAAc,KAAK;YAC7B,iBAAiB;YACjB,cAAc;YACd,KAAK;YACL,UAAU;YACV,+BAA+B;YAC/B,GAAG;SACJ,CAAC;QACF,MAAM,KAAK,GAAG,KAAK,CAAC,QAAQ,EAAE,IAAI,CAAC,CAAC;QACpC,IAAI,MAAM,GAAG,EAAE,CAAC;QAChB,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,MAAM,IAAI,CAAC,CAAC,QAAQ,EAAE,CAAC,CAAC,CAAC;QACzD,KAAK,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,CAAC,EAAE,EAAE,CACtB,MAAM,CAAC,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC,OAAO,sCAAsC,CAAC,CAAC,CAC5F,CAAC;QACF,KAAK,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,EAAE,IAAI,EAAE,EAAE;YAC9B,IAAI,IAAI,KAAK,CAAC,EAAE,CAAC;gBACf,MAAM,CAAC,IAAI,KAAK,CAAC,eAAe,IAAI,KAAK,MAAM,CAAC,KAAK,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;gBAChE,OAAO;YACT,CAAC;YACD,IAAI,CAAC;gBACH,MAAM,EAAE,GAAG,MAAM,MAAM,CAAC,IAAI,CAAC,CAAC;gBAC9B,6BAA6B;gBAC7B,MAAM,UAAU,GAAG,EAAE,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,aAAa,OAAO,EAAE,CAAC,CAAC,CAAC;gBAC9F,MAAM,OAAO,GAAG,UAAU,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC;gBAC3D,IAAI,CAAC,OAAO;oBAAE,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC;gBACzD,MAAM,GAAG,GAAG,EAAE,CAAC,YAAY,CAAC,QAAQ,OAAO,EAAE,EAAE,OAAO,CAAC,CAAC;gBACxD,MAAM,QAAQ,GAAG,QAAQ,CAAC,GAAG,CAAC,CAAC;gBAC/B,OAAO,CAAC;oBACN,QAAQ,EAAE,OAAO;oBACjB,KAAK;oBACL,OAAO;oBACP,QAAQ,EAAE,MAAM,CAAC,cAAc;oBAC/B,QAAQ;oBACR,SAAS,EAAE,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC;iBACjD,CAAC,CAAC;YACL,CAAC;YAAC,OAAO,CAAC,EAAE,CAAC;gBACX,MAAM,CAAC,CAAC,CAAC,CAAC;YACZ,CAAC;QACH,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;AACL,CAAC;AAED,SAAS,QAAQ,CAAC,GAAW;IAC3B,MAAM,QAAQ,GAA+B,EAAE,CAAC;IAChD,MAAM,EAAE,GAAG,8EAA8E,CAAC;IAC1F,IAAI,CAAyB,CAAC;IAC9B,OAAO,CAAC,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,KAAK,IAAI,EAAE,CAAC;QACnC,MAAM,KAAK,GACT,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,EAAE,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC;QAC/E,MAAM,GAAG,GACP,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,EAAE,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC;QAC/E,QAAQ,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,QAAQ,EAAE,GAAG,GAAG,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;IACrE,CAAC;IACD,OAAO,QAAQ,CAAC;AAClB,CAAC"}

package/dist/memory/modules/youtube/watcher.d.ts

@@ -0,0 +1,30 @@
+import type { EmbeddingsConfig, YoutubeConfig } from '../../../config/schema.js';
+import type { MemoryStore } from '../../core/store.js';
+/**
+ * Accepts a channel ID (UCxxx…) or a full URL and returns the channel ID.
+ * For handle-based URLs, fetches the channel page and extracts the canonical
+ * channel ID from the embedded JSON.
+ */
+export declare function resolveChannelId(input: string): Promise<string>;
+/**
+ * Fetches the RSS feed for a channel, diffs against the last known video,
+ * and ingests any new videos (up to 15 per poll tick — the RSS feed max).
+ */
+export declare function pollChannel(channelId: string, channelName: string, store: MemoryStore, embeddingsConfig: EmbeddingsConfig, youtubeConfig: YoutubeConfig): Promise<{
+    ingested: number;
+}>;
+/**
+ * Fetches a YouTube playlist page (public only), extracts video IDs from
+ * ytInitialData JSON blob embedded in the HTML, and ingests new ones.
+ */
+export declare function importPlaylist(playlistUrl: string, store: MemoryStore, embeddingsConfig: EmbeddingsConfig, youtubeConfig: YoutubeConfig, limit?: number): Promise<{
+    imported: number;
+    skipped: number;
+}>;
+/**
+ * Starts the channel polling loop for all watched YouTube channels.
+ * Called once at server startup from the YouTube module's onBoot().
+ * The interval is configurable via config.youtube.channelPollIntervalMs (default 6h).
+ */
+export declare function startChannelCron(store: MemoryStore, embeddingsConfig: EmbeddingsConfig, youtubeConfig: YoutubeConfig): void;
+//# sourceMappingURL=watcher.d.ts.map

package/dist/memory/modules/youtube/watcher.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"watcher.d.ts","sourceRoot":"","sources":["../../../../src/memory/modules/youtube/watcher.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,EAAE,gBAAgB,EAAE,aAAa,EAAE,MAAM,2BAA2B,CAAC;AACjF,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AASvD;;;;GAIG;AACH,wBAAsB,gBAAgB,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAkBrE;AAGD;;;GAGG;AACH,wBAAsB,WAAW,CAC/B,SAAS,EAAE,MAAM,EACjB,WAAW,EAAE,MAAM,EACnB,KAAK,EAAE,WAAW,EAClB,gBAAgB,EAAE,gBAAgB,EAClC,aAAa,EAAE,aAAa,GAC3B,OAAO,CAAC;IAAE,QAAQ,EAAE,MAAM,CAAA;CAAE,CAAC,CA6E/B;AAGD;;;GAGG;AACH,wBAAsB,cAAc,CAClC,WAAW,EAAE,MAAM,EACnB,KAAK,EAAE,WAAW,EAClB,gBAAgB,EAAE,gBAAgB,EAClC,aAAa,EAAE,aAAa,EAC5B,KAAK,SAAK,GACT,OAAO,CAAC;IAAE,QAAQ,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE,CAAC,CA6DhD;AAGD;;;;GAIG;AACH,wBAAgB,gBAAgB,CAC9B,KAAK,EAAE,WAAW,EAClB,gBAAgB,EAAE,gBAAgB,EAClC,aAAa,EAAE,aAAa,GAC3B,IAAI,CA8CN"}

package/dist/memory/modules/youtube/watcher.js

@@ -0,0 +1,198 @@
+// src/memory/modules/youtube/watcher.ts
+import { XMLParser } from 'fast-xml-parser';
+import { createLogger } from '../../../logger.js';
+import { getDb } from '../../../db/index.js';
+import { ingestYouTube } from './ingest.js';
+const log = createLogger('youtube:watcher');
+const FEED_BASE = 'https://www.youtube.com/feeds/videos.xml?channel_id=';
+const CHANNEL_ID_RE = /^UC[A-Za-z0-9_-]{22}$/;
+const CHANNEL_URL_RE = /(?:youtube\.com\/(?:channel\/|@))([A-Za-z0-9_.-]+)/;
+// ── resolveChannelId ───────────────────────────────────────────────────────
+/**
+ * Accepts a channel ID (UCxxx…) or a full URL and returns the channel ID.
+ * For handle-based URLs, fetches the channel page and extracts the canonical
+ * channel ID from the embedded JSON.
+ */
+export async function resolveChannelId(input) {
+    if (CHANNEL_ID_RE.test(input))
+        return input;
+    // Try to extract handle or channel path from URL
+    const match = CHANNEL_URL_RE.exec(input);
+    if (match) {
+        const candidate = match[1];
+        if (CHANNEL_ID_RE.test(candidate))
+            return candidate;
+        // It's a handle — fetch the channel page to get the real channel ID
+        const channelUrl = input.startsWith('http') ? input : `https://www.youtube.com/${input}`;
+        const html = await fetchText(channelUrl);
+        const idMatch = /"channelId":"(UC[A-Za-z0-9_-]{22})"/.exec(html);
+        if (idMatch)
+            return idMatch[1];
+    }
+    throw new Error(`Cannot resolve channel ID from "${input}". Provide a full channel URL or a channel ID starting with UC.`);
+}
+// ── pollChannel ────────────────────────────────────────────────────────────
+/**
+ * Fetches the RSS feed for a channel, diffs against the last known video,
+ * and ingests any new videos (up to 15 per poll tick — the RSS feed max).
+ */
+export async function pollChannel(channelId, channelName, store, embeddingsConfig, youtubeConfig) {
+    const feedUrl = `${FEED_BASE}${channelId}`;
+    let xml;
+    try {
+        xml = await fetchText(feedUrl);
+    }
+    catch (e) {
+        log.warn(`Failed to fetch RSS for channel ${channelId}: ${e instanceof Error ? e.message : String(e)}`);
+        return { ingested: 0 };
+    }
+    const parser = new XMLParser({ ignoreAttributes: false, attributeNamePrefix: '_' });
+    const parsed = parser.parse(xml);
+    const entries = parsed.feed?.entry
+        ? Array.isArray(parsed.feed.entry)
+            ? parsed.feed.entry
+            : [parsed.feed.entry]
+        : [];
+    const db = getDb();
+    const lastSyncedRow = db
+        .prepare(`SELECT last_synced_at FROM watched_sources
+       WHERE module_id = 'youtube' AND external_id = ?`)
+        .get(channelId);
+    const lastSyncedAt = lastSyncedRow?.last_synced_at ?? 0;
+    let ingested = 0;
+    for (const entry of entries) {
+        const videoId = entry['yt:videoId'];
+        const publishedStr = entry.published;
+        if (!videoId || !publishedStr)
+            continue;
+        const publishedAt = new Date(publishedStr).getTime();
+        if (publishedAt <= lastSyncedAt)
+            continue;
+        const videoUrl = `https://www.youtube.com/watch?v=${videoId}`;
+        try {
+            await ingestYouTube(videoUrl, store, embeddingsConfig, youtubeConfig, {
+                tags: [channelName, 'channel-watch'],
+                sourceContext: `YouTube channel: ${channelName}`,
+            });
+            ingested++;
+            log.info(`Ingested video ${videoId} from channel ${channelId}`);
+        }
+        catch (e) {
+            log.warn(`Failed to ingest video ${videoId} from channel ${channelId}: ${e instanceof Error ? e.message : String(e)}`);
+        }
+    }
+    // Update last_synced_at
+    db.prepare(`UPDATE watched_sources SET last_synced_at = ? WHERE module_id = 'youtube' AND external_id = ?`).run(Date.now(), channelId);
+    return { ingested };
+}
+// ── importPlaylist ─────────────────────────────────────────────────────────
+/**
+ * Fetches a YouTube playlist page (public only), extracts video IDs from
+ * ytInitialData JSON blob embedded in the HTML, and ingests new ones.
+ */
+export async function importPlaylist(playlistUrl, store, embeddingsConfig, youtubeConfig, limit = 50) {
+    const html = await fetchText(playlistUrl);
+    // Extract ytInitialData JSON
+    const match = /var ytInitialData = ({.+?});<\/script>/s.exec(html);
+    if (!match) {
+        throw new Error('Could not parse playlist page. The playlist may be private or YouTube has changed its page structure.');
+    }
+    let data;
+    try {
+        data = JSON.parse(match[1]);
+    }
+    catch {
+        throw new Error('ytInitialData JSON parse failed — YouTube may have changed page format.');
+    }
+    // Walk the deeply nested structure to extract video IDs
+    const videoIds = extractVideoIds(data);
+    const unique = [...new Set(videoIds)].slice(0, limit);
+    if (!unique.length) {
+        return { imported: 0, skipped: 0 };
+    }
+    // Check which are already ingested
+    const db = getDb();
+    const existing = new Set(db
+        .prepare(`SELECT source_id FROM memories WHERE type = 'youtube' AND source_id IN (${unique.map(() => '?').join(',')})`)
+        .all(...unique).map((r) => r.source_id));
+    let imported = 0;
+    let skipped = 0;
+    for (const videoId of unique) {
+        if (existing.has(`youtube:${videoId}`)) {
+            skipped++;
+            continue;
+        }
+        const url = `https://www.youtube.com/watch?v=${videoId}`;
+        try {
+            await ingestYouTube(url, store, embeddingsConfig, youtubeConfig, {
+                tags: ['watch-later', 'playlist-import'],
+            });
+            imported++;
+        }
+        catch (e) {
+            log.warn(`Failed to ingest ${videoId}: ${e instanceof Error ? e.message : String(e)}`);
+            skipped++;
+        }
+    }
+    return { imported, skipped };
+}
+// ── startChannelCron ───────────────────────────────────────────────────────
+/**
+ * Starts the channel polling loop for all watched YouTube channels.
+ * Called once at server startup from the YouTube module's onBoot().
+ * The interval is configurable via config.youtube.channelPollIntervalMs (default 6h).
+ */
+export function startChannelCron(store, embeddingsConfig, youtubeConfig) {
+    const intervalMs = youtubeConfig.channelPollIntervalMs ?? 21_600_000; // 6h default
+    const db = getDb();
+    async function tick() {
+        const channels = db
+            .prepare(`SELECT external_id, display_name FROM watched_sources
+         WHERE module_id = 'youtube' AND enabled = 1`)
+            .all();
+        if (!channels.length)
+            return;
+        log.info(`YouTube channel cron: polling ${channels.length} channel(s)`);
+        for (const ch of channels) {
+            try {
+                const { ingested } = await pollChannel(ch.external_id, ch.display_name, store, embeddingsConfig, youtubeConfig);
+                if (ingested > 0)
+                    log.info(`  ${ch.display_name}: ingested ${ingested} new video(s)`);
+            }
+            catch (e) {
+                log.error(`Channel poll error for ${ch.external_id}: ${e instanceof Error ? e.message : String(e)}`);
+            }
+        }
+    }
+    // Run once immediately at startup, then on interval
+    tick().catch((e) => log.error(`Initial channel poll failed: ${e instanceof Error ? e.message : String(e)}`));
+    setInterval(() => tick().catch((e) => log.error(`Channel poll error: ${e instanceof Error ? e.message : String(e)}`)), intervalMs);
+    log.info(`YouTube channel cron started (${intervalMs / 3_600_000}h interval)`);
+}
+// ── helpers ────────────────────────────────────────────────────────────────
+async function fetchText(url) {
+    const res = await fetch(url, {
+        headers: {
+            'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 Chrome/124 Safari/537.36',
+            'Accept-Language': 'en-US,en;q=0.9',
+        },
+        signal: AbortSignal.timeout(15_000),
+    });
+    if (!res.ok)
+        throw new Error(`HTTP ${res.status} for ${url}`);
+    return res.text();
+}
+function extractVideoIds(obj) {
+    const ids = [];
+    JSON.stringify(obj, (_key, value) => {
+        if (typeof value === 'object' &&
+            value !== null &&
+            'videoId' in value &&
+            typeof value.videoId === 'string') {
+            ids.push(value.videoId);
+        }
+        return value;
+    });
+    return ids;
+}
+//# sourceMappingURL=watcher.js.map