import { type InferSchema } from 'xmcp'
import { z } from 'zod'
import { extractYouTubeContent, checkDependencies } from '@/utils/youtube-extractor'
// Define the schema for tool parameters
export const schema = {
url: z.string().url().describe('YouTube video URL to extract content from'),
includeTitle: z.boolean().optional().default(true).describe('Include video title in the output'),
includeDescription: z.boolean().optional().default(true).describe('Include video description in the output'),
includeTranscript: z
.boolean()
.optional()
.default(false)
.describe('Include video transcript/audio content in the output'),
}
// Define tool metadata
export const metadata = {
name: 'extract-youtube',
description:
'Extract comprehensive content from YouTube videos including title, description, and transcript. Automatically uses available subtitles with language priority or falls back to optimized Whisper speech-to-text conversion with language detection.',
annotations: {
title: 'Extract YouTube Video Content',
readOnlyHint: true,
destructiveHint: false,
idempotentHint: true,
},
}
/**
* Format duration from seconds to readable format
*/
const formatDuration = (seconds?: number): string => {
if (!seconds) return 'Unknown'
const hours = Math.floor(seconds / 3600)
const minutes = Math.floor((seconds % 3600) / 60)
const secs = seconds % 60
if (hours > 0) {
return `${hours}:${minutes.toString().padStart(2, '0')}:${secs.toString().padStart(2, '0')}`
} else {
return `${minutes}:${secs.toString().padStart(2, '0')}`
}
}
/**
* Format number with commas
*/
const formatNumber = (num?: number): string => {
if (num === undefined || num === null) return 'Unknown'
return num.toLocaleString()
}
/**
* Format upload date
*/
const formatUploadDate = (dateStr?: string): string => {
if (!dateStr) return 'Unknown'
// yt-dlp returns date in YYYYMMDD format
const year = dateStr.substring(0, 4)
const month = dateStr.substring(4, 6)
const day = dateStr.substring(6, 8)
return `${year}-${month}-${day}`
}
// Tool implementation
export default async function extractYoutube({
url,
includeTitle,
includeDescription,
includeTranscript,
}: InferSchema<typeof schema>) {
try {
console.log(`[extract-youtube] Starting extraction for URL: ${url}`)
console.log(
`[extract-youtube] Options: title=${includeTitle}, description=${includeDescription}, transcript=${includeTranscript}`,
)
// Check dependencies first
console.log(`[extract-youtube] Checking dependencies...`)
const deps = await checkDependencies()
console.log(`[extract-youtube] Dependencies check result:`, deps)
if (!deps.ytdlp) {
return {
content: [
{
type: 'text',
text:
'❌ Error: yt-dlp is not installed. Please install it first:\n\n' +
'pip install yt-dlp\n' +
'# or\n' +
'brew install yt-dlp',
},
],
}
}
if (!deps.ffmpeg) {
return {
content: [
{
type: 'text',
text:
'❌ Error: ffmpeg is not installed. Please install it first:\n\n' +
'brew install ffmpeg\n' +
'# or\n' +
'sudo apt install ffmpeg',
},
],
}
}
if (includeTranscript && !deps.whisper) {
return {
content: [
{
type: 'text',
text:
'⚠️ Warning: Whisper is not installed. Transcript extraction will only work if subtitles are available.\n\n' +
'To install Whisper:\n' +
'pip install openai-whisper\n' +
'# or\n' +
'brew install openai-whisper',
},
],
}
}
// Extract content
console.log(`[extract-youtube] Starting content extraction...`)
// Add timeout to prevent hanging
const extractionPromise = extractYouTubeContent(url, {
includeTitle: includeTitle ?? true,
includeDescription: includeDescription ?? true,
includeTranscript: includeTranscript ?? false,
})
const timeoutPromise = new Promise<never>((_, reject) => {
setTimeout(() => reject(new Error('Extraction timeout after 60 seconds')), 60000)
})
const result = await Promise.race([extractionPromise, timeoutPromise])
console.log(`[extract-youtube] Content extraction completed successfully`)
// Create a formatted text output
let formattedOutput = '🎬 YouTube Content Extraction Results\n'
formattedOutput += `${'='.repeat(50)}\n\n`
// Video Information Section
if (result.title || result.uploader) {
formattedOutput += '📺 Video Information:\n'
if (result.title) {
formattedOutput += ` 📝 Title: ${result.title}\n`
}
if (result.uploader) {
formattedOutput += ` 👤 Channel: ${result.uploader}\n`
}
if (result.duration) {
formattedOutput += ` ⏱️ Duration: ${formatDuration(result.duration)}\n`
}
if (result.uploadDate) {
formattedOutput += ` 📅 Upload Date: ${formatUploadDate(result.uploadDate)}\n`
}
if (result.viewCount) {
formattedOutput += ` 👀 Views: ${formatNumber(result.viewCount)}\n`
}
if (result.likeCount) {
formattedOutput += ` 👍 Likes: ${formatNumber(result.likeCount)}\n`
}
formattedOutput += '\n'
}
// Description Section
if (result.description) {
formattedOutput += '📄 Description:\n'
// Truncate very long descriptions
const maxDescLength = 1000
const description =
result.description.length > maxDescLength
? `${result.description.substring(0, maxDescLength)}...`
: result.description
formattedOutput += `${description}\n\n`
}
// Tags and Categories
if (result.tags && result.tags.length > 0) {
formattedOutput += '🏷️ Tags:\n'
formattedOutput += ` ${result.tags.slice(0, 10).join(', ')}${result.tags.length > 10 ? '...' : ''}\n\n`
}
if (result.categories && result.categories.length > 0) {
formattedOutput += '📂 Categories:\n'
formattedOutput += ` ${result.categories.join(', ')}\n\n`
}
// Transcript Section
if (result.transcript) {
formattedOutput += `📜 Transcript (${result.transcriptSource}):\n`
// Add source-specific information
if (result.transcriptSource === 'whisper') {
formattedOutput += ` 🤖 Whisper Model: ${result.whisperModel || 'unknown'}\n`
if (result.detectedLanguage) {
formattedOutput += ` 🌐 Detected Language: ${result.detectedLanguage}\n`
}
} else if (result.transcriptSource === 'subtitles') {
if (result.detectedLanguage) {
formattedOutput += ` 🌐 Subtitle Language: ${result.detectedLanguage}\n`
}
}
formattedOutput += ` ${'-'.repeat(40)}\n`
formattedOutput += ` ${result.transcript.replace(/\n/g, '\n ')}\n\n`
}
// Footer
formattedOutput += '🔗 Source URL:\n'
formattedOutput += ` ${url}\n\n`
formattedOutput += `⏰ Extracted at: ${new Date().toISOString()}\n\n`
// Limit total response size to prevent issues
const maxResponseSize = 50000 // 50KB limit
if (formattedOutput.length > maxResponseSize) {
formattedOutput = `${formattedOutput.substring(0, maxResponseSize)}\n\n[Response truncated due to size limit]`
}
return {
content: [
{
type: 'text',
text: formattedOutput,
},
],
}
} catch (error) {
console.error(`[extract-youtube] Error occurred:`, error)
console.error(`[extract-youtube] Error stack:`, error instanceof Error ? error.stack : 'No stack trace')
return {
content: [
{
type: 'text',
text: `❌ Error extracting YouTube content: ${error instanceof Error ? error.message : 'Unknown error'}`,
},
],
}
}
}
