SRT Translation MCP Server

# SRT Translation MCP Server - AI Usage Guide ## 🚨 CRITICAL FOR AI SYSTEMS 🚨 The `detect_conversations` function has been **COMPLETELY REDESIGNED** to return **METADATA ONLY** - **NO ACTUAL TEXT CONTENT**. ## ❌ WHAT IT DOES NOT RETURN - ❌ Actual subtitle text content - ❌ Full conversation text - ❌ Any readable text for translation - ❌ Translated text content - ❌ Source text content - ❌ Any actual subtitle text ## ✅ WHAT IT DOES RETURN - ✅ Chunk boundaries and timing information - ✅ Language detection (Arabic vs English) - ✅ Content complexity assessment - ✅ Translation priority ranking - ✅ Topic keywords (first 3 words only) - ✅ Speaker information - ✅ Content type classification ## 🤖 AI TRANSLATION WORKFLOW ### Step 1: Analyze File Structure (METADATA ONLY) ```javascript // Call detect_conversations first - RETURNS METADATA ONLY! const metadata = await mcp_call('detect_conversations', { content: srtFileContent }); // ⚠️ This returns metadata only - NO actual text content! ``` ### Step 2: Review Metadata for Planning ```javascript const analysis = JSON.parse(metadata.content[0].text); // Check language distribution console.log(analysis.languageDistribution); // {ar: 59, en: 0} // Review chunk priorities analysis.chunks.forEach(chunk => { console.log(`Chunk ${chunk.id}: ${chunk.translationPriority} priority, ${chunk.complexity} complexity`); }); ``` ### Step 3: Plan Translation Strategy ```javascript // Prioritize high-priority chunks const highPriorityChunks = analysis.chunks.filter(c => c.translationPriority === 'high'); // Check language detection const arabicChunks = analysis.chunks.filter(c => c.languageInfo.primary === 'ar'); ``` ### Step 4: Execute Translation (ACTUAL TEXT CONTENT) ```javascript // Use translate_srt for full file translation - THIS RETURNS ACTUAL TRANSLATED TEXT! const translated = await mcp_call('translate_srt', { content: srtFileContent, targetLanguage: 'en', sourceLanguage: 'ar' // Based on metadata analysis }); // ✅ This returns the complete translated SRT file with actual text content! // OR use AI workflow for intelligent translation: const aiTranslated = await mcp_call('translate_with_ai_workflow', { content: srtFileContent, targetLanguage: 'en', sourceLanguage: 'ar', aiWorkflow: workflowConfig }); // ✅ This also returns actual translated text with AI workflow! ``` ## 📊 METADATA STRUCTURE EXPLAINED ### File-Level Metadata ```json { "chunkCount": 59, // Total chunks to process "totalDuration": 7525000, // Duration in milliseconds "languageDistribution": {"ar": 59}, // Language counts "speakerDistribution": {"unknown": 59}, // Speaker counts "chunks": [...] // Individual chunk data } ``` ### Chunk-Level Metadata ```json { "id": "chunk-0", // Unique identifier "startTime": "0:0:2", // Start timestamp "endTime": "0:0:13", // End timestamp "duration": 11000, // Duration in milliseconds "subtitleCount": 2, // Number of subtitles "speaker": null, // Speaker name (if detected) "languageInfo": { // Language detection "primary": "ar", // Primary language "confidence": 0.8, // Detection confidence "indicators": ["arabic_script"] // Detection indicators }, "contentType": "general", // dialogue/narration/question/general "complexity": "medium", // low/medium/high "translationPriority": "low", // low/medium/high "topicKeywords": ["hello", "this"] // First 3 words only } ``` ## 🎯 AI DECISION MAKING GUIDE ### Language Detection - Use `languageInfo.primary` to determine source language - Use `languageInfo.confidence` to assess reliability - Arabic: `"ar"`, English: `"en"` ### Translation Planning - **High Priority**: Dialogue and questions (translate first) - **Medium Priority**: Complex chunks (need more attention) - **Low Priority**: Simple narration (translate last) ### Complexity Assessment - **Low**: Simple chunks (< 5 subtitles, < 10 seconds) - **Medium**: Moderate chunks (5-10 subtitles, 10-30 seconds) - **High**: Complex chunks (> 10 subtitles, > 30 seconds) ### Content Type Handling - **Dialogue**: Maintain speaker identity and conversational tone - **Narration**: Use formal language style - **Questions**: Preserve question structure and intonation - **General**: Standard translation approach ## ⚠️ COMMON AI MISTAKES TO AVOID 1. **DON'T** try to extract text from metadata - use `translate_srt()` instead 2. **DON'T** use `topicKeywords` for translation - they're for context only 3. **DON'T** ignore `translationPriority` - it helps optimize translation order 4. **DON'T** skip language detection - use `languageInfo.primary` for source language 5. **DON'T** expect translated text from `detect_conversations()` - it's metadata only! 6. **DON'T** ask for "actual Arabic translation" from metadata functions 7. **DO** use `translate_srt()` or `translate_with_ai_workflow()` for actual translation ## 🎯 WHICH FUNCTION TO USE FOR WHAT ### For Planning and Analysis (METADATA ONLY): - ✅ `detect_conversations()` - Returns metadata for AI decision-making - ✅ `analyze_content_for_ai()` - Returns content analysis - ✅ `generate_ai_workflow()` - Returns workflow configuration ### For Actual Translation (TEXT CONTENT): - ✅ `translate_srt()` - Returns complete translated SRT file - ✅ `translate_with_ai_workflow()` - Returns translated SRT with AI workflow - ✅ `translate_chunk()` - Returns translated chunk data ## 🔄 RECOMMENDED AI WORKFLOW 1. **Analyze**: Call `detect_conversations()` to get metadata 2. **Plan**: Review metadata to understand file structure 3. **Prioritize**: Focus on high-priority chunks first 4. **Translate**: Use `translate_srt()` for actual translation 5. **Validate**: Check results against original metadata expectations ## 📈 BENEFITS FOR AI - **Efficiency**: No need to process full text for analysis - **Intelligence**: Informed translation decisions based on metadata - **Scalability**: Works with large files without overwhelming AI - **Quality**: Better translation results through structured planning --- **Remember**: The `detect_conversations` function is your **planning tool**, not your **translation tool**. Use it to understand the file structure, then use `translate_srt()` for actual translation work.