Add MCP Server for Video Navigation with Keyword Search and Visual Search

[Copilot]

This PR implements a comprehensive Model Context Protocol (MCP) Server for video navigation that provides advanced search capabilities and intelligent video sampling job generation.

🎯 Features Implemented

Resources

• Video Resource (video://video_id) - Access individual video metadata, subtitles, and content
• ImageSet Resource (imageset://all_frames) - Collection of all sampled video frames across loaded videos

Tools

Search Tool - Multi-strategy video content search:

• Keyword search in subtitles using existing KeywordExtractor with spacy-based NLP, gracefully falling back to simple text matching when spacy is unavailable
• Visual search foundation using keyframe sampling (ready for CLIP integration)
• YouTube URL support with automatic signed URL generation via YTDLPPlugin
• Time range filtering with start/stop parameters for precise navigation
• Multi-video search when no specific video path is provided

Video Sampling Job Tool - Intelligent code generation:

• Auto-proposes Python code using existing SamplerConfig and Worker classes
• Auto-proposes CLI commands for all supported sampling methods (hash, entropy, gzip, buffer, grid)
• Interactive mode asks for user preference when output format isn't specified
• Full parameter customization (frame intervals, hash sizes, buffer sizes, keyframes-only mode)

🏗️ Implementation Details

The MCP server is built on top of existing video-sampler components:

• Leverages video_sampler.language.keyword_capture for subtitle processing
• Uses video_sampler.integrations.yt_dlp_plugin for YouTube support
• Integrates with video_sampler.samplers for frame extraction
• Maintains backward compatibility with all existing functionality

📦 Usage

Start the MCP server:

video_sampler_mcp
# or 
python -m video_sampler.mcp_server.server

Search for content:

{
  "name": "search",
  "arguments": {
    "query": "wildlife documentary",
    "video_path": "https://youtube.com/watch?v=abc123",
    "start": 30.0,
    "stop": 120.0
  }
}

Generate sampling code:

{
  "name": "video_sampling_job", 
  "arguments": {
    "video_path": "/path/to/video.mp4",
    "sampling_method": "hash",
    "output_format": "python",
    "frame_interval": 2.0
  }
}

✅ Testing

• 14/14 tests passing with comprehensive coverage
• Tests include YouTube URL detection, code generation, video loading, subtitle search with time filtering, and error handling
• Includes demonstration script (demo_mcp.py) showcasing all features
• Graceful handling of missing optional dependencies (spacy, yt-dlp, openai)

📚 Documentation

• Complete README in video_sampler/mcp_server/README.md with usage examples
• Inline documentation for all MCP resources and tools
• Integration examples for AI assistants and MCP clients

This implementation provides a standardized, production-ready interface for video content analysis and processing, enabling AI assistants to intelligently navigate and process video content using the full power of the video-sampler framework.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.