MCP Video Parser

# Configuration Guide ## Configuration File The main configuration file is located at `config/default_config.json`. You can also create environment-specific configs: - `config/config.json` - Custom configuration (overrides defaults) - `config/production.json` - Production settings - `config/development.json` - Development settings ## Configuration Options ### Storage Settings ```json { "storage": { "base_path": "./video_data", // Where to store all video data "max_cache_size_gb": 50, // Maximum cache size before cleanup "cleanup_after_days": 30 // Delete old processed data after N days } } ``` **Environment Variables:** - `VIDEO_DATA_PATH` - Override base_path ### Processing Settings ```json { "processing": { "frame_sample_rate": 30, // Extract 1 frame every N frames "enable_scene_detection": true, // Auto-detect scene changes "scene_threshold": 0.3, // Sensitivity (0.0-1.0, lower = more sensitive) "max_frames_per_video": 1000, // Maximum frames to extract per video "frame_quality": 85, // JPEG quality (1-100) "frame_format": "jpg", // Frame format (jpg, png) "audio": { "model": "base", // Whisper model size "language": "en", // Audio language "enable_timestamps": true // Include word timestamps } } } ``` ### LLM Settings ```json { "llm": { "ollama_host": "http://localhost:11434", // Ollama server URL "vision_model": "llava:latest", // Model for analyzing frames "text_model": "llama2:latest", // Model for text generation "embedding_model": "nomic-embed-text", // Model for embeddings "max_context_length": 4096, // Maximum context window "temperature": 0.7, // Generation temperature "timeout_seconds": 60 // Request timeout } } ``` **Environment Variables:** - `OLLAMA_HOST` - Override Ollama server URL ### Performance Settings ```json { "performance": { "max_concurrent_videos": 2, // Process N videos simultaneously "frame_batch_size": 10, // Analyze N frames per batch "enable_gpu": true, // Use GPU acceleration if available "thread_count": 4 // Worker threads for processing } } ``` ### Logging Settings ```json { "logging": { "level": "INFO", // DEBUG, INFO, WARNING, ERROR "format": "json", // json or text "file": "logs/mcp-video-server.log" } } ``` ## Configuration Examples ### High-Quality Analysis For detailed analysis with more frames: ```json { "processing": { "frame_sample_rate": 15, "max_frames_per_video": 2000, "frame_quality": 95, "scene_threshold": 0.2 } } ``` ### Storage Optimization For limited storage environments: ```json { "storage": { "max_cache_size_gb": 10, "cleanup_after_days": 7 }, "processing": { "frame_sample_rate": 60, "max_frames_per_video": 500, "frame_quality": 70, "frame_format": "jpg" } } ``` ### Fast Processing For real-time or near real-time processing: ```json { "processing": { "frame_sample_rate": 120, "enable_scene_detection": false, "max_frames_per_video": 100, "audio": { "model": "tiny", "enable_timestamps": false } }, "llm": { "timeout_seconds": 30, "temperature": 0.5 } } ``` ### Multi-Camera Setup For surveillance or multi-camera environments: ```json { "storage": { "base_path": "/mnt/surveillance/video_data", "max_cache_size_gb": 500, "cleanup_after_days": 90 }, "performance": { "max_concurrent_videos": 4, "frame_batch_size": 20, "thread_count": 8 } } ``` ## Location Configuration Locations are auto-detected from video paths or can be specified during processing. Common patterns: - `/path/to/shed/video.mp4` → Location: "shed" - `/cameras/front_door/recording.mp4` → Location: "front_door" - Custom: Specify with `--location` flag ## Advanced Configuration ### Custom Model Configuration To use different models: ```bash # Pull custom models ollama pull llava:13b ollama pull mistral:latest # Update config { "llm": { "vision_model": "llava:13b", "text_model": "mistral:latest" } } ``` ### Network Configuration For remote Ollama servers: ```json { "llm": { "ollama_host": "http://192.168.1.100:11434" } } ``` ### Resource Limits To limit resource usage: ```json { "performance": { "max_concurrent_videos": 1, "frame_batch_size": 5, "enable_gpu": false, "thread_count": 2 }, "processing": { "max_frames_per_video": 200 } } ``` ## Monitoring Configuration Check current configuration: ```bash # View effective configuration python -c "from src.utils.config import get_config; print(get_config().model_dump_json(indent=2))" # Test configuration python scripts/test_config.py ``` ## Configuration Validation The system validates configuration on startup: - Path existence and permissions - Model availability - Ollama connectivity - Storage space Invalid configurations will show clear error messages with suggestions for fixes.