# AI Knowledge Distillation Platform Configuration
# تكوين منصة تقطير المعرفة للذكاء الاصطناعي

# System Configuration
system:
  # Memory management settings
  memory:
    max_memory_gb: 14.0  # Maximum memory usage (leave 2GB for system)
    chunk_size_mb: 500.0  # Chunk size for large model loading
    cleanup_threshold: 0.85  # Memory usage threshold for cleanup
    emergency_threshold: 0.95  # Emergency cleanup threshold
  
  # CPU optimization settings
  cpu:
    max_threads: 8  # Maximum number of threads
    use_intel_extension: true  # Use Intel Extension for PyTorch if available
    enable_mkl: true  # Enable Intel MKL
    enable_openmp: true  # Enable OpenMP
  
  # Storage settings
  storage:
    cache_dir: "./cache"
    models_dir: "./models"
    database_dir: "./database"
    logs_dir: "./logs"
    temp_dir: "./temp"
    max_cache_size_gb: 20.0  # Maximum cache size

# Model Loading Configuration
models:
  # Default settings for model loading
  default_settings:
    torch_dtype: "float32"  # Use float32 for CPU
    low_cpu_mem_usage: true
    device_map: "cpu"
    trust_remote_code: false
  
  # Chunk loading settings
  chunk_loading:
    enabled: true
    max_chunk_size_mb: 500.0
    max_cached_chunks: 3
    auto_cleanup: true
  
  # Supported model types
  supported_formats:
    - ".pt"
    - ".pth" 
    - ".bin"
    - ".safetensors"
  
  # Model size limits
  size_limits:
    small_model_mb: 1000  # Models under 1GB load normally
    large_model_mb: 2000  # Models over 2GB use chunking

# Training Configuration
training:
  # Default training parameters
  default_params:
    learning_rate: 0.0001
    batch_size: 4  # Small batch size for memory efficiency
    max_steps: 1000
    temperature: 3.0
    alpha: 0.7
    save_steps: 100
    eval_steps: 50
  
  # Memory optimization during training
  memory_optimization:
    gradient_accumulation_steps: 4
    gradient_checkpointing: true
    mixed_precision: false  # Disable for CPU
    dataloader_num_workers: 2

# Medical Datasets Configuration
medical:
  # Supported medical datasets
  datasets:
    roco_v2:
      repo_id: "eltorio/ROCOv2-radiology"
      streaming_supported: true
      estimated_size_gb: 8.5
    ct_rate:
      repo_id: "ibrahimhamamci/CT-RATE"
      streaming_supported: true
      estimated_size_gb: 12.3
    umie_datasets:
      repo_id: "lion-ai/umie_datasets"
      streaming_supported: true
      estimated_size_gb: 15.7
  
  # DICOM processing settings
  dicom:
    memory_limit_mb: 1000.0
    default_window_center: 40
    default_window_width: 400
    default_output_size: [512, 512]
  
  # Medical preprocessing settings
  preprocessing:
    target_size: [512, 512]
    normalize_images: true
    enhance_contrast: true

# Token Management Configuration
tokens:
  # Encryption settings
  encryption:
    key_file: ".token_key"
    algorithm: "Fernet"
  
  # Token types and their properties
  types:
    read:
      security_level: "medium"
      recommended_for: "development"
    write:
      security_level: "high"
      recommended_for: "production"
    fine_grained:
      security_level: "very_high"
      recommended_for: "enterprise"

# Database Configuration
database:
  # SQLite settings
  sqlite:
    database_dir: "./database"
    backup_interval_hours: 24
    cleanup_days: 30
  
  # Connection settings
  connection:
    timeout: 30
    check_same_thread: false

# Web Server Configuration
server:
  # FastAPI settings
  host: "0.0.0.0"
  port: 8000
  workers: 1  # Single worker for memory efficiency
  reload: false
  
  # CORS settings
  cors:
    allow_origins: ["*"]
    allow_methods: ["GET", "POST", "PUT", "DELETE"]
    allow_headers: ["*"]
  
  # Upload settings
  uploads:
    max_file_size_mb: 5000  # 5GB max file size
    allowed_extensions: [".pt", ".pth", ".bin", ".safetensors"]
    temp_dir: "./temp"

# Logging Configuration
logging:
  # Log levels
  level: "INFO"
  format: "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
  
  # File logging
  file:
    enabled: true
    filename: "logs/app.log"
    max_size_mb: 100
    backup_count: 5
  
  # Console logging
  console:
    enabled: true
    level: "INFO"
  
  # Specific logger levels
  loggers:
    uvicorn: "INFO"
    transformers: "WARNING"
    datasets: "WARNING"
    torch: "WARNING"

# Performance Monitoring
monitoring:
  # System metrics collection
  system_metrics:
    enabled: true
    interval_seconds: 30
    store_in_database: true
  
  # Memory monitoring
  memory_monitoring:
    enabled: true
    alert_threshold: 0.85
    emergency_threshold: 0.95
  
  # Performance recommendations
  recommendations:
    enabled: true
    check_interval_minutes: 5

# Security Configuration
security:
  # Token validation
  token_validation:
    enabled: true
    cache_results: true
    cache_duration_minutes: 60
  
  # File upload security
  file_uploads:
    scan_uploads: true
    max_file_size_mb: 5000
    allowed_mime_types:
      - "application/octet-stream"
      - "application/x-pytorch"

# Feature Flags
features:
  # Advanced features
  memory_management: true
  chunk_loading: true
  cpu_optimization: true
  medical_datasets: true
  token_management: true
  
  # Experimental features
  experimental:
    auto_model_optimization: true
    progressive_loading: true
    smart_caching: true

# Environment-specific overrides
environments:
  development:
    logging:
      level: "DEBUG"
    server:
      reload: true
  
  production:
    logging:
      level: "INFO"
    server:
      reload: false
    security:
      token_validation:
        enabled: true