@startuml Crowd Analyzer Architecture

  title Архитектура Crowd Analyzer

  package "External Systems" {
    [Redis Queue] as redis
    [Storage Service] as storage
    [Tevian API] as tevian_api
    [Camera Server] as camera_server
  }

  package "Crowd Analyzer" {
    [analyze_crowd()] as entry_point
    [_analyze_crowd()] as main_logic
    [_run_detectors()] as detector_runner
    [_get_triggered_zones()] as trigger_logic
    [_build_zones_info()] as zones_builder
  }

  package "Detectors" {
    [TevianHeadsDetector] as tevian_detector
  }

  package "Data Access" {
    [frames.pull()] as frame_puller
    [central.send()] as central_sender
    [services_db] as services_db
    [zones_db] as zones_db
  }

  package "Storage" {
    [MongoDB Local] as mongo_local
    [MongoDB Crowd] as mongo_crowd
  }

  ' External connections
  redis --> entry_point : tasks
  camera_server <-- frame_puller : get frames
  tevian_api <-- tevian_detector : AI requests
  storage <-- main_logic : upload images

  ' Internal flow
  entry_point --> main_logic
  main_logic --> detector_runner
  main_logic --> zones_builder
  main_logic --> trigger_logic
  detector_runner --> tevian_detector

  ' Data access
  frame_puller --> camera_server
  central_sender --> redis : results
  services_db --> mongo_local
  zones_db --> mongo_local

  ' Key relationships
  main_logic --> frame_puller : get frames
  main_logic --> central_sender : send results
  main_logic --> zones_db : update status
  detector_runner --> services_db : check config
  tevian_detector --> tevian_api : detect heads

  note right of entry_point
    Entry point:
    - Semaphore control
    - Error handling
    - Metrics tracking
  end note

  note right of main_logic
    Main logic:
    - Frame processing
    - Zone analysis
    - Result storage
    - Notification sending
  end note

  note right of detector_runner
    Detector runner:
    - AI service calls
    - Error handling
    - Performance tracking
  end note

  @enduml

  Диаграмма состояний зоны:

  @startuml Zone State Diagram

  title Состояния зоны в процессе анализа

  [*] --> Inactive : zone created

  state Inactive {
    Inactive : length_by_ai = 0
    Inactive : no triggers
  }

  state Active {
    Active : length_by_ai > 0
    Active : analyzing people count
  }

  state Triggered {
    Triggered : trigger condition met
    Triggered : schedule active
    Triggered : not in grace period
  }

  state GracePeriod {
    GracePeriod : trigger condition met
    GracePeriod : but in grace period
    GracePeriod : suppressing notifications
  }

  Inactive --> Active : people detected
  Active --> Inactive : no people detected
  Active --> Triggered : trigger_at threshold reached\nAND schedule active\nAND not in grace period
  Active --> GracePeriod : trigger_at threshold reached\nBUT in grace period
  Triggered --> Active : trigger condition not met
  GracePeriod --> Triggered : grace period expired\nAND trigger still met
  GracePeriod --> Active : trigger condition not met

  note right of Triggered
    Actions:
    - Zone highlighted on image
    - Notification sent
    - Grace period started
  end note

  note right of GracePeriod
    Grace period prevents
    spam notifications for
    zones that constantly
    trigger
  end note

  @enduml