api_version: cmvr.edge.ai/v1

runtime:
  # PyAV plus the PPE and phone-use YOLO branches share this bounded pool.
  thread_workers: 4
  shutdown_timeout_s: 8

endpoints:
  cmvr_es:
    transport: grpc
    # cmvr-es gRPC address. Change this value for each deployed robot.
    target: 192.168.0.119:50052
    tls: false
    timeout_s: 5
    options:  
      max_receive_mb: 32

  ppe_alert_platform:
    transport: http
    # Detection-alert platform HTTP base URL. The alert payload model_id and
    # rule_id distinguish PPE violations from phone-use violations.
    base_url: http://192.168.0.222:13080
    timeout_s: 3

pipelines:
  detection:
    enabled: true
    nodes:
      camera:
        uses: cmvr.grpc.camera_rgb_stream@1
        with:
          endpoint: cmvr_es
          device_id: wrist_cam
          pixel_format: BGR8
          reconnect: true
          reconnect_initial_s: 0.5
          reconnect_max_s: 10
          # Log stream state immediately on first frame and emit a periodic
          # progress/stall heartbeat without printing every encoded frame.
          stream_log_interval_s: 5

      decoder:
        uses: media.video_decoder.pyav@1

      detector:
        uses: detection.model@1
        with:
          model: construction-ppe-yolov8@1
          # Omitting detect_labels means all registered labels.  This example
          # asks the backend to return only PPE violations used by the rules.
          detect_labels:
            - No-Boots
            - No-Ear-Protection
            - No-Glass
            - No-Glove
            - No-Helmet
            - No-Mask
            - No-Vest
          confidence: 0.50
          max_fps: 10
          # Log the first completed inference immediately, then aggregate one
          # heartbeat every 5 seconds so model activity is visible without
          # printing every frame. Set to 1 for one-second debugging, or omit to disable.
          inference_log_interval_s: 5
          # Keep the decoded threshold frame available to repeat_gate so an
          # annotated alert image can be rendered only when a rule triggers.
          attach_frame: true
          model_options:
            # Model artifact and inference device are deployment configuration,
            # not process environment requirements. This repository-relative
            # path requires launching cmvr-edge-ai from the repository root.
            weights: models/detection/construction-ppe-yolov8/v1/best.pt
            device: cpu
            imgsz: 640
            iou: 0.70
            half: false
            max_det: 100

      repeat_gate:
        uses: detection.repeat_gate@1
        with:
          # cmvr-es currently omits capture timestamps on successful stream
          # frames, so received time is the deterministic deployment default.
          time_source: received
          alert_image:
            enabled: true
            jpeg_quality: 85
          rules:
            - id: no-boots
              labels: [No-Boots]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source
            - id: no-ear-protection
              labels: [No-Ear-Protection]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source
            - id: no-glass
              labels: [No-Glass]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source
            - id: no-glove
              labels: [No-Glove]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source
            - id: no-helmet
              labels: [No-Helmet]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source
            - id: no-mask
              labels: [No-Mask]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source
            - id: no-vest
              labels: [No-Vest]
              min_hits: 3
              window_ms: 2000
              cooldown_ms: 30000
              min_confidence: 0.50
              scope: source

      # phone_detector:
      #   uses: detection.model@1
      #   with:
      #     model: people-talking-yolov8x@1
      #     # The checkpoint also contains a generic class named "label". It is
      #     # retained in model registration for class-ID safety but is not an
      #     # actionable phone-use event, so this branch selects only class 1.
      #     detect_labels:
      #       - talking on phone
      #     confidence: 0.50
      #     # YOLOv8x is substantially heavier than the PPE model. Start with a
      #     # conservative CPU rate and tune only after measuring target hardware.
      #     max_fps: 5
      #     inference_log_interval_s: 5
      #     attach_frame: true
      #     model_options:
      #       weights: models/detection/people-talking-yolov8x/v1/best.pt
      #       device: cpu
      #       imgsz: 640
      #       iou: 0.70
      #       half: false
      #       max_det: 100

      # phone_repeat_gate:
      #   uses: detection.repeat_gate@1
      #   with:
      #     time_source: received
      #     alert_image:
      #       enabled: true
      #       jpeg_quality: 85
      #     rules:
      #       - id: talking-on-phone
      #         labels: [talking on phone]
      #         min_hits: 3
      #         window_ms: 2000
      #         cooldown_ms: 30000
      #         min_confidence: 0.50
      #         scope: source

      alert_platform:
        uses: platform.http_json_sink@1
        with:
          endpoint: ppe_alert_platform
          path: /v1/detection-alerts
          # Platform outages must not stop camera capture or inference. After
          # bounded retries, log a WARNING and drop only this report.
          failure_mode: log_and_drop
          max_attempts: 3
          retry_initial_s: 0.25
          retry_max_s: 2

    edges:
      # Encoded H264/H265 packets must remain contiguous before decode.
      - from: camera.frames
        to: decoder.frames
        qos:
          profile: video_contiguous
          capacity: 8
          overflow: block

      # Once frames are decoded, keeping only the newest frame bounds latency.
      - from: decoder.frames
        to: detector.frames
        qos:
          profile: realtime_latest
          capacity: 1
          overflow: drop_oldest

      # Fan out the already decoded image; do not open a second camera stream
      # or decode the same H264/H265 packet twice.
      - from: decoder.frames
        to: phone_detector.frames
        qos:
          profile: realtime_latest
          capacity: 1
          overflow: drop_oldest

      - from: detector.detections
        to: repeat_gate.detections
        qos:
          profile: telemetry
          # DetectionResult carries a decoded frame when attach_frame is enabled;
          # keep this queue short so raw image buffers cannot accumulate. Under
          # sustained overload drop_oldest also means dropped results do not count.
          capacity: 2
          overflow: drop_oldest

      # - from: phone_detector.detections
      #   to: phone_repeat_gate.detections
      #   qos:
      #     profile: telemetry
      #     # attach_frame carries the decoded image until the rule is evaluated.
      #     capacity: 2
      #     overflow: drop_oldest

      - from: repeat_gate.alerts
        to: alert_platform.input
        qos:
          profile: telemetry
          capacity: 64
          overflow: block

      # - from: phone_repeat_gate.alerts
      #   to: alert_platform.input
      #   qos:
      #     profile: telemetry
      #     capacity: 64
      #     overflow: block

  talk:
    enabled: true
    nodes:
      audio_stream_placeholder:
        # Replace with cmvr.grpc.microphone_audio_stream@1 when its proto lands.
        uses: core.sequence_source@1
        with:
          items:
            - simulated-audio-chunk
          schema_name: AudioChunk
          schema_version: 1

      dialogue_placeholder:
        # The real chain will be VAD -> ASR -> dialogue -> TTS.
        uses: core.passthrough@1

      output:
        uses: core.log_sink@1
        with:
          logger: cmvr_edge_ai.talk

    edges:
      - from: audio_stream_placeholder.output
        to: dialogue_placeholder.input
        qos:
          profile: audio_contiguous
          capacity: 16
          overflow: block

      - from: dialogue_placeholder.output
        to: output.input
        qos:
          profile: request
          capacity: 8
          overflow: block