Claude/folder structure burmese 7apslx

docker/docker-compose-backend.yml

@@ -0,0 +1,101 @@
+version: '3.8'
+
+services:
+  db:
+    image: postgres:16-alpine
+    environment:
+      POSTGRES_DB: namucam
+      POSTGRES_USER: namucam
+      POSTGRES_PASSWORD: namucam
+    volumes:
+      - pg_data:/var/lib/postgresql/data
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U namucam"]
+      interval: 5s
+      timeout: 5s
+      retries: 5
+
+  redis:
+    image: redis:7-alpine
+    volumes:
+      - redis_data:/data
+
+  minio:
+    image: minio/minio
+    command: server /data --console-address ":9001"
+    environment:
+      MINIO_ROOT_USER: minioadmin
+      MINIO_ROOT_PASSWORD: minioadmin
+    volumes:
+      - minio_data:/data
+    ports:
+      - "9000:9000"
+      - "9001:9001"
+
+  go2rtc:
+    image: alexxit/go2rtc
+    ports:
+      - "1984:1984"
+    volumes:
+      - ./go2rtc.yaml:/config/go2rtc.yaml
+
+  backend:
+    build:
+      context: ../src/backend
+      dockerfile: Dockerfile
+    ports:
+      - "8000:8000"
+    environment:
+      DATABASE_URL: postgresql+asyncpg://namucam:namucam@db:5432/namucam
+      REDIS_URL: redis://redis:6379/0
+      MINIO_ENDPOINT: minio:9000
+      MINIO_ACCESS_KEY: minioadmin
+      MINIO_SECRET_KEY: minioadmin
+      GO2RTC_URL: http://go2rtc:1984
+      VLM_BASE_URL: ${VLM_BASE_URL:-http://host.docker.internal:5405}
+      LLM_BASE_URL: ${LLM_BASE_URL:-http://host.docker.internal:5407}
+    depends_on:
+      db:
+        condition: service_healthy
+      redis:
+        condition: service_started
+    volumes:
+      - snapshots:/tmp/snapshots
+
+  worker:
+    build:
+      context: ../src/backend
+      dockerfile: Dockerfile
+    command: celery -A backend.workers.celery_app worker --loglevel=info --concurrency=4
+    environment:
+      DATABASE_URL: postgresql+asyncpg://namucam:namucam@db:5432/namucam
+      REDIS_URL: redis://redis:6379/0
+      MINIO_ENDPOINT: minio:9000
+      MINIO_ACCESS_KEY: minioadmin
+      MINIO_SECRET_KEY: minioadmin
+      GO2RTC_URL: http://go2rtc:1984
+      VLM_BASE_URL: ${VLM_BASE_URL:-http://host.docker.internal:5405}
+      LLM_BASE_URL: ${LLM_BASE_URL:-http://host.docker.internal:5407}
+    depends_on:
+      - db
+      - redis
+    volumes:
+      - snapshots:/tmp/snapshots
+
+  frontend:
+    build:
+      context: ../src/frontend
+      dockerfile: Dockerfile
+      args:
+        NEXT_PUBLIC_API_URL: ${NEXT_PUBLIC_API_URL:-http://localhost:8000}
+        NEXT_PUBLIC_WS_URL: ${NEXT_PUBLIC_WS_URL:-ws://localhost:8000}
+    ports:
+      - "3000:3000"
+    depends_on:
+      - backend
+
+volumes:
+  pg_data:
+  redis_data:
+  minio_data:
+  snapshots:

docker/go2rtc.yaml

@@ -0,0 +1,12 @@
+api:
+  listen: ":1984"
+
+# Streams are added dynamically via the go2rtc REST API
+# Example:
+# streams:
+#   cam_001: rtsp://192.168.1.100/stream1
+#   cam_002: rtsp://192.168.1.101/stream1
+streams: {}
+
+webrtc:
+  listen: ":8555"

skills/analysis/traffic-vlm-analysis/SKILL.md

@@ -0,0 +1,143 @@
+---
+name: traffic-vlm-analysis
+description: "Traffic & public safety VLM analysis — accident, anomaly, and crowd detection using vision-language models"
+version: 1.0.0
+category: analysis
+runtime: python
+entry: scripts/analyze.py
+install: pip
+
+requirements:
+  python: ">=3.9"
+  platforms: ["linux", "macos", "windows"]
+
+parameters:
+  - name: analysis_mode
+    label: "Analysis Mode"
+    type: select
+    options:
+      - "traffic_accident"
+      - "crowd_anomaly"
+      - "suspicious_behavior"
+      - "wrong_way"
+      - "road_obstruction"
+      - "fire_smoke"
+      - "full_scan"
+    default: "full_scan"
+    description: "Focus the VLM on a specific threat type, or run all checks at once"
+    group: Analysis
+
+  - name: sensitivity
+    label: "Sensitivity"
+    type: select
+    options: ["low", "medium", "high"]
+    default: "medium"
+    description: "low = only obvious incidents | high = flag edge cases and near-misses"
+    group: Analysis
+
+  - name: fps
+    label: "Processing FPS"
+    type: select
+    options: [0.1, 0.2, 0.5, 1, 2]
+    default: 0.5
+    description: "Frames per second sent to VLM — VLM is slower than YOLO, keep low"
+    group: Performance
+
+  - name: min_confidence
+    label: "Min Confidence to Report"
+    type: number
+    min: 0.1
+    max: 1.0
+    default: 0.6
+    description: "Suppress incidents below this confidence score"
+    group: Analysis
+
+  - name: yolo_prefilter
+    label: "YOLO Pre-filter"
+    type: boolean
+    default: true
+    description: "Only send frames to VLM when YOLO detects relevant objects (saves VLM tokens)"
+    group: Performance
+
+  - name: camera_location
+    label: "Camera Location Description"
+    type: string
+    default: ""
+    description: "e.g. 'Intersection of Main St and 5th Ave, downtown' — injected into VLM prompt for context"
+    group: Analysis
+
+  - name: language
+    label: "Report Language"
+    type: select
+    options: ["english", "burmese", "both"]
+    default: "english"
+    description: "Language for incident descriptions in VLM output"
+    group: Analysis
+
+capabilities:
+  live_detection:
+    script: scripts/analyze.py
+    description: "Real-time VLM scene analysis for traffic and public safety"
+---
+
+# Traffic VLM Analysis
+
+Real-time traffic and public safety scene analysis using Vision-Language Models (VLMs). Runs on top of the Aegis camera frame pipeline and outputs structured incident events.
+
+## What It Detects
+
+| Mode | Examples |
+|------|---------|
+| `traffic_accident` | Collisions, overturned vehicles, post-crash debris |
+| `crowd_anomaly` | Stampedes, unusually dense crowds, dispersal events |
+| `suspicious_behavior` | Loitering, package abandonment, erratic movement |
+| `wrong_way` | Vehicles driving against traffic |
+| `road_obstruction` | Stopped vehicles blocking lanes, fallen objects |
+| `fire_smoke` | Visible flames, smoke plumes from vehicles or structures |
+| `full_scan` | All of the above in one pass |
+
+## Sensitivity Levels
+
+| Level | Behavior |
+|-------|---------|
+| `low` | Only clear, confirmed incidents (fewer false positives) |
+| `medium` | Balanced — flags near-misses and developing situations |
+| `high` | Flags edge cases, unusual patterns, and potential precursors |
+
+## VLM Compatibility
+
+Works with any OpenAI-compatible vision endpoint:
+- **Local:** Qwen-VL, LLaVA, DeepSeek-VL, SmolVLM (via llama-server or Ollama)
+- **Cloud:** GPT-4o Vision, Claude Sonnet (vision), Gemini Pro Vision
+
+## Protocol
+
+### Aegis → Skill (stdin)
+```jsonl
+{"event": "frame", "frame_id": 42, "camera_id": "cam_001", "frame_path": "/tmp/frame.jpg", "timestamp": "2026-07-01T10:30:00Z", "width": 1920, "height": 1080}
+```
+
+### Skill → Aegis (stdout)
+```jsonl
+{"event": "ready", "model": "qwen-vl", "mode": "full_scan", "sensitivity": "medium", "fps": 0.5}
+
+{"event": "analysis", "frame_id": 42, "camera_id": "cam_001", "timestamp": "...",
+ "incident_detected": true,
+ "incident_type": "traffic_accident",
+ "severity": "high",
+ "confidence": 0.91,
+ "description": "Two-vehicle collision detected at intersection. One vehicle has crossed the center line. Airbags may have deployed.",
+ "objects": ["car", "truck"],
+ "suggested_action": "Dispatch emergency services",
+ "skipped_reason": null}
+
+{"event": "analysis", "frame_id": 43, "camera_id": "cam_001", "timestamp": "...",
+ "incident_detected": false, "skipped_reason": "no_trigger_objects"}
+
+{"event": "error", "message": "VLM endpoint unreachable", "retriable": true}
+```
+
+### Stop Command
+```jsonl
+{"command": "stop"}
+```