Sujit224/aria-response
GitHub: Sujit224/aria-response
一套融合计算机视觉与大语言模型的多模态安防应急响应系统,实现对 CCTV 视频流和住客聊天中危机事件的自动检测、分级与实时协调通知。
Stars: 0 | Forks: 0
# aria-response
## 目录结构参考
```
aria-response/
├── .env.example
├── .env
├── requirements.txt
├── README.md
│
├── app/ # FastAPI backend
│ ├── __init__.py
│ ├── main.py # Entry point, lifespan, routes mount
│ │
│ ├── api/
│ │ ├── __init__.py
│ │ └── routes.py # REST: incidents, ack, session history
│ │
│ ├── db/
│ │ ├── __init__.py
│ │ └── session.py # Async engine, get_db, init_db
│ │
│ ├── models/
│ │ ├── __init__.py
│ │ ├── tables.py # SQLAlchemy: all 12 DB tables
│ │ └── schemas.py # Pydantic: PipelineState + all payloads
│ │
│ ├── graph/ # Chat detection LangGraph pipeline
│ │ ├── __init__.py
│ │ ├── pipeline.py # Graph wiring + conditional routing
│ │ └── nodes/
│ │ ├── __init__.py
│ │ ├── enricher.py # Attach room/block/floor from guest profile
│ │ ├── nlp_classifier.py # Claude-powered threat classification
│ │ ├── zone_resolver.py # Map to zones 1/2/3, persist Incident
│ │ ├── llm_responder.py # Generate all role-specific messages
│ │ └── alert_dispatcher.py # Redis pub/sub + DB dispatch logs
│ │
│ ├── vision/ # YOLO detection pipeline
│ │ ├── __init__.py
│ │ ├── schemas.py # YOLODetection, ThreatEvent, ContextFilterResult
│ │ ├── pipeline_state.py # VisionPipelineState
│ │ ├── pipeline.py # Vision LangGraph graph
│ │ ├── camera_worker.py # RTSP reader, YOLOv8 inference, per-frame logic
│ │ ├── camera_manager.py # Loads all active cameras, spins up workers
│ │ ├── context_filter.py # Guard post suppression + SuppressionLog
│ │ ├── threat_classifier.py # YOLO class → ThreatEvent + severity
│ │ ├── zone_resolver.py # Vision-path zone resolver node
│ │ └── llm_responder.py # Vision-path LLM responder node
│ │
│ ├── ws/
│ │ ├── __init__.py
│ │ └── chat.py # WebSocket handler + Redis listener
│ │
│ └── services/
│ └── __init__.py # (reserved: ack watchdog, push notify)
│
├── alembic/ # DB migrations
│ ├── env.py
│ ├── script.py.mako
│ └── versions/
│ └── 001_initial_schema.py
│
├── frontend/
│ │
│ ├── guest-pwa/ # Guest emergency chat (installable PWA)
│ │ ├── package.json
│ │ ├── vite.config.js
│ │ ├── index.html
│ │ ├── public/
│ │ │ ├── icon-192.png
│ │ │ └── icon-512.png
│ │ └── src/
│ │ ├── main.jsx
│ │ ├── hooks/
│ │ │ └── useARIASocket.js # WebSocket hook with auto-reconnect
│ │ ├── lib/
│ │ │ └── session.js # Session/guest ID helpers, getVenueId
│ │ ├── components/
│ │ │ ├── SOSButton.jsx # One-tap panic button
│ │ │ ├── AlertBanner.jsx # Severity-colored alert overlay
│ │ │ ├── ChatBubble.jsx # Message thread bubble
│ │ │ └── StatusBar.jsx # Connection status + room location
│ │ └── pages/
│ │ └── GuestChat.jsx # Main PWA screen
│ │
│ └── staff-dashboard/ # Staff ops dashboard
│ ├── package.json
│ ├── vite.config.js
│ ├── index.html
│ └── src/
│ ├── main.jsx
│ ├── hooks/
│ │ └── useStaffSocket.js # Staff WebSocket hook
│ ├── lib/
│ │ └── api.js # REST client: incidents, resolve, ack
│ ├── components/
│ │ ├── IncidentCard.jsx # Severity-colored incident list card
│ │ ├── IncidentDetail.jsx # Zone map, dispatch log, resolve button
│ │ ├── FloorMap.jsx # SVG zone 1/2/3 visualizer
│ │ ├── DispatchLog.jsx # Per-incident ack tracking table
│ │ └── StatusBar.jsx # Live connection + venue status
│ └── pages/
│ ├── Dashboard.jsx # Main page: live feed + detail panel
│ └── Hotel3D.jsx # Embedded 3D hotel navigator
│
└── public/
└── hotel3d.html # Standalone 3D hotel navigator (Three.js)
```
标签:AIoT, AI危机检测, AI安防, AV绕过, CCTV视频监控, Claude, CVE检测, FastAPI, LangGraph, Pydantic, Redis, SQLAlchemy, YOLO, 人工智能安全, 公共安全, 合规性, 员工协调, 园区安全, 威胁分类, 实时报警, 实时通信, 库, 应急响应, 应急响应系统, 应急指挥调度, 异步数据库, 搜索引擎查询, 智能安防, 目标检测, 聊天监控, 视频分析, 计算机视觉, 逆向工具, 酒店安全