System Overview Hub¶

This hub provides high-level architecture understanding of the Home Security Intelligence system.

Component Architecture¶

Backend Layer Organization¶

flowchart TB
    subgraph Frontend["Frontend Layer"]
        FE["React + TypeScript<br/>Vite + Tailwind + Tremor<br/>:5173 (dev) / :8080 (prod)"]
    end

    subgraph Backend["Backend Layer"]
        BE["FastAPI<br/>Python 3.14+<br/>:8000"]
        WS["WebSocket<br/>/ws/events, /ws/system"]
    end

    subgraph AI["AI Services Layer"]
        DET["YOLO26<br/>Object Detection<br/>:8095"]
        LLM["Nemotron 30B<br/>Risk Analysis<br/>:8091"]
        FLO["Florence-2<br/>Vision Extraction<br/>:8092"]
        CLIP["CLIP ViT-L<br/>Re-identification<br/>:8093"]
        ENR["Enrichment API<br/>Multi-Model Hub<br/>:8094"]
    end

    subgraph Data["Data Layer"]
        PG[(PostgreSQL<br/>:5432)]
        RD[(Redis<br/>:6379)]
        FS[("/export/foscam/<br/>Camera Storage")]
    end

    FE <-->|REST API| BE
    FE <-->|WebSocket| WS
    BE --> DET
    BE --> LLM
    BE --> FLO
    BE --> CLIP
    BE --> ENR
    BE <--> PG
    BE <--> RD
    BE --> FS

Service Inventory¶

Technology Stack¶

flowchart TB
    subgraph Stack["Technology Stack"]
        direction TB

        subgraph FE["Frontend"]
            React["React 19"]
            TS["TypeScript 5.3"]
            Tailwind["Tailwind CSS 3.4"]
            Tremor["Tremor 3.17"]
            Vite["Vite 5.0"]
        end

        subgraph BE["Backend"]
            Python["Python 3.14+"]
            FastAPI["FastAPI 0.104+"]
            SQLAlchemy["SQLAlchemy 2.0"]
            Pydantic["Pydantic 2.0"]
        end

        subgraph DB["Database"]
            PG["PostgreSQL 15+"]
            Redis["Redis 7.x"]
        end

        subgraph ML["AI/ML"]
            PyTorch["PyTorch 2.x"]
            YOLO26["YOLO26"]
            Nemotron["Nemotron-3-Nano-30B"]
            LlamaCpp["llama.cpp"]
        end

        subgraph Infra["Infrastructure"]
            Docker["Docker/Podman"]
            NVIDIA["NVIDIA Container Toolkit"]
            Prom["Prometheus"]
            Graf["Grafana"]
        end
    end

Quick Links¶

Data Flow Summary¶

1. Image Capture: Foscam cameras FTP upload images to /export/foscam/{camera}/
2. Detection: FileWatcher queues images, YOLO26 performs object detection
3. Batching: BatchAggregator groups detections (90s window, 30s idle timeout)
4. Analysis: Nemotron LLM analyzes batches, assigns risk scores
5. Broadcast: Events pushed via Redis pub/sub to WebSocket clients
6. Display: React dashboard updates in real-time

Script Dependencies¶

The project includes various scripts for development, testing, and deployment. This dependency graph illustrates how scripts relate to each other and their execution order.

Related Documentation¶