An enterprise-grade, hybrid cloud-to-edge multi-agent security infrastructure designed to automate physical checkpoint auditing, biometrics, and threat detection in real time. Built under Track 3: Security & Compliance for the LabLab AI Hackathon.
VigilOS bridges on-site physical security hardware with secure cloud networks and decentralized agent protocols, minimizing local exposure while maximizing analytical computing power.
- Demo Application URL: VigilOS - Live Security Operations Dashboard
- Decentralized Node Profile: Ralphi IA on BAND Network
The ecosystem relies on an asynchronous, highly available infrastructure decoupled into three distinct layers:
- Speechmatics Streaming ASR: Captures real-time audio streams directly from physical gate intercoms, performing zero-latency speech-to-text to extract visitor data, intentions, and credentials without manual keyboard entry.
- Cloudflare Zero Trust Tunnels: Establishes a secure, outbound-only encrypted tunnel from on-site edge hardware (Hikvision cameras and local servers). This allows cloud components to pull biometric feeds securely without opening incoming firewall ports.
- AI/ML API: Functions as our centralized cognitive traffic controller. It orchestrates and triggers multiple LLM agents in parallel (delegating heavy vision/LPR analysis to Gemini Pro and complex risk scoring loops to Claude 3.5 Sonnet).
- Supabase Cloud: Acts as our real-time persistence layer. Edge deltas are securely pushed here to update the live operations dashboard and maintain an unalterable log audit trail.
- Triggerware Middleware Engine: Polls database state changes continuously. Whenever a high-risk perimeter breach or anomalous entry is recorded, it fires asynchronous worker pipelines.
- Bright Data (Web Data API): Conducts surgical, non-intrusive OSINT and public verification scans via residential proxies to check background indicators for flag anomalies.
- Atlassian Jira Integration: Automatically maps threat cards and injects them as critical compliance/security incidents within Jira Cloud for institutional visibility and auditing.
The engineering lifecycle of VigilOS utilized a highly iterative, dual-environment AI-assisted pipeline to evaluate model constraints, context-window sizes, and execution speeds.
This current repository houses the fully stabilized, production-ready backend built with FastAPI and Supabase. Cursor was selected as the definitive development platform due to its deep local codebase indexation and large context window handling. This environment enabled rapid debugging of multi-endpoint API integration payloads and middleware data mapping in record time.
In parallel, initial discovery, exploratory agent testing, and modular model behavior benchmarking were developed natively on NativelyAI.com and deployed utilizing Google’s AntiGravity agent workflows.
If you wish to examine our autonomous deployment experiments and secondary pipeline testing repos, you can access the alternative repository here: 👉 VigilOS - Experimental AntiGravity Deployment Repository (Note: Core production features remain maintained in the current Cursor branch).
├── .gitignore # Strict exclusion patterns for secrets management
├── main.py # FastAPI Main Application Gateway
├── core/
│ ├── config.py # Environment validation layer
│ ├── database.py # Supabase & Client connection pooling
│ └── security.py # Cloudflare payload validation & token authentication
├── services/
│ ├── aiml_orchestrator.py # AI/ML API dynamic model switcher (Gemini / Claude)
│ ├── speechmatics_asr.py # Streaming audio transcription service
│ ├── triggerware_hook.py # Asynchronous Jira event emitter payload
│ └── brightdata_osint.py # Automated web scraping and verification loops
└── requirements.txt # Production software dependencies