A defense-grade, edge-cloud hybrid AI assistant built from scratch by Robair Farag. Deployed on NVIDIA Jetson Orin, orchestrated with Kubernetes, secured with JWT, and monitored with Prometheus and Grafana. Live at https://axiom.bitshadow.dev
AXIOM is a production-grade AI assistant system that goes far beyond a simple chatbot. It combines voice interaction, persistent memory, real-time web search, system control, file analysis, a holographic interface with an animated face, multi-agent debate system, morning intelligence briefings, and a 24/7 web watchdog — all running across a hybrid edge-cloud architecture. Built to demonstrate real-world AI engineering skills for roles in tech, aerospace, and defense.
CLOUD LAYER (DigitalOcean Droplet) — REST API, JWT Auth, Rate Limiting, Prometheus Metrics ORCHESTRATION LAYER (Kubernetes K3s) — AXIOM Service, Grafana, Prometheus, Auto-healing DEVELOPMENT LAYER (HP Pavilion i7, Ubuntu 24.04) — Voice Assistant, Wake Word, System Control EDGE LAYER (NVIDIA Jetson Orin Nano) — On-device AI, Offline Capable, ARM Architecture
AI Brain — Powered by Claude Sonnet with streaming responses, natural conversational personality, and context-aware interactions across all sessions.
Persistent Memory — Remembers conversations across sessions using JSON-based memory store with timestamp tracking. Wipe memory on command.
Voice System — Wake word detection (say Hey AXIOM to activate), voice input via Google Speech Recognition, natural voice output via ElevenLabs, fully hands-free operation.
Web Search — Autonomous real-time web search via DuckDuckGo. AXIOM decides when to search without being told and synthesizes results into natural language.
System Control — Live CPU, memory, and disk monitoring. Open applications, create folders, list files, and control volume by voice command.
File Intelligence — Read and summarize PDF documents, Word (.docx) files, and plain text files on command.
Security — JWT authentication on all protected endpoints, role-based access control, rate limiting at 60 requests per minute per IP, environment-based secret management.
Containerization — Full Docker containerization with Docker Compose for local orchestration and an optimized multi-step Dockerfile.
Kubernetes Orchestration — K3s lightweight Kubernetes cluster with Persistent Volume Claims, auto-healing deployments, and service mesh with ClusterIP routing.
Monitoring — Prometheus metrics collection with Grafana visualization dashboard, custom chat request counters, and response time tracking.
Edge AI Deployment — Runs natively on NVIDIA Jetson Orin Nano with ARM architecture support, offline capable with the same codebase as cloud deployment.
Holographic UI — Iron Man inspired interface with animated holographic face, real-time system stats, live chat, voice waveform, and full dark theme. Mobile responsive — works perfectly on any phone or tablet browser.
Animated Face — Real-time animated AI face that reacts when AXIOM speaks. Eyes move, mouth syncs to voice, particles and scan lines create a holographic presence.
Multi-Agent Debate System — Three specialized AXIOM agents (Alpha analytical, Beta creative, Gamma critical) debate any question and synthesize a unified answer. Triggered by the DEBATE button.
Proactive System Monitoring — AXIOM watches your system 24/7 and speaks up on his own when CPU spikes, memory is low, disk is critical, or unusual network activity is detected. No asking required.
Morning Intelligence Briefing — AXIOM scans world news, tech news, and weather every morning and delivers a spoken briefing automatically. Trigger manually by saying brief me.
Web Watchdog — Tell AXIOM to monitor any topic, website, or keyword. He checks every 30 minutes and alerts you the moment something new happens.
Conversation Mode — Toggle hands-free conversation mode for natural back-and-forth voice interaction. AXIOM listens, responds, and prompts you for your next message automatically.
Interrupt Control — Stop AXIOM mid-speech anytime with the stop button. Full control over the conversation flow.
Cloudflare SSL — Production HTTPS deployment via Cloudflare tunnel with automatic SSL certificate and DDoS protection.
24/7 Auto-restart — Systemd service on Droplet ensures AXIOM restarts automatically on any crash or server reboot.
AI Model: Claude Sonnet by Anthropic Voice Input: Google Speech Recognition and PyAudio Voice Output: ElevenLabs TTS Backend: Python and Flask Security: JWT and Flask-CORS Monitoring: Prometheus and Grafana Containerization: Docker Orchestration: Kubernetes K3s Edge Device: NVIDIA Jetson Orin Nano Cloud: DigitalOcean Droplet DNS and SSL: Cloudflare Frontend: HTML, CSS, JavaScript Memory: JSON and ChromaDB Web Search: DuckDuckGo Scheduling: Python Schedule OS: Ubuntu 22.04 and 24.04 LTS
-AXIOM/ ├── src/ │ ├── brain/ │ │ ├── axiom.py # Voice assistant laptop │ │ ├── axiom_headless.py # Cloud API service │ │ ├── axiom_edge.py # Jetson edge deployment │ │ ├── axiom_agents.py # Multi-agent debate system │ │ ├── axiom_monitor.py # Proactive system monitor │ │ └── axiom_butler.py # Morning briefings and watchdog │ ├── voice/ │ │ ├── listener.py # Microphone input │ │ ├── speaker.py # ElevenLabs TTS output │ │ └── wakeword.py # Wake word detection │ ├── memory/ │ │ └── memory.py # Persistent memory system │ ├── tools/ │ │ ├── search.py # Web search │ │ ├── system.py # System control │ │ └── file_reader.py # Document analysis │ ├── security/ │ │ └── security.py # JWT auth and rate limiting │ └── ui/ │ ├── index.html # Holographic interface │ └── face.html # Animated AI face ├── deploy/ │ └── kubernetes/ │ ├── axiom-deployment.yaml │ ├── axiom-service.yaml │ ├── axiom-pvc.yaml │ └── monitoring.yaml ├── Dockerfile ├── docker-compose.yml ├── requirements.txt └── README.md
Clone the repository: git clone https://github.com/Robair26/-AXIOM.git then cd into -AXIOM
Set up environment: python3 -m venv axiom-env then source axiom-env/bin/activate then pip install -r requirements.txt
Configure API keys: cp .env.example .env then add your Anthropic and ElevenLabs API keys
Run voice assistant on laptop: python src/brain/axiom.py
Run as cloud API service: python src/brain/axiom_headless.py
Deploy with Docker: docker build -t axiom . then docker-compose up
Deploy to Kubernetes: kubectl apply -f deploy/kubernetes/
Run on Jetson Edge: python src/brain/axiom_edge.py
GET /health — No auth required — System health check POST /auth — No auth required — Get JWT token POST /chat — JWT required — Send message to AXIOM POST /debate — JWT required — Trigger multi-agent debate POST /speak — JWT required — Text to speech via ElevenLabs GET /stats — JWT required — Get live system stats GET /alerts — No auth required — SSE stream for proactive alerts POST /watch — JWT required — Add topic to watchlist GET /watchlist — JWT required — Get current watchlist POST /briefing — JWT required — Trigger morning briefing now DELETE /memory/clear — JWT required — Wipe AXIOM memory GET /metrics — No auth required — Prometheus metrics
Local Voice Assistant — Runs on laptop with full voice input and output using wake word activation and ElevenLabs voice.
Cloud API on DigitalOcean — Headless Flask API running as a Docker container on a DigitalOcean Droplet with systemd service management for auto-restart and Cloudflare SSL tunnel.
Kubernetes Cluster — Full K3s deployment with AXIOM, Prometheus, and Grafana pods running with persistent storage and auto-healing on any crash.
Edge Device on NVIDIA Jetson Orin Nano — AXIOM Edge runs directly on ARM hardware using the Tegra Orin GPU for local inference with no cloud required.
Access Grafana: kubectl port-forward service/grafana-service 3000:3000 then open http://localhost:3000 and login with admin/admin
Access Prometheus: kubectl port-forward service/prometheus-service 9090:9090 then open http://localhost:9090
All chat and stats endpoints require a valid JWT token obtained via the /auth endpoint. Rate limiting is enforced at 60 requests per minute per IP address. All API keys are stored in environment variables and never committed to version control. Cloudflare tunnel provides additional security layer with DDoS protection.
Say or type these to see AXIOM in action:
brief me — triggers a live morning briefing from real web search watch AI news — AXIOM starts monitoring AI news 24/7 watch SpaceX — AXIOM monitors SpaceX updates and alerts on changes what are you watching — lists all active watchlist topics stop watching AI news — removes topic from watchlist debate: is AI going to replace humans — triggers multi-agent debate how is my system doing — live CPU memory and disk report open firefox — opens application by voice command create a folder called projects — creates folder on your machine what is the latest news in AI — autonomous web search read this file — AXIOM reads and summarizes any document Hey AXIOM — wake word to activate voice assistant on laptop exit — shuts down AXIOM forget — wipes all memory and starts fresh
Robair Farag — AI Engineer and Builder GitHub: https://github.com/Robair26 Project: https://github.com/Robair26/-AXIOM Live: https://axiom.bitshadow.dev
Computer vision via Jetson camera for real-time object detection and face recognition Voice cloning for fully personalized AXIOM voice Autonomous code writing and execution Browser automation and control Gmail and Google Calendar integration