Hardware
Gigabyte Brix — FeroServer
Intel Core i7-8550U · 15GB RAM · Arch Linux. The main workhorse — runs Hermes, PolyWhales, ATLAS V2, and all Cloudflare tunnels 24/7 as a headless server.
Raspberry Pi 4 — 8GB
Headless Raspberry Pi 4 dedicated to running a personal Minecraft server. Online since January 2025 — the first real Linux/self-hosted infrastructure project.
Mac Mini — Prometheus Node
Arriving soon. Will take over as the AI inference node — running Ollama with local LLMs for Prometheus, the autonomous trade decision agent.
Active Services
Hermes
Kalshi prediction market trading bot. FastAPI backend, SQLite database, live WebSocket dashboard. Systemd service, auto-restarts on failure.
View Project →PolyWhales
Real-time Polymarket whale scanner. Streams on-chain events from Polygon via Alchemy WebSocket, scores wallets by win rate, surfaces signals.
View Project →ATLAS V2
Autonomous AI Twitch streamer. Groq LLM inference, Piper TTS, FAISS memory, OBS WebSocket control, self-learning pipeline.
View Project →Cloudflare Tunnels
Named tunnel routing all services to public subdomains — hermes.ferocity.trade, whales.ferocity.trade, ferobuilds.dev. Zero open ports.
Software Stack
OS & System
Arch Linux on the Brix, Raspberry Pi OS on the Pi 4. All services managed via systemd with linger enabled for boot-time autostart.
Backend
Python 3 across all projects — FastAPI for web servers, SQLite for persistence, asyncio for concurrent loops, uvicorn for ASGI serving.
AI & Inference
Groq API (Llama 3.3 70B) for ATLAS, Anthropic Claude for Prometheus (planned). Ollama with llama3.2:3b ready for local inference on the Mac Mini.
Networking
Cloudflare for DNS, tunnels, and SSL. Alchemy for Polygon RPC and WebSocket. Zero exposed ports on the home network.