The Architecture
A multi-node infrastructure running local LLMs, semantic memory, agent orchestration, and home automation. Everything here is in active use—not experiments.
Systems
Tools and infrastructure I design, build, and operate daily.
Coquina
Self-Healing Memory Platform for AI Agents (formerly Cortex)
Memory that fixes itself. Every write passes A-MAC admission control — five scoring signals quality-gate what gets in, so the graph never fills with noise. Search runs two paths at once: Postgres full-text and ChromaDB vectors, ranked by recency. An auto-linker wires typed edges between memories with zero manual curation.
Four self-evolution subsystems keep it honest: embedding-drift detection, forgetting metrics, two-stage diffusion retrieval, and graph-diffusion consensus. It runs natively on Apple M4 Silicon with embedded ChromaDB — no Docker, no zombie containers — and heals across three tiers, back in about five seconds.
Forge
Queue-Based Autonomous Worker System
I built this to run while I sleep. 43 autonomous workers on Apple M4 Silicon, coordinated by Redis-backed queues and 81 macOS LaunchAgents — GPU scheduling, health monitoring, and overnight orchestration, all local.
Every night at 5 AM a 28-step DAG runs security review, code review, auto-fix with draft PRs, research scouts across arXiv and the ecosystem, memory consolidation, perf benchmarks, and a morning briefing. A self-healing daemon sweeps every 60 seconds and restarts anything that died.
The sandbox builder ships whole apps on its own: Ollama writes Swift, xcodegen builds it, and a fix loop closes the errors. Zero cloud.
Thalamus
MQTT Gateway to AI Memory
The bridge between the physical house and the memory layer. Sensor readings, automation triggers, and system heartbeats ride MQTT off Home Assistant straight into Coquina, where agents can query the state of the real world like any other memory.
Nerve
MCP Server Registry & Guardian
The guardian for the MCP layer. Nerve holds the canonical registry of server configs, watches for drift, verifies connectivity, and auto-restores anything missing or broken. When a config gets silently wiped, this is what puts it back — the structural safeguard that keeps the multi-agent stack intact.
Forge Command Center
Real-Time System Monitoring
One pane of glass over the whole runtime: live worker status, queue depths, chain run history, and system health. A React front end on the Forge API, auto-refreshing, with task dispatch built in — watch the fleet and push work to it from the same screen.
Hook Pipeline
6-Hook Auto-Persist System
Six Claude Code hooks fire across every session: context recall, a pre-tool safety gate, pre-tool injection, Slack commit alerts, pre-compaction memory extraction, and session-end persistence. Ollama reads each transcript, pulls out structured memories, and writes them to Coquina — no copy-paste, nothing lost between sessions.
Local LLM Fleet
30 Models on M4 Metal
30 models running locally on Apple M4 Silicon via Ollama. Each has a job: forge-gemma4-12b handles code review and local vision, qwen3.5:9b does overnight reasoning, gemma4:e4b runs fast structured/JSON and voice, and nomic-embed-text holds embeddings frozen. Add 5 custom LoRA adapters on top. It powers everything from Home Assistant voice control to autonomous code generation — zero cloud API calls.
System Architecture
The homelab maps to brain structures—each system mirrors a cognitive function.
Philosophy
Everything here exists because I use it. Coquina runs in every coding session. Forge orchestrates overnight chains while I sleep. The homelab hosts real services my household depends on.
Most “AI-powered” features are demos pretending to be products. This is where I test what actually works under real constraints: latency, failure modes, cost, and whether anyone would use it twice.
I build infrastructure so I can recommend it honestly. Every system I suggest to a client is one I've operated myself.
Interested in this kind of infrastructure?
I build these systems for clients too. Let's talk about what you need.
Start a Conversation