Local compute. Deterministic control. Verifiable outcomes.
Scott Hardie · Solutions Architect · AI Systems Builder · Toronto, Canada
Explore the platform · View the storefront · Connect
AI demos are easy. Production systems must survive retries, partial failure, hostile inputs, runaway spend, model drift, and an auditor asking exactly what happened.
Hardonia is a working portfolio of control planes, runtimes, security boundaries, and commercial systems designed around one idea:
Intelligence can be probabilistic. Infrastructure cannot.
Every decision in the platform flows through TypeSafe JEV — a deterministic decision engine that classifies intent, routes workloads, gates tool calls, and verifies outcomes at $0.042 per million tokens. No black boxes. No vibes-based routing.
JEV runs between every layer of the Hardonia platform:
| Layer | What JEV does | Cost |
|---|---|---|
| Chat widget | Classifies visitor intent in 300ms — canned response or deep inference | $0.00002/msg |
| Context engine | Decides which tool calls to keep or drop during compaction | $0.001/compaction |
| Tool gating | Gates every MCP tool call through allowlist + hard rules before execution | $0.00001/call |
| Caveman compression | 62% token reduction on long sessions | Auto-triggered |
| GPU routing | Routes inference to the right GPU lane (V100 · P40 · RTX 3060) | Zero cost |
The chat widget on aiautomatedsystems.ca classifies every visitor message through JEV before deciding whether to serve a canned response or route to local Ollama inference. Zero cloud API costs.
Seven monorepos, each a coherent subsystem. Every monorepo has an ARCHITECTURE.md showing how it fits the platform.
|
Automation
|
Infrastructure
|
┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ autopilot │ │ agent-infra │ │ agent-edge │
│ ops/finops/ │──│ control- │──│ mesh-edge/ │
│ growth/ │ │ plane/ │ │ pcap/ │
│ support/ │ │ mission- │ └──────────────┘
└──────┬───────┘ │ ledger/ │
│ │ agent-mesh/ │
└──────────│ mcpwall/ │
└──────┬───────┘
│
┌─────▼──────┐
│ model-tools │
│ inference/ │
│ ollama/ │
└──────┬──────┘
│
⚡ JEV decides
│
┌──────▼──────┐
│ GPU FLEET │
│ V100 · P40 │
│ RTX 3060 │
└──────────────┘
| Repo | Stack | What it does |
|---|---|---|
| Settler | TS · TigerBeetle | Reconciliation intelligence and audit OS |
| veridag | Rust · Quint | Formally specified distributed trust DAG |
| Zeo | TS · Edge | Local-first, signed, composable agent pipelines |
| Reach | Rust · Runtime | Deterministic execution and transcript replay |
| Requiem | C++ · Native | Native execution and operator-console contracts |
| truthcore | Python | Verification kernel and evidence reports |
| ReadyLayer | TS · CI | Delivery governance and provenance export |
| MortgageMatchPro | TypeScript | Mortgage matching platform |
| Keys | TypeScript | Auditable mission control for constrained agents |
| Nautilus | Docker | Containerized operational AI infrastructure |
| TokenGoblin | Go · ClickHouse | AI token-spend observability and routing guardrails |
| SawyerCore | Node · Python | Deterministic edge-AI runtime and simulation engine |
| WorldForge | Rust | Deterministic, moddable simulation operating system |
| World26 | Python | Open planetary-systems simulator |
| FlexibleAccessible | TypeScript | WCAG accessibility compliance |
| Repo | What it does |
|---|---|
| ai-lab | Lab config, scripts, monitoring, GPU fleet management |
| agent-governance | Agent laws, spec, 159 tests, 35 spec sections |
| hardonia-checkout-api | Stripe checkout + webhook verification |
| JupyterNotebooks | Applied AI notebooks: quantization, vision, fine-tuning |
| Repo | What it does |
|---|---|
| hardonia-store | Storefront · aiautomatedsystems.ca |
| comfyui-workflow-packs | 20+ ComfyUI workflow packs on Gumroad |
| content-repo | 28 SEO blog posts, email sequences, social content |
| ai-prompt-templates | 200+ tested prompt templates |
| ai-ops-toolkit | CLI tools for AI lab operations |
The Hardonia AI lab runs a 3-GPU sovereign stack with 2,354 ComfyUI nodes across video generation, image processing, and audio synthesis — all running locally with zero cloud API costs.
| GPU | VRAM | Role | Model Capacity |
|---|---|---|---|
| NVIDIA V100 | 32 GB | Heavy inference · Wan 2.2 14B video | 14B params |
| NVIDIA P40 | 24 GB | ComfyUI primary · HunyuanVideo · LoRA training | 8.3B params |
| NVIDIA RTX 3060 | 12 GB | Vision models · embeddings · LTX 2.3 | 5B params |
Live capabilities:
| Capability | Model | Nodes | Output |
|---|---|---|---|
| Product photo → video | Wan 2.2 | 168 nodes | 5-sec rotating product demo |
| Text → video + audio | LTX 2.3 | 111 nodes | Synced video + narration |
| Photo → cinematic video | HunyuanVideo 1.5 | 19 nodes | 1080p with super-resolution |
| Image → portrait | SDXL + ControlNet | 400+ nodes | Professional headshots |
| Batch product photos | ComfyUI + IP-Adapter | 300+ nodes | 100+ images/hour |
Infrastructure: Ollama 4-lane inference · LiteLLM routing · Prometheus + Grafana monitoring · NATS event mesh · n8n automation · Agent governance (159 tests, 35 spec sections)
The Video Gen API on
:8085wraps all video models behind a single REST endpoint. Upload a product photo, pick a style, get a video.
| Principle | Working rule | |
|---|---|---|
| 🔍 | Evidence over confidence | If a run cannot be inspected or replayed, it is not production-ready. |
| 🏠 | Local-first by design | Own the compute, data boundary, fallback path, and cost model wherever practical. |
| 🔒 | Determinism at the edges | Keep probabilistic intelligence inside explicit policy, schema, and execution constraints. |
| ⚙️ | Boring reliability wins | Idempotency, RLS, state machines, and observable queues beat clever hidden behavior. |
| 💵 | Revenue is a reconciled event | A dashboard row is not money; provider-correlated settlement evidence is money. |
| ⚡ | JEV before vibes | Every decision is classified, routed, and verified — not guessed. |
If you are working on a serious AI, SaaS, integration, reliability, or revenue system, start with a specific bottleneck, a measurable outcome, and a verifiable path to production.
© Scott Hardie · Hardonia Sovereign Systems · Toronto, Canada