Every user intent flows through the Router Kernel, gets classified with confidence scoring, routes to the right specialized hub, and produces governed memory writes with a full audit trail. No black boxes.
No magic. No black box. You describe what you need, LUNA routes it to the right hub, and you get production-ready output with full traceability.
Tell LUNA in plain language: a contract review, a code module, a training plan, a decision framework. Natural language in, structured intent out.
The Router classifies your intent and delegates to the specialized hub: Legal, Coding, Academy, Compass, Sales, or Brain. Context flows automatically.
Get complete deliverables with tests, documentation, audit trails, and full traceability. Every output is governed, every decision is captured.
Specialized hubs
The Router Kernel classifies your intent and routes to the right hub. Hubs do not leak responsibilities. They do not silently override each other. When in doubt, a hub escalates — it does not guess.
Decision scoring — GREEN / YELLOW / RED by founder-dependency.
Knowledge transfer — from tribal knowledge to structured curricula.
Work Orders — describe once, get production-ready code with tests and docs.
Contract review, compliance monitoring — automated with governance trails.
Voice and text interface — local-first with Whisper.cpp and Ollama.
Supply chain security — gates risky dependencies before they land.
Work Order system
Work Orders are the structural unit that distinguishes ODIN from a chatbot. You describe the high-level goal; ODIN decomposes it into Sub-Work Orders and atomic tasks — each traceable back to its source.
Example Work Order
WO-001 Implement user authentication
├── SO-001 JWT middleware
│ ├── T-001 Token validation
│ └── T-002 Refresh logic
├── SO-002 Session store
│ └── T-003 Redis adapter
└── SO-003 Tests + docs
rationale: "Replace session tokens with JWT for stateless scaling"
owner: "engineering:backend"
audit_events: 11
Frequently asked questions
A chatbot is a conversational interface. ODIN is infrastructure. Six specialized hubs with strict contracts, governance-aware memory (BrainDB), and complete audit trails. Every output includes rationale, ownership, and dependencies.
"If ODIN depends on the founder to function, ODIN isn't done yet." Everything must be transferable, documentable — not just explainable — and operable by others without quality degradation.
Compass Hub scores every decision by whether it increases or decreases dependency on the founder. GREEN = reduces dependency. YELLOW = neutral. RED = increases dependency and requires explicit justification.
No. ODIN deploys on your servers. The Assistant Hub uses Whisper.cpp and Ollama locally. Cloud inference is optional and requires explicit opt-in. No background processing without consent.
router.route, hub.process, memory.write, memory.read, approval.request, approval.grant, approval.deny, assistant.capture, sentinel.scan, sentinel.block, sentinel.allow. Every event carries correlation IDs and risk flags.
Let's talk about how ODIN can capture your organization's context and prevent the bottleneck problem.