plans: rev 2 — Go rewrite, ontology-first, DB-native config, learning loop
Major revision of the Docker-based homelab OS plan:
1. Go instead of Python — all services rewritten as Go binaries
(Gin web framework, sqlc for DB access, goroutines for probes)
2. Ontology-first design — systems modeling with 3 layers:
- Infrastructure (physical, compute, network, storage, software)
- Governance (identity, secrets, policy)
- Cognition (observation, decision, action, knowledge, learning)
7 Mermaid diagrams: layer map, ER diagram, 3 lifecycle state machines,
feedback loop, policy model
3. DB-native config — inventory.yaml/ontology.yaml/policy.yaml become
seed manifests (bootstrap + DR). The DB is the runtime source of truth,
editable via API. Ontology IS the DB schema (entity_types,
relationship_types, lifecycle_defs tables).
4. Feedback loop — agent learns from execution:
execution → outcome → feedback → pattern → skill → classification
Patterns accumulate from execution history, skills codify proven
procedures, classifier uses pattern confidence for auto-act decisions.
Cold start: agent starts cautious, earns autonomy through evidence.
5. 5 migration groups: ontology meta-schema, entity instances, operations,
learning model, policy. Recursive blast_radius SQL function.
6. Phase 0 added: ontology design before any code.
dtoro
d44979aca7