{
  "term": {
    "id": "lex:ocn-hybrid-inference-architecture",
    "slug": "ocn-hybrid-inference-architecture",
    "term": "OCN Hybrid Inference Architecture",
    "aliases": [
      "OCN Hybrid Inference Architecture — Local Bonsai + Cloud Groq Task Routing"
    ],
    "kind": "canon",
    "type": "framework",
    "definition": "Hybrid local/cloud inference architecture for the OCN stack. Routes tasks by complexity: Bonsai 4B/8B (local, zero cost) handles classification, scoring, short inference. Groq Llama 3.3 70B (cloud) handles heavy reasoning as fallback. Hot-swappable LoRA adapter routing enables domain-specific inference without model retraining. Multi-agent local stack: 4-5 Bonsai 4B instances (0.57GB each) run simultaneously on M1. Apache 2.0 base model clears white-label path.",
    "definition_basis": "verbatim",
    "definition_source": "https://app.notion.com/p/225eef13788641d584eb87799fd3fc08",
    "provenance": [
      {
        "source": "notion",
        "ref": "https://app.notion.com/p/225eef13788641d584eb87799fd3fc08",
        "captured_at": "2026-07-25"
      }
    ],
    "related": [],
    "status": "published",
    "first_seen": "2026-03-31",
    "source_ids": [
      "ip:OCN-FW-HYBRID-001"
    ]
  },
  "url": "https://lexicon.neuruh.com/t/ocn-hybrid-inference-architecture",
  "cite": "lex:ocn-hybrid-inference-architecture — Neuruh Lexicon v1.0.0",
  "lexicon_version": "1.0.0",
  "content_hash": "0a2bb72426e976507b89a3172a8c5671b6fe136f6e54ce8499ec1d890ac7e851"
}