OCN Hybrid Inference Architecture

Also: OCN Hybrid Inference Architecture — Local Bonsai + Cloud Groq Task Routing

Frameworkverbatimlex:ocn-hybrid-inference-architecture

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.

Quoted verbatim from Notion, captured 2026-07-25. First seen 2026-03-31.

Provenance

Every source that names this term. Links into the Notion workspace require workspace access.

Notionapp.notion.com/p/225eef13788641d584eb87799fd3fc08 (private workspace)2026-07-25

Cite

OCN Hybrid Inference Architecture — lex:ocn-hybrid-inference-architecture — Neuruh Lexicon v1.0.0 (0a2bb72426e9) — https://lexicon.neuruh.com/t/ocn-hybrid-inference-architecture

Agents: GET /api/term/lex:ocn-hybrid-inference-architecture or the lexicon_resolve MCP tool.