[REQUEST] local-llm-integration — Ollama, llama.cpp, on-device AI

## Skill: local-llm-integration
**Category**: offline | **Geo**: Global | **Priority**: Medium

Guide integration modeles locaux: Ollama, llama.cpp, llama.rn.
Choix de modele selon RAM/device. Matrice taille × performance.