How does LLM-agnostic positioning interact with on-premise models?

Local models are first-class targets in the routing layer. From the skill's perspective, a local Llama is just another provider option. This makes the on-prem / cloud decision a policy decision per workload rather than a foundational architecture commitment. a