Hacker News Reader

Does 2.0 expose explicit capability negotiation, or does it infer capabilities from the selected model?

We have an extensive model registry with cost and capability tracking:

thx