amalv

If a team adopts this pattern and later decides to remove ModelRiver, how hard is it to unwind? Are the request and event models close to provider APIs or fairly opinionated?

At what point does adding this layer become more complex than just handling streaming failures directly in the app?