What broke when we ran an AI coding agent on local Ollama models (including a 4GB GPU)

We build an AI app builder, and one feature we wanted was "bring your own Ollama": point the agent at your own models and pay nothing for tokens. Getting a chat answer out of a local model is easy. Ge