Local classification in OpenClaw with Ollama and Tev1

Some agent tasks have a small, structured output: choose a department, estimate whether an incident...