Latency and Time to First Token

Benchmark real streamed Ollama responses and diagnose configuration that makes an assistant feel slow.

In this module