Rewrite or edit with Ollama returns "Empty Ollama response" even though Ollama is running. Why?

This happens with thinking models such as Qwen 3 or Gemma 4. They reason before they answer, and Ollama counts the reasoning against the output budget BookyAI sets for each request. Short edits have a small budget, so the model used it all on reasoning and returned no text. Update to version 1.9.4 or later: thinking is now off by default for Ollama, so the whole budget goes into prose. If you prefer to keep thinking on, tick "Let thinking models think before writing" in Settings, under the Ollama model field; if a model then still returns only reasoning, BookyAI tells you so in the error message instead of "Empty Ollama response".