Certain API parameters are not currently supported by NovelCrafter. It is impossible to control thinking mode on Gemini Flash or Anthropic Sonnet because the "max_tokens" parameter is not supported. Those two models always default to no reasoning. NovelCrafter should allow you to configure the max_tokens and other API parameters. See here for more info on the API parameters needed to control reasoning on various models: https://openrouter.ai/docs/use-cases/reasoning-tokens#using-max-tokens-for-reasoning