Custom model settings for AI providers

Tenant owners can now fine-tune how Tines talks to customer-provisioned models. Head to Settings → AI settings → Providers, then open model actions → Model settings for OpenAI (including OpenAI-compatible endpoints), Anthropic, or AWS Bedrock.

From there, you can:

• Control temperature. Turn it off for models that reject a temperature value, like GPT-5.
• Set thinking mode. Choose Off, Budget, or Adaptive for Claude models on Anthropic and Bedrock. Models that always think only offer Adaptive.
• Choose reasoning efforts. Tick the levels a model accepts — None, Low, Medium, High, or Extra high. The effort picker in Workbench, story copilot, case copilot, and the apps builder will only show those levels.
• Test your changes. Check unsaved settings against the real model, covering plain responses, tool calling, and story copilot. If something fails, you'll see the provider's actual error.