Usage Info Display
- Chat UI: Each assistant message now shows response time, token counts (input → output, thinking), and estimated cost - Workflow Execution: Command node details display usage info in the execution progress modal - Execution History: Usage info shown per step in the history modal and included in Canvas exports
- Workflow usage info display: Token usage, cost, and response time are now shown in workflow generation (Notice popup), execution modal (per node), history modal (per step), and Canvas export
Bug Fix: enableThinking not respected
Now explicitly sets thinkingBudget: 0 to suppress thinking when disabled, and respects the explicit option over keyword detection.
- Thinking mode not disabled when explicitly turned off: enableThinking: false in workflow command nodes was ignored because Gemini 2.5+/3 models have thinking enabled by default. Now explicitly sets thinkingBudget: 0 to disable
- Unnecessary thinking in simple chat: Thinking was active even without trigger keywords. Now fully keyword opt-in — no keyword match means thinking is disabled
- Thinking always enabled for workflow generation: AI workflow generation now always uses thinking mode for better quality output, regardless of keywords
