Better usage stats

Feature Description

Currently, the Usage section shows overall token consumption, but it doesn't provide a breakdown by model.

It would be very helpful to include per-model token usage (input, output, and total tokens) so users can:

• Identify which models are consuming the most tokens.
• Better estimate and optimize costs.
• Compare usage across different models.
• Debug unexpected token spikes from specific models or workflows.

This would be especially useful for users who frequently switch between multiple models or providers within the same project.

Use Case

No response

Additional Context

No response

How important is this to you?

Nice to have

首先定位 Usage 部分以及现有的整体 token 消耗聚合。跟踪使用情况的记录和显示方式，然后确定可以在哪里显示模型标识以及输入、输出和总 token 值。完成的标准是用户可以按模型查看这三个指标，同时现有的整体使用情况仍然可用。