[BUG] How should the consumed tokens be counted when the inference is interrupted halfway or canceled

[JonSmall]

📋 CheckList

• I have searched existing issues to avoid duplicates
• I am using a recently maintained version of Coze Loop
• I have provided all required information
• I understand this is a bug report and not a feature request
• I have submitted this report in English (otherwise it will not be processed)

🐛 Bug Description

How should the consumed tokens be counted when the inference is interrupted halfway or canceled
The current statistics are based on the last output of usage_age. If the inference fails halfway, how can it be completed.

🔄 Steps to Reproduce

Do it by yourself !

[lsy357]

We use the eino to call the LLM. Most models only return tokens in the last chunk, so if an exception occurs midway, the tokens will not be counted.