Problem

LLM usage is not invoiced by characters, but by tokens. Therefore those need to be counted, regardless if usage is triggered by translation, quality estimation or other usages.

Solution

Count the number of tokens for every AI request.

Extend the currently existing resource usage log:

Add a sheet "Resource usage per month and type".

This sheet duplicates the sheet "Resource usage per month" (also with the additional columns mentioned below), but adds another column "type".

This column "type" can have the following values:

• MT (char-based)
• AI - translation
• AI - quality estimation
• AI - alternative translations
• ... (further values in the future)

The sheet "Resource usage per month" summons all different usage types of an AI, while the sheet "Resource usage per month and type" shows separate rows for each type.

Add the column type also to the sheet "Resource usage log".

Add a column "Tokens" to the  sheets "Resource usage per month", "Resource usage per month and type" and "Resource usage log".

In case of an AI resource the column "Characters translated" stays empty and only a value is shown in "Tokens" - and vice versa.