Where did the mystery tokens come from?

General
1

I had an experience like this today. I added a task. The task was for the AI to extract text from the Web page and convert the text into a pdf file. Since I used missing words, the AI went and found the pdf books and tried to extract the text (big data, that is). Afterwards, I wanted to correct the sentence of the mission and try it. I got an error. It said I ran out of AI tokens.

I checked the tokens and saw that I had completed the cost.

Now I logged into thunk.AI and I see that the cost quota, which I saw as 10 dollars, has decreased to 5 dollars. And I don’t get an alert that I’m out of cost.

Where did the mystery tokens come from?

a- is this a bug?
b- Did you install the free tokens?

Below are screenshots of the incident I described.

image
image
image
image
@praveen-Thunk.AI

2

Yes, we put in an approximate usage counter and a limit on free usage.
Those token counts seem wildly wrong though! So we’ll take a closer look at that. It should be roughly 1c for every 100 tokens (and a token is very very roughly one word). For the amount you’ve used it, I’m not surprised you are close to your limit.

As a note, this is not only about the tokens/words produced. Most of it is the tokens/words consumed by the language model. There is a lot of information that has to be sent to it and then cross-checked by it in order for the work to happen. Those are the “prompt tokens”.