Settings → History shows what you have spent on artificial intelligence since JUDLEX started keeping track. Each call to the provider is recorded on your computer with what it cost.
JUDLEX does not tell you what something will cost before you click. Buttons say that they use artificial intelligence and how much will be sent (documents, pages); the amount appears afterwards, here, as money spent.
At the top
A line reminds you that the figures are worked out with each provider's public rates and that the provider's invoice is the one that counts. Next to it, How the AI is billed opens again the notice you accepted when you added your key, with links to each provider's pricing page.
Below that, four figures:
Last day with spending (or week, or month), with how much it went up or down compared with the previous one.
Calls in the period: how many times artificial intelligence was called in that period.
Total recorded: everything spent.
Tokens used: the units of text sent and received, which is what providers charge for.
Amounts are in dollars, because that is how providers charge.
Calls without an amount
If you use your own provider with a model that has no rate, a note appears below the figures: And 12 calls without an amount, from your own providers without a rate. Those calls are counted, but they are not added to any total, because JUDLEX does not know what they cost; nor are they recorded as free. In the tables below, the model's row reminds you. To give them an amount, give the model a rate from Settings → Connections: The price.
Spending over time
By day, By week or By month. The table lists only the periods with spending; the chart covers the whole calendar and leaves blank the ones with no activity. Alongside, Highest day and Total shown.
Where it went
Three tables, each with the calls, the units of text, the amount and the percentage:
Where it goes: by operation. Classifying and analysing exhibits is usually where the money goes; asking the case file questions hardly ever is. The chat's running commentary while papers come in has its own line.
With which model: to see whether the expensive model is worth it. Models you have not used are marked unused.
Which case file it's charged to: useful if you pass the cost on to a client. Anything spent outside a case file, such as a general question, has its own line.
Cap per question
Cap per question is the most one question to the chat can spend, counting what the helpers it shares the work with spend too ($0.25 by default). It is not what it will cost: it is a limit on what has already been spent. When it is reached, the chat makes no more calls and answers with what it has, saying what was left unchecked.
What these figures mean
Each amount is worked out on the day of the call, at the provider's public rate at that time, with no mark-up. If the provider changes its prices, what has already been recorded does not change.
Calls to your own provider use the rate you gave it or the one the service publishes. Without a rate, they have no amount; with a server marked as not charging, they are zero.
It is a guide. The invoice that counts is your provider's, since the provider is the one who charges you. The JUDLEX licence does not include artificial intelligence usage.
What you have spent today also shows at the foot of the consultation panel: Panel options.