These micro-apps are intentionally not "chatbots" and they aren't "custom GPTs". Those things are great, but too open and unstructured for assessment and organizational use (where costs need to be tightly controlled).
That being said, I have seen several use cases now where the AI gives an output, but the user has to revise that output. I.e. the AI provides a video script but the user says, "cut that down to 500 words" or something.
So I'd like to build in an option to "revise" the output, and a "max revisions" setting.
Here is a user story.
If the "revisable" option is True, then the user get a "revise" button option after the outputs have been generated by AI.
If the student clicks revise, they get a text area that allows them to ask for a change in the outputs.
When submitted, that text gets sent to AI.
The AI sends a new output.
This process is repeatable for as many times as the user wants up to the "max revisions" field. Once revisions are used up, the user is forced to accept the output.
I think the way we calculate costs for this template is just by looking at the tokens used for the one prompt, since that was the max for this template in the past. Now, we might have multiple prompts, so you'll have to adjust the cost numbers to "add on" to each other if a user runs multiple revisions.
These micro-apps are intentionally not "chatbots" and they aren't "custom GPTs". Those things are great, but too open and unstructured for assessment and organizational use (where costs need to be tightly controlled).
That being said, I have seen several use cases now where the AI gives an output, but the user has to revise that output. I.e. the AI provides a video script but the user says, "cut that down to 500 words" or something.
So I'd like to build in an option to "revise" the output, and a "max revisions" setting.
Here is a user story.
I think the way we calculate costs for this template is just by looking at the tokens used for the one prompt, since that was the max for this template in the past. Now, we might have multiple prompts, so you'll have to adjust the cost numbers to "add on" to each other if a user runs multiple revisions.