Feeding a 1k/10k/100k tokens document always results in a summary of the same length.
Which is a nice feature if you're looking for an executive summary style.
But I sometimes wish I could have simply a way to parse the info faster, with a lower level of details.
I wonder if there would be a way to customize the output, based on the needs.
Maybe we could have a slider where you could configure the output to be:
- executive summary (the current default fixed length)
- 25% of the text size
- 50% of the text size
- 75% of the text size
Or maybe from the default summary, tell what is the current token density, and offer the possibility to change it:
"This summary is 3% of the original document. Click here to get more details/Use the slider to customize the level of details"
Useful information density is imperfectly correlated to the amount of tokens, so maybe there's a better metric to work with.
Not sure it makes sense.^^"
