Open joshuasorkin opened 7 months ago
so maybe just do this for languages where I've already built up a sizable amount of word_image entries: english, russian, spanish.
What to do for languages that don't have a large vocabulary built up? Would it be possible to make a comma separated string of words and say "don't use these"?
Yes, it seems like that is possible
let's say we have the words that have been most recently shown at these times: firecracker - 13:30 aluminum - 6:45 smithy - 3:23 merriment - 18:21
we go through our max number of attempts to choose a word either through random selection from word_image, or generating via openai. all end up with words that have been shown today. From this list, then, we would choose "smithy", since it was shown the longest time ago.
this will be a problem if it's a language that has very few words, because then the same few words will end up getting cycled through earliest most-recent