leonardo scrape is finished except a few days that were missed for some reason, tasks are per hour (their pagination uses timestamps and an offset, using day ranges slowed down as offset increased)
>>> TASKS.count_documents(query_eq("complete", True))
18110
>>> TASKS.count_documents(query_not_exists("complete"))
116
>>> IMAGES.estimated_document_count()
953839291
checked distinct prompts last night, will be slightly off from the final count
>>> count_distinct(IMAGES, "generation.prompt")
228941006
shame they don't have 1B by themselves