Не смотря на простой код многопоточность, асинхронного, мультипроцессорность не делают скачивание быстрым. Не позволяют использовать всю пропускную способность сети. Количество файлов за единицу времени скачивается меньше, чем средняя скорость трафика сети клиента.

Должен быть способ получать файлы гораздо быстрее и больше, и не получить бан по ip за api abuse.

In some cases, when artist tag have Upper case letters.
post_ids = list(struct.unpack(f'!{total_ids}I', bytearray(response.content)))
breaks down to error

File "d:\ghd\python-nozomi-3.0.0\nozomi\api.py", line 174, in _get_post_ids
struct.error: unpack requires a buffer of 152 bytes

--buffer size problem
positive_tags = ['artist:PossumMachine⚠️']
positive_tags2 = ['artist:Libidoll']
positive_tags3 = ['artist:AliceBunnie']
positive_tags4 = ['artist:SPICYdias']

The list of extra tags ['1', '2', '3'] does urls with "...tags:1+2+3&pid=0"
It should be a lot of urls with one single extra tag.

Probably it can be done even faster than the implementation that I have done at the moment.

• negative tags doesn't work correctly.

• good tags may have problems, when tag character and tag artist have same name. negative tag don't except artist tag.
  for example:
  sherry
  

• Bad search: nozomi has tag search with blue tag, red tag as artist tag and green as character tag, but in some cases image searching no have difference in them. It can download extra post, that you don't need. Some tags of character tag are blue and green with low difference.
  
  and
  

• Also need something for good searching to avoid extra tags

positive_tags17 = ['marie']
positive_tags18 = ['marie_rose']
positive_tags19 = ['marierose']

or

positive_tags7 = ['Vanellope']
positive_tags11 = ['vanellope']
positive_tags4 = ['vanellope_von_schweetz']

or

positive_tags10 = ['Sarah']
positive_tags12 = ['sarah_(the_last_of_us)']

sometimes request file_type not equal content_type.