# Dependencies for reproducing this benchmark.
#
# The extraction code in domstrip_extract/ pulls in html2text, lxml and
# readability-lxml (which itself depends on lxml). httpx is used only to
# re-fetch pages with --refresh, and tiktoken only for real BPE token counts;
# without tiktoken pass --tokenizer whitespace, which is reproducible but not
# what a model bills you for.
httpx>=0.28
html2text>=2024.2.26
lxml>=5.2
readability-lxml>=0.8.1
tiktoken>=0.8
