疑问，lac.add_word如何使用

试了下，比如添加一个成语“红红火火”，
lac.add_word('红红火火', sep=None)
lac.run("他这一生红红火火了一把")
追后的结果还是把红红火火拆成了 '红红'和'火火'，
[['他', '这一生', '红红', '火火', '了', '一把'], ['r', 'r', 'a', 'a', 'u', 'm']]
请教下各位大神，是我操作不对么

Start by reproducing the reported Python snippet with lac.add_word('红红火火', sep=None) and lac.run(...). Read the add_word API behavior and relevant segmentation entry points to determine why the custom word is split; done means the example keeps “红红火火” as one token, with the expected output verified.

Written by the indexing model from the issue text.