rime-data-octagram for Debian
=============================

This package installs the compact-size octagram language models into
/usr/share/rime-data, together with grammar.yaml:

  zh-hant-t-essay-bgw.gram   traditional Chinese, word bigrams
  zh-hant-t-essay-bgc.gram   traditional Chinese, character bigrams
  zh-hans-t-essay-bgw.gram   simplified Chinese, word bigrams
  zh-hans-t-essay-bgc.gram   simplified Chinese, character bigrams

They are only used by librime-plugin-octagram, and only by schemas that
include a section of grammar.yaml. Some schemas already do so optionally,
for example luna_pinyin.schema.yaml (rime-data-luna-pinyin) contains

  __include:
    - grammar:/hant?

so they use the traditional Chinese word model as soon as this package is
installed. For other schemas, create <schema_id>.custom.yaml in the Rime
user data directory (for example ~/.local/share/fcitx5/rime/ or
~/.config/ibus/rime/):

  patch:
    __include: grammar:/hans

Use grammar:/hant for dictionaries based on traditional characters, and
grammar:/hant_char or grammar:/hans_char for character-based schemas.
Then redeploy Rime from your input method frontend.

The standard-size models (about 30 MB for each character model and 200 MB
for each word model) can be downloaded from the upstream release page,
https://github.com/lotem/rime-octagram-data/releases . They use the same
file names, and a copy in the Rime user data directory takes precedence
over the one in /usr/share/rime-data.

The .gram files are rebuilt at package build time (see debian/rules), since
the upstream files are memory images in little-endian byte order.

 -- Boyuan Yang <byang@debian.org>, Wed, 30 Sep 2026