DICTIONARIES · LICENSES

Dictionary licenses and imports

Licenses for bundled, optional and user-imported dictionary data

In one line: the project code is MIT, but dictionary data is not MIT. The main Simplified dictionary is distributed as GPL-3.0 because it includes GPL-3.0-only data. Traditional, stroke and Wubi 86 are LGPL-3.0; Wubi 98 is Unlicense; Kangxi is MIT; the rare-character index uses the Unicode License. Optional imports CustomPinyinDictionary (CC BY-SA 4.0) and rime-wanxiang (CC BY 4.0) require attribution, a license link and a change notice when redistributed.

1. Dictionaries shipped with the installer

DictionaryContentsLicense and obligations
Simplified main
luna.opid
Characters + common phrases + merged external dataGPL-3.0. Character layer from rime-luna-pinyin (LGPL-3.0); phrase layer includes rime-pinyin-simp (Apache-2.0); GPL-3.0-only data is also merged, so the product is distributed as GPL-3.0.
Traditional
trad.opid
Traditional characters, phrases and a modern conversion layerLGPL-3.0. rime-terra-pinyin (LGPL-3.0) + rime-pinyin-simp / OpenCC (Apache-2.0) + Unicode License + project-authored entries (MIT).
Stroke table
stroke.opid
Five-stroke input and stroke lookup for rare charactersLGPL-3.0. From rime-stroke, pinned commit and compiled as-is.
Wubi 86
wubi86.opid
Wubi 86 code tableLGPL-3.0. From rime-wubi, pinned commit and converted.
Wubi 98
wubi98.opid
Wubi 98 characters + phrasesUnlicense. From 98wubi-tables, pinned commit and converted.
Kangxi
kangxi.tsv
Readings for all Kangxi headwordsMIT. Keep Copyright (c) Shunshi.AI. Frequencies are project ranking constants, not upstream data.
Rare-character index
rare_index.tsv
Radical / stroke-count indexUnicode License. From Unicode Unihan kRSUnicode and kTotalStrokes.

The emoji panel's common subset is adapted from OpenMoji and distributed under CC BY-SA 4.0 (credit the OpenMoji project and contributors). Kaomoji come from rime_kaomoji_dict (MIT, Copyright (c) 2023 aoguai).

2. Optional imported dictionaries (not bundled)

These dictionaries are downloaded, converted and imported by the user. The project repository does not contain them and does not download them at build time. Truncated 200k-entry files are available for direct import:

DataUpstreamLicense and obligations
CustomPinyinDictionarywuhgit/CustomPinyinDictionaryCC BY-SA 4.0. Credit dictionary data by wuhgit and contributors, link the license and state changes. Adapted data redistributed must remain under CC BY-SA 4.0 (or a compatible license), with no extra restrictions.
rime-wanxiangamzxyz/rime-wanxiangCC BY 4.0. Credit rime-wanxiang data by amzxyz and contributors, link the license and state changes. There is no ShareAlike requirement.

The conversion scripts are MIT; the converted data keeps its upstream license. The header comment in each output file is the attribution and change notice — do not remove it. CustomPinyinDictionary has no usable frequencies, so imported entries use the default frequency for ranking.

3. Import limits per client

Imported entries participate in candidate ranking: duplicate words take the larger of the main-dictionary frequency and the import boost, while unique words receive the boost. Large dictionaries can change candidate order; disable or delete them at any time in Settings.

4. Redistribution notes

This page explains licenses only and does not change any upstream license terms. © 2026 erik.xyz · Home · Privacy · Visual asset license