Skip to main content

Crate kiri_engine

Crate kiri_engine 

Source

Re-exports§

pub use crate::shared::DictData;

Modules§

dictionary
lattice
oov
shared
SharedDictionary: pre-parsed trie + raw data that can be shared across multiple tokenizer instances via Arc, avoiding ~150 MB per-tokenizer trie duplication.
tokenizer
buildLattice main loop: for each byte position, checks canBow, hasPreviousNode, performs lexicon lookup, MeCab OOV, simple OOV, and inserts into lattice.
types
Pure Rust types for Kiri’s dictionary structures (no NAPI dependency).

Functions§

inhibit_connection_in_data
Inhibit a connection by writing INHIBITED_CONNECTION (0x7fff) into the connection matrix region of the raw dictionary data.