Crate cognee_chunking 

Text chunking for Cognee, ported from the Python chunking hierarchy.

Splits text through a word → sentence → paragraph hierarchy into token-bounded chunks. Zero-copy where possible (chunks borrow &str slices via byte-offset tracking).

• text_chunker / cognify_pipeline — the chunking entry points (the latter is a plain code span, not an intra-doc link: it is gated off wasm32, where the link would be unresolved on a --target wasm32 doc build)
• token_counter — the token_counter::TokenCounter trait and its WordCounter / HuggingFaceTokenCounter / TikTokenCounter impls, selected by config (TokenCounterKind::from_env)

Re-exports

pub use chunk_by_row::chunk_by_row;

pub use cognify_pipeline::ExtractTextChunksPipeline;

pub use config::TokenCounterKind;

pub use cut_type::CutType;

pub use error::ChunkingError;

pub use text_chunker::NAMESPACE_OID;

pub use text_chunker::chunk_text;

pub use token_counter::TokenCounter;

pub use token_counter::WordCounter;

Modules

chunk_by_paragraph

Paragraph-level text chunker.

chunk_by_row

Row-based chunking for CSV and DLT data.

chunk_by_sentence

Sentence-level text chunker.

chunk_by_word

Word-level text chunker.

cognify_pipeline

Extract text chunks pipeline.

config

Chunking configuration — tokenizer selection via environment variables.

cut_type

error

text_chunker

Top-level text chunker producing DocumentChunk output.

token_counter