chonkie

What it is

Chonkie is a Python library designed for fast, efficient chunking of text to support retrieval-augmented generation (RAG) pipelines. It provides multiple chunking strategies, a modular set of integrations, and an API server option for serving pipelines.

How it works

The project exposes several chunkers (e.g., RecursiveChunker, SemanticChunker, FastChunker, etc.) that split text according to different strategies. It offers refineries (e.g., OverlapRefinery, EmbeddingsRefinery) to post-process chunks, porters (JSON, Datasets) to export results, and handshakes to ingest chunks into vector stores (e.g., Chroma, Elastic, MongoDB, Milvus, Pinecone, Qdrant, Weaviate, LanceDB). It also supports multiple tokenizers (character, word, byte, tokenizers, tiktoken, transformers).

Usage examples show how to initialize a chunker, process text, and iterate over chunks, as well as a pipeline example chaining chunking, refinement, and embedding steps. The project also provides a REST API server option with installation commands for API and dependencies.

Getting started

Installation basics shown:

pip install chonkie

Or for full installation via extras:

pip install "chonkie[all]"

And API server usage:

pip install "chonkie[api,semantic,code,catsu]"
chonkie serve

Or run with uvicorn:

uvicorn chonkie.api.main:app --host 0.0.0.0 --port 8000

Docker option:

docker compose up

Recent releases

• v1.7.0 v1.7.0 (2026-07-07): # Chonkie v1.7.0
• v1.6.8 v1.6.8 (2026-06-01)
• v1.6.8-alpha.1 (2026-05-29)
• v1.6.7 (2026-05-19)
• v1.6.6 (2026-05-13)

Traction

4637 stars, 349 forks, 55 open issues

Behind the repo

Chonkie is developed by the chonkie-inc organization and released under the MIT license. The repository contains documentation and examples demonstrating integration with 45+ components across tokenizers, embeddings, LLMs, refineries, porters, and vector databases.

Caveats

License: MIT Created: 2025-03-29 Last push: 2026-08-04