Chonkie

Chonkie is an open source ingestion library for retrieval augmented generation pipelines. Its core job is chunking: splitting documents into pieces sized for embedding and retrieval. Around that it covers the rest of ingestion, so a single library can fetch, clean, chunk, refine, embed and send data to a vector database. A pipeline interface chains steps together, for example a recursive chunker tuned for Markdown, a semantic chunker, an overlap refinery and an embeddings step. It is for developers building RAG systems who want a lightweight alternative to large frameworks. It follows a rule of minimum installs, so users add only the chunkers and integrations they need. It offers more than 32 integrations with tokenizers, embedding providers and vector databases, supports 56 languages, and is available in Python and JavaScript as well as through a hosted API. Agent skills are also published so coding assistants such as Claude Code, Cursor and Copilot can use the library correctly. The library is free under the MIT licence and installs with pip or uv. A cloud option exists for chunking through an API, but the company's pricing link now redirects to another site and no cloud prices could be confirmed, so buyers should check current terms before relying on the hosted service. The notable change is corporate: the chonkie.ai domain now redirects to Feyn Labs, a company focused on training specialist models on a customer's own data, which suggests the team has shifted its commercial focus while the open source library and its documentation remain available.