Chonkie

The company shares updates about what it is building, learning, and shipping.

Chonkie provides native chunking and embeddings for LlamaIndex pipelines.

Introducing Pulpie, a family of Pareto-optimal models for extracting main content from HTML pages.

Introducing SQRL, text-to-SQL models trained to inspect a database before writing SQL.

The company supports training custom models on user data, continued learning in production, faster and cheaper inference, and ownership of model weights.

The policy describes how Chonkie, Inc. collects, uses, and protects information.

The terms govern use of Feyn's website and public materials.

GitHubRepository owner

Chonkie introduces native chunking and embeddings support for LlamaIndex retrieval pipelines. The integration is powered by Chonkie libraries and brings those capabilities to LlamaIndex pipelines.

Feb 2026chonkie.aiNews article

Chonkie’s engineering team builds Memchunk, a high-throughput text chunker for retrieval pipelines. It uses single-instruction-multiple-data (SIMD) delimiter search to process text at very high speeds.

Jan 2026chonkie.aiNews article

Shreyash Nigam and Bhavnick Minhas are building Chonkie, an open-source library that splits large documents into semantically meaningful chunks for retrieval-augmented generation. The founders trace the project to context-window limits they encountered in industry, and Chonkie has progressed into a Y Combinator company.

Nov 2025every.ioNews article