Skip to content
GitHub

Chunkwise

An open-source, self-managed platform to evaluate chunking strategies and deploy ETL pipelines for RAG applications.

Evaluate, compare, and deploy chunking strategies

Chunkwise provides a complete experimentation platform for RAG data pipelines. Test different chunking strategies, benchmark retrieval performance with ground truth datasets, and deploy optimized ETL pipelines to production.

  1. Visualize

    Create workflows, upload documents, and select chunking strategies. Adjust chunking configurations and see how documents are split into chunks.

  2. Evaluate

    Benchmark your chunking strategy’s retrieval performance using a ground truth dataset to optimize for your use case.

  3. Compare

    Create multiple workflows with different chunking strategies and compare them side-by-side. View differences in chunk metrics and retrieval performance.

  4. Deploy

    Deploy your tested workflows as production ETL pipelines. Automatically provision infrastructure to ingest and process documents from your S3 bucket.

Learn how we built Chunkwise

Read our case study to understand the architecture, design decisions, and challenges we faced while building a production-ready platform to evaluate chunking strategies and deploy ETL pipelines for RAG applications.

Read the case study →