Data engineers building RAG and analytics pipelines over complex unstructured documents like PDFs and presentations.

Sycamore is an open source document processing engine from Aryn that segments, OCRs, and extracts tables and images from documents for ETL, RAG, and analytics.

Used as a Python library, optionally paired with the Aryn DocParse GPU powered API, loading results into vector databases like OpenSearch or Pinecone.

Released by Aryn around 2023 and actively maintained as of 2026.

Handles messy real world documents with embedded tables and figures more reliably than simple text extraction, improving downstream RAG quality.

Can run fully locally or depend on the hosted Aryn DocParse API for GPU accelerated segmentation and OCR.

Open source under Apache-2.0 for the core library; the optional Aryn DocParse component is a hosted API that requires credentials and sends document data to Aryn's servers.