Skip to content

spRAG

A RAG framework for dense unstructured documents

Visit official site

What spRAG is

spRAG is described as a RAG framework built for dense unstructured documents, aimed at complex open ended questions over long texts.

What you can do with it

  • Create a knowledge base directly from a file
  • Query it with a saved knowledge base id
  • Combine related chunks into longer passages

Who it is for

  • Data scientists and machine learning engineers
  • Financial analysts reading financial reports
  • Legal professionals searching legal files

Getting started

  • Install the package with pip
  • Set environment variables for the API providers
  • Create a KnowledgeBase from a file
  • Load it again by its id

Pros & cons

✓ What we like

  • KnowledgeBase objects persist and save configuration
  • Embedding and reranking can use OpenAI, Claude 3 Haiku and Cohere

! What to watch out for

  • API keys for several providers must be configured
  • The card reports accuracy above a traditional RAG baseline on FinanceBench

FAQ

How is a knowledge base created?

The card names create_kb_from_file to build a KnowledgeBase from a file.

Can the model choice change?

The feature list states custom configuration lets users pick different models and APIs.

Last reviewed: 2026-11-15

More AI writing assistant tools

View all →

How we review