raggy - a scraping and querying library

shhhhhh... I'm just here for the code

pip install raggy

see the tutorial for a quick start

What is this RAG thing people keep talking about?

R.A.G. stands for Retrieval Augmented Generation. Enough hot air has been blown about it, so just read this if you're not sure what it is.

Why is it useful?

Large language models are trained on discrete data sets, at points in time. They do not update their knowledge sources inherently. Humans want LLMs to know about their data, right now.

What does raggy do?

Non-technical

• efficiently scrape data from the web into rich documents
• throw these documents into a place where they're grouped by similarity and labelled with metadata
• retrieve documents from this place on command, based on similarity to a query or specific metadata (e.g. dates, authors, etc.)

More technical

• coerce arbitrary data formats into a list of Document objects, possibly enriched with metadata
• upsert these Document objects into a vectorstore like Chroma or Turbopuffer
• expose human / LLM-friendly query interfaces to these vectorstores that allow metadata filtering