RAG Workflows with Weaviate
End-to-End RAG with Weaviate
JP Hwang
Senior Developer Educator, Weaviate
Looking back...
Coming up!
From strings to PDFs
PDFs
Headings
Tables
Images → Chapter 3
docling
→ convert PDFs to markdown
Goal
: Be able to ask documents questions
1
Stanford AI Index Report 2025
Back to the "R" in RAG
Searching with vectors
Create text embeddings
Compare cosine distances
Retrieve the most similar document(s)
Robust to synonyms, typos, alternative wordings
Good choice for "messy" inputs
Searching with keywords
Look at frequency of term usage in documents
Retrieve the document(s) with the most usage
Simpler and faster
Less flexible to "messy" inputs
The best of both words: hybrid search
Performs both searches simultaneously
Weights both rankings by value $\alpha$
Let's practice!
End-to-End RAG with Weaviate
Preparing Video For Download...