Tutorial: Build AI-Supercharged RAG Apps with a Vector Database
summary
Hands-on build of a RAG app using an open source stack: a Weaviate vector database plus local Ollama for embeddings and a small Gemma LLM, ingesting 100,000 Twitter support rows into an HNSW index with quantization. The demo shows precomputed vectors import hitting about ~2,500 objs/s, searches via vector, keyword and hybrid queries with an alpha blend, and a post-index memory footprint that drops to ~300 MB while retrieval augmented generation returns grounded summaries.
AI-written summary · about
chapters (16)
- 0:00Workshop introduction
- 1:26Weaviate vector database
- 5:17Search types and RAG basics
- 9:50Vectors and meaning
- 13:19Why vector search wins
- 18:10Build vector database
- 22:03Movie data setup
- 31:57BM25 filtering
- 42:13Semantic vector search
- 51:04Vector distances and indexes
- 55:39Performance and scalability
- 1:01:18Streamlit demo setup
- 1:06:54Hybrid search in app
- 1:14:51Memory optimization
- 1:21:12Workshop resources
- 1:21:57Q&A
related talks
13mRAG Without the Hassle: Building AI-Powered Applications
33mVector Databases in the Age of AI: Embracing RAG for Enhanced Search Capabilities
53mBuild RAG from Scratch
20mProduction RAG apps made easier with Astra
47mReal-Time AI with Open Source
06mWhat makes vector databases essential for RAG and AI apps?
47mExtendable by Design: Building Generative AI Apps with Postgres and Vector Storage and Azure AI
35mThe Hidden Jewels to Solve Confabulation Challenges in Generative AI’s RAG Promise