Smart Search with RAG: Elasticsearch Meets Language Models
summary
Shows how Elasticsearch becomes a RAG retriever: data chunks (demo uses chunk size 50) go into an index named gs_rag_demo, Elser generates tokens and embeddings, and hybrid BM25 plus vector KNN ranking returns four context records that get stitched into a prompt for OpenAI/ChatGPT (demo uses LangChain-style code). Example answers come back specific: “work-from-office resumes May 1, 2023, three days a week” and sales goal “increase revenue by 20%.” Notebook runs in Python against Elastic Cloud ID and demonstrates quantization and reranking knobs.
AI-written summary · about
chapters (12)
- 0:00Introduction and speaker intro
- 1:25Challenges with generative AI apps
- 3:14What is RAG and how it works
- 4:21Overview of Elasticsearch
- 5:47Elasticsearch in RAG workflow
- 6:29Demo: Q&A on private data
- 7:56Lexical search explained
- 9:21Semantic and vector search
- 11:10Hybrid search and Elser
- 14:00Building RAG with Elasticsearch demo
- 19:05Why Elasticsearch for RAG
- 20:28Resources and closing
related talks
45mMaking semantic search & RAG real: How to build a prod-ready app
37mBuilding advanced RAGs with Elasticsearch OpenSearch
34mImplementing an open-source RAG with OpenSearch
46mBlow past benchmarks using vector search and RAG with Elasticsearch
42mTake Your RAG To The Max!
53mBuild RAG from Scratch
43mHarnessing DeepSeek-R1 and OpenSearch for RAG-Powered Intelligent Info...
20mChallenges in Building RAG Systems (and Solutions Using OpenSearch)