To RAG or not to RAG
summary
Explains when RAG (retrieval augmented generation) actually fixes hallucinations by showing a live bash demo that runs a local LLM (Llama), creates embeddings, chunks web pages, and stores them in an Elastic vector store so queries about Java stream “gatherers” return sourced text not made‑up prose. Compares RAG to MCP/agents for read‑only retrieval versus tool-driven actions, shows why embedding model choice and chunking matter, and demonstrates that a working pipeline needs only HTTP calls, embeddings, vector search, and prompt assembly.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 5:06AI hallucination problems
- 8:18What is RAG
- 12:35Why context matters
- 15:47RAG retrieval mechanisms
- 20:09Vector search for semantics
- 25:24Building embeddings and nearest neighbor
- 28:18RAG with Elasticsearch and local models
- 33:00When to use RAG
- 35:27RAG vs MCP differences
- 39:19Combining MCP and RAG
- 40:43AI job impact and staying relevant
- 41:45Resources and community
related talks
39mRAGs to Riches: How Our Content Affects Retrieval Augmented Generation
01h 33mRAG How To Talk To Your Data
21mHow RAG is Redefining AI Capabilities
48mCrafting your own RAG system: Leveraging 30+ LLMs for enhanced performance
50mRetrieval Augmented Generation Beyond Toy Demos
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
22mThe Power of RAG at Scale: A Year of Explosive Growth and Observab...
51mDeveloping a Custom RAG System