A Deep Dive Into Retrieval-Augmented Generation with LlamaIndex
summary
LlamaIndex is a Python and TypeScript framework that loads any data source, builds embeddings, and connects to LLMs; it supports 25+ models and 30+ vector databases, plus hundreds of connectors (Google Drive, S3, Slack, SQL). Features include a one‑click app generator via create-llama CLI, pipeline caching, hybrid search with tunable alpha, metadata filtering, subquestion query engine, small‑to‑big retrieval, recursive retrievers for tables, a pandas/SQL natural language query engine, and an SEC‑insights agent demo that composes tools across years of documents.
AI-written summary · about
chapters (15)
- 0:00RAG and LlamaIndex intro
- 2:51What is LlamaIndex
- 4:37RAG application stages
- 6:23Data ingestion pipeline
- 7:28Indexing and vector storage
- 8:53Querying and prompting
- 9:57Basic query engines
- 11:19Sub-question engines
- 12:43Retrieval precision techniques
- 14:07Metadata filtering
- 15:30Hybrid search methods
- 17:18Querying tables
- 19:25Querying SQL databases
- 22:13Building multi-doc agents
- 24:17Summary and future
related talks
42mUsing LlamaIndex for building Generative AI apps
31mRetrieval Augmented Generation
29mUsing LangChain and LlamaIndex to customize LLMs...
35mAdvanced RAG Techniques with LlamaIndex
01h 14mAn Introduction to Retrieval Augmented Generation
27mAI Demystified Unveiling the Intelligence Behind the Machine
31mSponsored Session: Building a Production Open Source RAG Application using LlamaIndex and Cassandra
28mA Retrieval Augmented Generation system to query the scikit-learn documentation