A deep dive into Retrieval-Augmented Generation with Llamaindex
In this talk we cover:
What LlamaIndex is
What LlamaHub and create-llama are
The stages of Retrieval-Augmented Generation (RAG)
LlamaIndex's ingestion pipeline with caching
The set of vector stores, LLMs and embedding models available in LlamaIndex
Inspecting and customizing your prompts
And then seven advanced querying strategies, including
SubQuestionQueryEngine for complex questions
Small-to-big retrieval for improved precision
Metadata filtering, also for improved precision
Hybrid search including traditional search engine techniques
Recursive Retrieval for complex documents
Text to SQL
Multi-document agents that can combine all of these techniques