A data framework for LLM applications focused on data preparation, indexing, and access for RAG scenarios. It provides abstractions for indexes, query engines, and index graphs, enabling flexible pipelines on top of files, databases, APIs, and vector stores. Key ideas include: a unified representation of heterogeneous sources, separating the data layer from agent logic, extensible retrievers, automatic query rewriting, and multi-hop reasoning via chained indexes. LlamaIndex is a good choice when you need to combine structured and unstructured data, run complex multi-step queries involving retrieval, aggregation, and grounding, and finely control how the LLM “navigates” your data.