LlamaIndex LLM Data Framework

LlamaIndex is a data framework for building LLM (Large Language Model) applications, providing tools for data ingestion, structuring, and retrieval to connect custom data sources to LLMs. Created by Jerry Liu in 2022, it enables developers to build Retrieval-Augmented Generation (RAG) applications, agents, and data-driven AI workflows. Key features: data connectors (LlamaHub) for loading data from over 160 sources including PDFs, databases, APIs, Notion, Google Drive, Slack, and web pages. Data indexing for transforming raw documents into queryable structures including vector indices (embeddings), keyword indices (BM25), tree indices (hierarchical), list indices (sequential), and knowledge graph indices. Query engines for answering questions over indexed data using retrieval-augmented generation with similarity search, hybrid search, and custom retrievers. Chat engines for multi-turn conversational interfaces with memory and context management. Agents powered by LLMs that can use tools, make decisions, and execute multi-step reasoning with ReAct and function-calling patterns. Response synthesizers for combining retrieved information into coherent answers using refine, compact, and tree_summarize modes. Evaluation framework for assessing retrieval quality, response faithfulness, and relevance using LLM-based evaluators. Over 100 LLM integrations including OpenAI, Anthropic, local models, Azure, and Cohere. Vector store integrations with Pinecone, Chroma, Weaviate, Milvus, Qdrant, and pgvector. Customizability via callbacks, event systems, and pluggable components. Python and TypeScript SDKs. Used by enterprises for enterprise search, knowledge management, and AI assistants.