LlamaIndex

LlamaIndex is a comprehensive toolkit designed to build context-augmented agents. It bridges the gap between Large Language Models and private data through robust retrieval and indexing components.

Key Components

• QueryEngine: A component that retrieves relevant information (RAG) and provides a synthesized answer.
• VectorStoreIndex: A searchable data structure for embeddings.
• LlamaHub: A vast registry of community-contributed loaders, tools, and agent templates.

Agency in LlamaIndex

LlamaIndex supports multiple agent patterns:

• Function Calling Agents: For models with native tool-calling APIs.
• ReAct Agents: For general-purpose reasoning over any LLM.
• Agentic RAG: Using a QueryEngine as a tool, allowing the agent to decide when and how to search the data.

State & Workflows

• Context: A state-management object that allows agents to remember past interactions.
• Workflows: An event-driven, async-first way to define agentic behavior as a sequence of discrete steps and events.

See Also

• agentic-rag
• graph-orchestration
• agentic-frameworks-moc