LlamaIndex

LlamaIndex is a framework focused on connecting LLMs with private data, the mainstream choice for RAG applications. More "focused" than LangChain with cleaner API design.

Core abstractions

• Document: raw data from any source (PDF, webpage, DB row).
• Node: minimal unit after chunking, with metadata.
• Index: organization of nodes (vector index / keyword index / knowledge graph index / tree index).
• QueryEngine: executes a query against an index + LLM.
• ResponseSynthesizer: composes retrieved nodes into the final answer.

Workflow

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader

documents = SimpleDirectoryReader("data").load_data()
index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine()
response = query_engine.query("What does the doc say?")

Advanced features

• Query Pipelines: chain retrieval + rerank + LLM into debuggable pipelines.
• Agents / Tools: built-in ReAct, OpenAI Function Calling agents.
• Multi-modal: native support for images and PDF tables.
• Workflows: event-driven orchestration (v0.10+).

vs LangChain

Newcomers get a better RAG experience starting with LlamaIndex; complex agents go with LangChain / LangGraph.