Skip to content
← Tools / Skills
ToolLlamaIndex Inc./10 trustVerified

LlamaIndex

Data framework for connecting LLMs with your data — ingestion, indexing, and retrieval for RAG and agents.

Visit official site →

About

LlamaIndex is an MIT-licensed Python/TypeScript framework purpose-built for RAG: 300+ data connectors, vector indexing, hybrid retrieval, reranking, and agent workflows over private data.

Important context: the company's public focus has shifted to its commercial LlamaParse agentic-OCR platform — the OSS framework remains maintained (Oct 2026) but is no longer the company's primary bet.

What it can do

  • 300+ data connectors (LlamaHub) for PDFs, APIs, SQL, docs, and more
  • Vector, keyword, hybrid, and hierarchical retrieval with reranking
  • Structured data extraction and multimodal RAG
  • Agent workflows and tool-use over indexed data
  • 300+ integration packages for any LLM, embedding, or vector-store provider
  • Commercial sibling LlamaParse for agentic OCR of complex documents (free tier available)

Setup

  1. pip install llama-index (starter) — or pip install llama-index-core plus only the integration packages you need.
  2. Set OPENAI_API_KEY (or configure Ollama/local LLM + HuggingFace embeddings for fully local use).
  3. Load docs: SimpleDirectoryReader("data").load_data()
  4. Build index: VectorStoreIndex.from_documents(documents)
  5. Query: index.as_query_engine().query("..."); persist with index.storage_context.persist().

Prerequisites

Python; LLM + embedding provider API keys (or local models via Ollama); storage backend optional.

Use cases

datasearchchatcode

Works with

Generic APIChatGPT

Trust score — 9/10

/10 trustVerified

52k stars, MIT, actively maintained with quality docs and the largest RAG-focused integration ecosystem. Docked one point because the vendor has publicly pivoted its primary focus to the commercial LlamaParse platform — the OSS framework is still updated, but its long-term trajectory is now enterprise-parse-centric. Vendor benchmark claims on the homepage are marketing — provider-claimed.

★ 52,469 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
8.9/10
Maintenance · recency, releases, issue load
9.3/10
Openness · license + docs
10.0/10
Manual review · sanity read + red-flag check
8.0/10
  • · 52,469 GitHub stars (log-scaled vs catalog peers)
  • · blended with ~2,937,852 monthly package downloads
  • · last push 4d ago; 19 releases/12mo; 908 open issues
  • · manual review: pivot to LlamaParse monetization; core stays MIT

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.