← Back to all projects

LlamaIndex

Retrieval-first framework for grounding agents in your documents

OfficialMIT
Stars
52.7k
Forks
8.5k
Open issues
1.1k
Last commit
26 Oct 2026

What is LlamaIndex?

Where most frameworks start from the agent loop, this one starts from the data: ingestion, indexing strategies, and the retrieval patterns that decide whether answers are actually grounded. Reach for it when the hard part of your problem is the corpus rather than the reasoning.

What can you do with LlamaIndex?

  • Query a folder of documents — SimpleDirectoryReader(dir).load_data() feeds VectorStoreIndex.from_documents(), and index.as_query_engine().query() returns an answer grounded in those files.
  • Persist an index to disk — Indexes are in-memory by default; index.storage_context.persist() writes to ./storage and load_index_from_storage(StorageContext.from_defaults(persist_dir="./storage")) reloads it without re-embedding.
  • Run on local models only — Setting Settings.llm to Ollama and Settings.embed_model to HuggingFaceEmbedding swaps OpenAI out for local inference and local embeddings without changing the indexing code.
  • Install only the integrations needed — Instead of the all-in-one llama-index starter package, install llama-index-core and add specific packages such as llama-index-llms-ollama from the 300+ on LlamaHub.
  • Parse hard document formats — LlamaParse, a separate hosted platform requiring an API key, does agentic OCR across 130+ formats and can feed the ingestion step of an index or be used on its own.

Star history

15 Aug to 30 Oct · +1.1k

51.7k52.7k

Frequently asked questions

Is LlamaIndex free for commercial use?

LlamaIndex is released under the MIT licence — OSI-approved open source, which permits commercial use.

How can LlamaIndex be deployed?

LlamaIndex is available as Self-hosted / Managed cloud.