The LlamaIndex framework is a Python toolkit for building LLM-powered agents over your data. LLMs are trained on public data, not yours. Your data is behind APIs, in databases, or trapped in PDFs and slide decks. Context augmentation makes that data available to the model at the moment it answers, and its best-known form is retrieval-augmented generation (RAG). The framework gives you the parts to build any context-augmentation use case, from prototype to production, and it is the shortest path from documents parsed by LlamaParse to an agent that can use them.

• Agents are LLM-powered assistants that use tools to complete tasks, from answering questions over your documents to taking actions in other systems. A RAG pipeline is one of many tools an agent can call.
• Workflows are event-driven, multi-step processes that combine agents, data connectors and tools, with branching, retries and human-in-the-loop review, and can be deployed as services.

The framework is built from composable parts. The high-level API gets you from documents to answers in a few lines; the lower-level API lets you customize or replace any of them.

New here? The starter tutorial builds an agent with a tool over your documents, the local-models tutorial does the same without any hosted API, and how to read these docs points you to the right section for your experience level. The Learn section then walks through a complete application step by step.

The framework’s built-in readers are fine for clean text. For scanned PDFs, forms, spreadsheets and slide decks, the quality of what goes into your index decides the quality of every answer. That is what LlamaParse is for: Parse (agentic OCR for scans, tables and charts), Extract (JSON in your schema), Classify, Split and Index (managed ingestion and retrieval). Parsed pages drop straight into a framework index:

pip install llama-index "llama-cloud>=2.8"
export LLAMA_CLOUD_API_KEY=llx-...
export OPENAI_API_KEY=sk-...

The OpenAI key is for the index, which embeds the pages with OpenAI by default. To embed locally instead, set Settings.embed_model as in the local-models tutorial.

from llama_cloud import LlamaCloud
from llama_index.core import Document, VectorStoreIndex

client = LlamaCloud()  # reads LLAMA_CLOUD_API_KEY

file = client.files.create(file="data/report.pdf", purpose="parse")
result = client.parsing.parse(
    file_id=file.id, tier="agentic", version="latest", expand=["markdown"]
)
pages = result.markdown.pages
documents = [Document(text=p.markdown) for p in pages if p.success]

index = VectorStoreIndex.from_documents(documents)

From there the index is a tool like any other: the starter tutorial shows how to hand it to an agent.

For fully managed retrieval, Index keeps an index in sync with your data sources and serves retrieval to your query engine or agent. For parsing on your own machine, LiteParse is the open-source, local option.

More on the use cases page: text-to-SQL, querying CSVs, prompting techniques and fine-tuning.

The framework is the oldest of our open-source projects, not the only one. The newer ones are smaller, focused on documents, and work with the framework or on their own.

LlamaIndex Framework is MIT-licensed and built with its community. The contributing guide covers everything from a documentation fix to a new integration. The FAQ answers the questions everyone asks first.

More from LlamaIndex: create-llama, full-stack projects and the Discover LlamaIndex video series.