Building an Advanced Knowledge Assistant 📚🤖 RAG was just the beginning. A huge theme we’ve noticed this year is the prevalence of companies building an advanced knowledge interface over unstructured data, to produce more unstructured outputs. Check out my talk at the AI

Any “agentic” application is basically a computer program with LLM calls in the mix. We think an agent orchestration framework should support any custom agentic use case, whether it’s some simple LLM routing calls or an entire ReAct agent. LlamaIndex 🦙 workflows have the

Introducing LlamaParse Premium 💎 LlamaParse is the best document parser for your LLM applications, and now it’s even better. LlamaParse Premium combines the visual understanding capabilities of SOTA multimodal models with long text/table content extraction capabilities of

Reading Long Tables Even Sonnet-3.5 hallucinates some values on the Caltrain schedule, a giant timetable of train times. But LlamaParse Premium gets it correct vs. raw multimodal models. See below (source doc, raw gpt-4o parsing, raw sonnet-3.5 parsing, llamaparse premium):

Additional Benefits - Parse diagrams into Mermaid, a text format you can directly pass to LLMs or render on its own (see below - source, diagram) - Parse equations into LateX A huge shoutout to Pierre-Loic Doulcet and Sacha Bron for the work on this. You can try it today here:

So excited for our first Agentic RAG-A-THON with LlamaIndex 🦙, Pinecone, VESSL AI, @box, Together AI, Mistral AI, OpenAI and more! Join us in Silicon Valley for an exciting hackathon focused on building RAG and advanced AI agents 🚀 $10K in cash prizes for the winners!

LlamaParse premium is the best document parser out there for your context-augmented LLM application. It can handle complex slide decks, diagrams, multi-table Excel sheets, interleaving scanned document text, and any other document type with lots of text, tables, and visual

An underrated issue with document parsing for RAG / agent use cases is dealing with multi-page tables - sometimes a big table spills over into multiple pages. This breaks chunking algorithms that generally operate at the page-level or smaller, and causes LLMs to lose the full

🚀 Full house!! We're so excited to see everyone on Sunday for the first-ever LLM x Law Hackathon! 🎉 ⚖️ 🧑‍💻 Get ready to tackle real-world legal challenges using LLMs! See you there Megan Ma, Pierre-Loic Doulcet, Atli Cervantes, Jonathan Richard Schwarz, International Chamber of Commerce, Jus Mundi, Pierce Kelaita,

How to parse only selected pages with LlamaParse When you have a long and complex document, but you only need a few specific pages, you can save time and credits by specifying which pages to select. This quick video from Ravi Theja walks you through it:

It's been a huge year for LlamaIndex! Tens of millions of pages parsed, millions of framework downloads, and tons of happy users! Check out our year-end review in the numbers, including a month-by-month breakdown of our biggest feature releases: llamaindex.ai/blog/the-year-…

Excited to be featured as an NVIDIA partner during Jensen's keynote at CES 🔥

I know everyone is talking about DeepSeek R1 and the "DeepSeek selloff" today... Taking a break from that, and getting back to business for a minute - we just released vectara-ingest with support for llama_parse from LlamaIndex 🦙.

Learn everything there is to know about LlamaParse in this comprehensive video! In this video, Sudarshan Koirala covers: ➡️ Multiple parsing modes ➡️ Using parsing instructions to improve quality ➡️ Output formats available ➡️ Parsing audio and images ➡️ JSON mode ➡️ Using it all in

Our co-founder Jerry Liu recently gave a guest lecture on building document workflow agents, and you can catch the recording! It covers: ➡️ LlamaIndex's evolution from pure RAG to knowledge agents with multi-step reasoning capabilities over enterprise data. ➡️ Advanced document

It's 2025, it's finally the year of linux on the desktop and RAG is dead!

What i also like about this paper is it implies new architectures for both retrieval/RAG and conversation history - by storing the data representation natively as image tiles, you can render "old" text to progressively lower-resolution images. This means you might be able to