LlamaIndex

LlamaExtract is now in public beta! Extracting structured data from long, complex documents is a common problem our customers face. We've solved it with LlamaExtract, a web UI and API that allow you to define a schema and automatically extract structured data from your documents. In this demo post, learn how to transform complex 10-K reports into structured, actionable data. Read the full post here: https://lnkd.in/dUa2YH7H Or sign up for LlamaExtract, part of LlamaCloud, today! https://lnkd.in/gd2Q8U4P

Gemini 2.0 Flash ⚡️ is surprisingly cracked at document parsing compared to its heavierweight siblings e.g. Sonnet 3.7. The key trick: adding agentic reasoning capabilities on top 🤖 We built a Gemini agent that outperforms most other models in document parsing and does super well on table reconstruction. Now available on LlamaCloud - check it out! https://lnkd.in/g9Wpqn7w To integrate with your code, check out our notebook here: https://lnkd.in/gtWatJhk Huge shoutout to Pierre-Loic Doulcet and Sacha Bron for working on this. 

ICYMI 💫 - here’s a fantastic step-by-step tutorial by Akshay Pachaar on how to build an agentic reasoning system for search and retrieval (specifically, corrective RAG) from scratch. Uses LlamaIndex workflows to let you orchestrate complex, customizable event-driven agent workflows. Combine this with Linkup, Comet, and Qdrant for the full e2e system. Workflow: https://lnkd.in/gvB764h2 Post: https://lnkd.in/gdpTHs7u

Wondering whether to use our Google Gemini or our Google Vertex AI integration? Wonder no more! Now there's a unified Google GenAI integration that supports both! It supports everything you need -- streaming, async, multi-modal, and structured prediction. It even supports image generation and editing! Check out the LLM handling: https://lnkd.in/g36dx_fT And also the corresponding embedding model: https://lnkd.in/gm4QBmgD

The 5th LLM x Law Hackathon is at Stanford on April 6! Bringing innovators together to develop AI solutions for legal work. With First-Build and VC Competition tracks, participants can create new tech or showcase ongoing projects to VCs. Hosted by Stanford CodeX with judges from General Catalyst, Thomson Reuters, Meta, and leading law firms. Sponsors include Jus Mundi, Koyeb , LlamaIndex, and Thomson Reuters. Sign up to participate here! https://lnkd.in/gSKCZfXd

Learn how AI is transforming the future of web development at WeAreDevelopers .org WebDev & AI Day! Join industry experts like Laurie Voss, Hannah Foxwell, and Thomas Steiner for insights on: • AI's role in platform engineering and DevEx • The future of developer tools in an AI-driven world • Practical AI applications in web development workflows Don't miss this virtual event on March 27. Learn more about the event: https://lnkd.in/dX3nZQH6 Get a 15% discount on us! https://lnkd.in/gGBpV-Yj

Connect your LlamaIndex agents to any Model Context Protocol server! We're closely following progress on the Model Context Protocol, an open-source effort to make it easy to discover and use tools. Using our MCP integration, you can use the tools exposed by any MCP-compatible server in your LlamaIndex agent, in just one line of code! Install the integration: https://lnkd.in/gY4ezNni See a full demo: https://lnkd.in/gfBuQsU9 Learn more about MCP: https://lnkd.in/gx6mvJgK

Build a multilingual, multimodal RAG system with LlamaIndex and Qdrant 🌐🖼️ Learn how to create a powerful Retrieval-Augmented Generation system that handles multiple languages and modalities: 🔍 Ingest and retrieve content in English, Spanish, Chinese, and more 🖼️ Process both text and images seamlessly ⚡ Achieve fast performance with minimal user friction Key components: • LlamaIndex for flexible agent-based workflows • Qdrant for high-performance vector storage • Langfuse (YC W23) for observability and analytics Dive into the architecture, implementation details, and industry applications in this comprehensive guide: https://lnkd.in/gNyacSsX

Last week at LlamaIndex #11, the weekly column that randomly goes monthly, then back to weekly maybe. Parkour! During the past few weeks we worked on a new, long awaited feature: custom start and stop events for workflows! Posting the link to the docs in the comments, but the TL;DR; version of the story is you're not limited to use the built-in StartEvent and StopEvent to implement input and output of a workflow, you can go wild passing the custom events you want. This will unblock some use cases, with one I personally care about (more on this later) and will overall provide better autocompletion in IDEs, proper typing and more flexibility in managing what a workflow run will return. Enjoy!

We’re curating a collection of templates 📖 that show you how to build task-specific agents to automate knowledge work - and it’s available to anyone to signup and use! With just a few lines of code, you can orchestrate agent workflows that do one or more of the following: 1. Process a 100+ page slide deck and QA over any part of it (charts, tables, diagrams, images) 🖼️ 2. Extract out all line items from an invoice 🧾 3. Review all the clauses within a contract for compliance 🧑⚖️ 4. Generate a full report containing interleaving text and images 📝 And much much more. A lot of this depends on the core document processing technology enabled by LlamaCloud as well as agent orchestration through LlamaIndex workflows. It lets you build workflow that can leverage advanced retrieval, extract from unstructured documents, and retrieve from both structured/unstructured knowledge bases. Repo: https://lnkd.in/g4DWzySf Signup for LlamaCloud: https://lnkd.in/gi8dxGnt