LlamaIndex

New Webinar 🚨: A Principled Approach to RAG Experimentation + Evaluation 🧪 After 1+ years of RAG development, proper evaluation is still hard. We’re excited to partner with Weights & Biases on a webinar showing you how to build, evaluate, and iterate on a RAG pipeline. Jerry Liu, Ayush Thakur, and more will highlight different evaluation strategies with an LLM Judge against a RAG pipeline. W&B Weave is a new toolkit that gives you a comprehensive way to track and evaluate LLM applications, with features like comparisons, annotations, tracing, and more. Happening next Wednesday 7/10, 9am PT: https://lu.ma/dywrdye5

OpenContracts - a fully open-source, AI-powered Document Analytics Tool ✨ OpenContracts is an exciting project by John Scrudato that allows you to analyze and annotate docs, and also share it with the public. It’s genAI native, and uses LlamaIndex on both the query-side and data extraction side. The query-side integration lets you answer questions over the hundreds of documents. The data extraction side lets you run structured extraction on all your documents, adding additional annotations beyond the human-labeled ones. A huge use case for this is legal analysis/engineering; allows end users to manage, process, and ask questions over contracts, corporate legal docs, and more. Best of all, it’s fully free and open-source 🔥 Stack: LlamaIndex, Marvin, pgvector, Repo: https://lnkd.in/gjcQ_Zjp Docs: https://lnkd.in/gP4_Uvzc

RAG on your Raspberry Pi 🍓🔎 This is a cool tutorial if you’re looking to build a RAG pipeline that can live fully locally on a small embedded device. 🧩 Kameshwara Pavan Kumar Mantha shows you how to build a RAG pipeline that lives on a Raspberry Pi device with Docker, Inc, Ollama, Qdrant, and using LlamaIndex as the orchestration layer. The model used is gemma-2b. You might be able to obtain further optimizations through LLM quantization and caching. https://lnkd.in/gVwYGcDc

The Future of Knowledge Assistants 🤖 At the AI Engineer World Fair, we covered what it means to build a better knowledge assistant beyond using naive RAG. There are three main components: 1️⃣ Advanced data and retrieval modules: Have an advanced set of capabilities for parsing, chunking, and retrieval even before you try out fancier orchestration techniques. 2️⃣ Advanced single-agent query flows: Treat all data interfaces as tools, use agentic reasoning to build personalized QA systems. 3️⃣ General multi-agent task solvers: Build a multi-agent system as event-driven microservices in order to better collectively solve a task, whether through an agentic orchestrator or through an explicitly defined orchestrator. Along the way we released some cool announcements: - Llama-agents + a sneak peek into our LlamaCloud waitlist Slides: https://lnkd.in/gkQikWJG llama-agents: https://lnkd.in/g37FkPyx

Let’s translate your multi-agent system from Python functions into microservices 💡 Mervin Praison ✅ has created the most comprehensive video tutorial we’ve seen on our brand-new `llama-agents` framework so far ✨. This not only includes a high-level overview of the concepts, but practical coding examples to show you how to 1) setup agent services that can run synchronously in a notebook, and then 2) setup an actual agent server and query it via a client. Afterwards check out the 10+ examples we have in our examples folder, covering RAG to reflection services. Video: https://lnkd.in/gbSRfnsZ Llama-agents repo: https://lnkd.in/g37FkPyx

Launching a Multi-Agent System with Docker and Kubernetes 🤖🛠️ We’re excited to feature a brand-new starter kit by Val Andrei Fajardo on deploying a multi-agent system into production. The seamless transition from local services to k8s deployment is one of our core goals with `llama-agents`. We present three ways of launching the multi-agent components: core services (message/control plane), agent services, and additional services. 1️⃣ Launching Services without Docker: This shows our standard flow of launching servers from python code 2️⃣ Launching Services with Docker: Use docker-compose to setup the services - we provide the Dockerfile and env variable template  3️⃣ Launching with K8s: Use kubectl to launch each service as a separate pod. Llama-agents is designed to let you build agents as microservices. Use this as a starting template for your own multi-agents project! 🔥 Check out the full guide right here: https://lnkd.in/ghC9CiQn Llama-agents repo: https://lnkd.in/g37FkPyx

llama-agents is our brand-new project aimed at helping you turn your multi-agent system into microservices in production, and we want to highlight AbdulMajedRaja RS (1littlecoder) for making a day 1 introductory video helping you to easily get started 🧑🏫 It walks through our architecture of the control plane/message queue/agent services, and also walks through examples of how to set this up with LlamaIndex/llama-agent abstractions. Check it out! 🔥 Video: https://lnkd.in/gWHjMix8 Repo: https://lnkd.in/g37FkPyx

A comprehensive tour through AI agents with LlamaIndex This ~1+ hour workshop led by Val Andrei Fajardo as part of Gen AI Philippines takes you through a comprehensive tour of all the common LLM applications you can build today, with LlamaIndex as the core abstractions. Contains plenty of Excalidraw diagrams making this a great resource for beginners. Topics include RAG, Agent components, Agentic RAG, and even a WIP “multi-hop agent” in the works. We’ve got the full video, cookbook, and slides to share. Video: https://lnkd.in/gbCsHVip Notebook: https://lnkd.in/gGmxzyCM

A Beginner’s Guide on Building a Report Generation Agent 📝🤖 This is a great introductory tutorial on the set of tools you need to plug into a ReAct agent in order to build a report generator:  1. A RAG tool over an existing corpus of guideline documents 2. A web search tool to look up information from a web page 3. A report generation tool that can take markdown formatted text and generate a PDF. By Getting Started with Jeff, check out his channel: https://lnkd.in/gNsu55X6 Check out the video here: https://lnkd.in/gD4j32sB

LlamaIndex users are big fans of the Jina AI reranker, and now they've released their best reranker yet!