LlamaIndex

Build a Financial Analyst Agent System using crewAI and LlamaIndex 📈🤖 Here’s a neat tutorial by Pavan Belagatti showing you how to build a multi-agent system. Starting with CrewAI, you can define a top-level financial analyst agent as well as a content strategist that synthesizes the final report. Our LlamaIndex integration lets you plug in any query engine as a tool into this agent. YouTube: https://lnkd.in/gmbidczF Notebook: https://lnkd.in/gXcrAGU3

A Step-by-Step Guide to Building Text-to-SQL Applications 👣🔎 Here’s a step-by-step guide by Avi kumar Talaviya on building robust text-to-SQL applications, shown in the flowchart below 👇 It extends beyond a simple prompt call - give a database collection with a lot of tables, you want to dynamically retrieve few-shot table examples so the LLM has the relevant metadata to query the right tables. LlamaIndex offers both low-level and higher-level modules around SQL retrieval. Check out the blog here: https://lnkd.in/gXvgiiXY For an even more advanced guide where you feed not only tables but also rows as few-shot examples, check out our query pipeline guide: https://lnkd.in/gXesdBuf

This weekend, learn about 5 different ways of evaluating your RAG systems. zhaozhiming takes you through a comprehensive tour of the different RAG evaluation methods using LLM-as-a-judge (which have corresponding LlamaIndex implementations): 1. Answer Relevance 2. Context Relevance 3. Faithfulness 4. Correctness 5. Pairwise Comparison Along with synthetic dataset generation. Check it out here: https://lnkd.in/gdQ3fZ2W

We see an emerging role needed to bring context-augmented LLM applications to production: an AI data engineer 🧠 Any agent requires a core data and tool layer to operate well. An AI data engineer needs the data engineering skills to understand how to build scalable, reliable data pipelines in production - like RAG parsing, chunking, indexing, and to add enterprise features like auth/access control. But they also need the AI chops to understand how to setup and experiment with data and tool parameters, from chunk size to function signatures, to understand which decisions best impact the performance of the e2e application being built. We’re looking for AI data engineers - come join us at LlamaIndex! 🦙 https://lnkd.in/gD-g9HqJ

We’re big on the LLM ETL stack. We’ve made two big releases in LLMs + structured outputs this week 💫 1. Open-source structured outputs with async + streaming: https://lnkd.in/gpHSTbCb 2. LlamaExtract: https://lnkd.in/gVg5gahy We want your input 📣 - what are the biggest enterprise use cases that you’re using structured extraction and output capabilities for? Whether you’re using a framework like LlamaIndex or Instructor or a tool like Deasie. Is it for a standalone use case or to make an e2e pipeline like RAG better? Some thoughts: 1. Add metadata for RAG 2. Invoice processing 3. Report generation 4. Form filling Let us know your thoughts 👇

Multi-modal RAG: search over text and images. In this video Engineer Prompt covers: ➡️ Using the CLIP model to create a unified vector space for text and images ➡️ Using OpenAI embeddings for text, and Qdrant as the multimodal vector store. ➡️ Retrieving relevant text chunks and images based on user queries Check out the video here: https://lnkd.in/gbKMrrrf Or head straight to the notebook: https://lnkd.in/gJpDuKaM

Access control over data is a big requirement for any enterprise building LLM applications. LlamaCloud makes it easy to set this up. LlamaCloud lets you natively index ACLs through our data connectors - for instance, we directly load in the user/org-level permissions as metadata in Sharepoint 🔐 It’s also easy to inject custom metadata through source documents, or programmatically have each user modeled by a separate index within the same project, and have this natively integrated with the downstream vector store. Check it out here: https://lnkd.in/gqkaWYuF Waitlist signup: https://bit.ly/llamacloud LlamaCloud: https://lnkd.in/gi8dxGnt

Automated Structured Extraction for RAG ⚙️⛏️ Combine automated metadata extraction + a RAG pipeline and get better retrieval results. Define any schema through Pydantic or infer it automatically over a “train” set of docs. Then use LlamaExtract for metadata extraction, attach it to each document (which is parsed through LlamaParse to help preserve text/images/charts/tables). This allows you to do metadata filtering (either manually or via an LLM) and also provide more context to the LLM for better synthesis results. Notebook: https://lnkd.in/g7hgr6mD LlamaExtract: https://lnkd.in/gVg5gahy

gpt-4o-mini is the best performing model for its weight class in multimodal parsing. It’s able to extract tables from complex bar charts 📊 and diagrams, such as from the Llama 3.1 blog post. It’s available as a core model in anyone to use in LlamaParse, at 1.5c per page (or 0.3c per page if you provide an API key). It provides a faster/cheaper option than bigger/heavier models like GPT-4o or Claude 3.5 Sonnet. Check out this notebook for a full tutorial on how to get started: https://lnkd.in/gm9pyMPX Signup for LlamaParse: https://lnkd.in/gi8dxGnt LlamaParse repo: https://lnkd.in/g3UmUkcD

Today we’re excited to introduce an early preview of LlamaExtract 🦙🔬, a managed service that lets you perform structured data extraction from unstructured documents. Infer a human-editable schema from a candidate set of documents. Given this schema, extract structured values from new documents. Structured extraction is a huge part of the LLM ETL stack for any RAG/agent use case. We built LlamaParse to solve document parsing but increasingly realized the 1) need for good quality metadata in RAG pipelines, and 2) the potential for LLMs to automate transformation. LlamaExtract is available in beta both via a UI playground as well as API to everyone. Usage is counted the exact same as LlamaParse! This is just the start; we’re making huge improvements soon on every front💡 Check out our launch blog post: https://lnkd.in/gzfR5xUW  Signup here to start using (no waitlist needed): https://lnkd.in/gi8dxGnt If you’re interested in the broader LlamaCloud waitlist, check here: https://lnkd.in/gFyTJeCc