LlamaIndex Newsletter 2024-05-07

Hello LlamaIndex fam! 🦙

If you’re in SF, join us for the first-ever Llama 3 Hackathon (invitation here)! Shack15 is an amazing venue and it’s sure to be a fun time. If you can’t make it, stay tuned for the rundown on the cool projects that come out of the event. Now, on to the highlights:

🤩 The highlights:

• LlamaIndex.TS hits v0.3! Loads of new features inside!
• LlamaIndex Python hits v0.10.34! A bumper release!
• That’s a lot!

✨ Feature Releases and Enhancements:

Two big releases this week!

• LlamaIndex.TS hit version 0.3! Tweet, Blog post
  • Features:
    • Agent support including ReAct, Anthropic and OpenAI agents, as well as a generic AgentRunner class
    • Standardized Web Streams compatible with React 19, Deno, and Node 22
    • More comprehensive type system
    • Enhanced support for deployment on Next.js, Deno, Cloudflare Workers and Waku
• LlamaIndex Python hit version 0.10.34! Tweet
  • Features:
    • Introspective agents that work through reflection
    • Support for huggingface text-generation-inference API
    • Structured planning agent
    • A chat summary memory buffer
    • Rust-based LLM support with Mistral.rs
    • Milvus sparse hybrid search
    • Google Firestore vector store support
• A new LlamaPack for the Reflection Agentic Pattern. Tweet

🎥 Demos:

• Filter AirBnB listings using natural language with this open-source demo! It uses Mistral AI’s Mixtral 8x7b and Qdrant engine, plus Streamlit to build UI. Tweet, Blog post
• Fully local RAG with Llama 3, Ollama and LlamaIndex! A short, sweet guide. Tweet, Blog post
• Fine-tune your embedding model using labels from a reranker. Tweet, Blog post

🗺️ Guides:

• Hanane Dupouy walks us through building an agent that can perform complex financial calculations. Tweet, Slides
• Plaban Nayak sets up a local, open-source RAG pipeline that uses Llama 3 and Qdrant to demonstrate how to improve the accuracy of your RAG with reranking. Tweet, Blog post
• Jason Zhou talks about the components needed for agentic RAG. Tweet
• Divyanshu Dixit walks us through agents dedicated to workflow automation. Tweet, Blog post

✍️ Tutorials:

• Tyler Hutcherson of Redis and our own Laurie Voss walk you through building agentic RAG with semantic caching and other production-ready techniques. Video, Notebook
• Cleanlab has a tutorial on getting trustworthiness scores from your RAG pipeline to allow you to avoid hallucinations and course-correct. Tweet, Notebook

🎥 Webinars:

• On May 8 we’ll be co-hosting a webinar with Pulumi on deploying AI applications to AWS. Tweet
• Our own Andrei and friends walk you from basic RAG through handling long-context RAG all the way to evaluating your RAG pipeline. Tweet, Video, Notebook