Jerry Liu
@jerryjliu0
co-founder/CEO @llama_index
Careers: https://t.co/EUnMNmbCtx
Enterprise: https://t.co/Ht5jwxSrQB
ID:369777416
https://www.llamaindex.ai/ 07-09-2011 22:54:31
4,2K Tweets
44,2K Followers
1,3K Following
hackathon time: MongoDB with support of Fireworks AI, Upstage, and LlamaIndex π¦ in SF
letβs see what I can learn 𧡠β it will be a long day: 9:30am breakfast and planned finish at 10pm
JSON structured output from invoice document with Llama3:8b Instruct. Running in Sparrow RAG with LlamaIndex π¦, Weaviate β’ vector database and ollama π
Katana
Excited to feature a fantastic tutorial on building a local coding agent π€π» by Tech With Tim!
This agent will take in user requests and write code.
β
It is RAG-augmented: equipped with a LlamaParse + LlamaIndex π¦ powered RAG tool to help look up relevant information from
Build a full-stack local RAG application with Llama 3 π§΅
Stack: LlamaIndex π¦, ollama, Streamlit, Lightning AI β‘οΈ
Llama 3 is the most powerful open-source LLM to date, and the best part is you can run it completely locally!
Amazing work by Akshay π for a day 1 tutorial
Learn how to build an agent that reads your documentation to write code for you! In our latest collaboration with Tech With Tim, you'll learn how to:
β‘οΈ Set up local LLMs with ollama
β‘οΈ Parse documentation with LlamaParse
β‘οΈ Build an agent
β‘οΈ Teach the agent to write code!
arxiv.org/abs/2404.11792
Real-Reasoning RAG, or Where to Get Performance Gains Out of RAG (from A I T O M A T I C & IBM Research, #AIAlliance ):
β’ Fine-tuning the retriever model gives you more bang for your buck than the generator model.
β’ Employing reasoning yields significant
To get the best performance out of Llama 3 (AI at Meta), you need to apply special tokens!
Our day 0 LlamaIndex π¦ cookbook lets you build RAG with Llama 3 with in-built support for special tokens, like end-of-turn and begin-of-text.
The capabilities are impressive. The 8B and
Meta'a Llama 3 model is released and we have day 0 support! Ravi Theja and Logan Markewich put together this cookbook showing how to use Llama 3 directly from Hugging Face, from running basic prompts to a full RAG pipeline:
docs.llamaindex.ai/en/latest/examβ¦
500 stars by Day 2 π€― We're working hard to implement all the Firecrawl feedback
Here is what we shipped today:
π§Ή - Remove non content elements (option)
π - PDF parsing powered by LlamaIndex π¦
πΌοΈ - Absolute paths on images
π - New docs website powered by Mintlify
π§΅π
Build RAG, Function Calling, and Agents with LlamaIndex π¦ and Mistral AI 8x22b - the most powerful Mistral AI model to date, and the best model in its (active) parameter class so far!
Features: 64k token context, multilingual, coding, function calling, math, and more. Uses 39B