Are LLMs the shiniest toys around that everyone wants added to their solution ASAP? Or are the potential hallucinations, data security concerns, and steep learning curve still putting you off from giving into the hype?
There’s no better fix than just trying it out yourself.

This is your chance to dedicate a few hours to understanding and experimenting with the necessary building blocks and running your own RAG solution. At the end of this workshop you’ll have your own running web-based retrieval-augmented assistant, grounded on your personal data. Not only that, since we’re designing a modular solution, you can easily swap models, data handling, prompts, or design choices to adapt and evolve your app whenever you want to experiment with the new kid on the block.
Cure your GenAI FOMO and check off that “figure out this LLM thing” off your to-do list today!

## Overview
* Quick breakdown of LLMs, vectors & embeddings, semantic search, GenAI, popular open source tools, models & frameworks
(i.e. Llama, Mistral, Gemma, HuggingFace, LangChain, etc)
* Model selection up to the user - customising your app according to desired use case w/ starter code blocks
* Experiment with chunking strategies & data input
(optional to integrate with data stored in a different source)
* Build RAG architecture with LangChain & ElasticSearch
(optional to explore other chain options)
* Build web interface with Python + Flask
(optional explore alternatives like JS and PHP for front end w/ additional repos)
* Test model & responses, prompt engineering, introduce testing pipeline & observability intro

## Requirements & Materials

* Class will follow a GitHub tutorial w/ worksheets & code samples
* Bring your laptop, have a code editor, Python installation, and Docker setup
* Pre-requisite work (options & instructions will be included in repo):
* Download the LLM(s) of choice (default will be ollama + llama3.1)
* Install ElasticSearch (locally or via cloud)