Learn RAG with LLMWare (2024)

Interested in learning more about Generative AI? Getting started with Generative AI is fast and simple with LLMWare!

To use Generative AI skillfully in most work environments, you need to be able to link a large language model (LLM) to a knowledge base. Retrieval Augmented Generation (RAG) is the method of linking such a knowledge base (usually some form of documents such as PDFs, PowerPoint, Word or Excel) to query to a model.

In this step-by-step course, learn all the key components of RAG such as knowledge ingestion, parsing, indexing, embedding and prompting.  This is a foundational course in working with Generative AI so that you can start to build effective workflows and chatbots.

The course has 7 short videos:

1. Intro Video:

Learn the basics of Retrieval Augmented Generation in this Step-by-Step series that will teach you all the fundamentals of how to get started with Generative AI.

Following our in-depth tutorial series of 7 short videos, you will learn the most important components of learning to work with an LLM-based framework using LLMWare’s easy, integrated open source library.

In this intro, you will learn how to pip install LLMWare, the basic requirements to get started, and get an overview of the different components of RAG. Learn how to create a state of the art data pipeline using the latest LLMs, all on your laptop.

2. Parsing, Text Chunking and Indexing

Create your first library for Retrieval Augmented Generation and get started by learning the first basic steps of parsing, text chunking and indexing documents in this easy-to-follow tutorial.

3. Build Embeddings

What are embeddings, embedding models, vectors and vector databases? In this Example 2 tutorial, learn the fundamental concepts behind embeddings, embedding models, vectors and vector databases.

With our easy-to-follow example, you will also build your first embeddings with embedding models from Hugging Face and store to a database, and use these embeddings to run your semantic queries.

4. Prompt Models

Learn how to prompt models. In this example, you will learn how to load and start prompting models from Hugging Face and OpenAI.

Start inferencing models using this example and see how you can check if the model is providing the right answer using the context that was passed with LLMWare’s similarity score. Learn also how to capture model usage data such as token consumption, the output, the total processing time, etc. in downloadable files for future auditability.

5. RAG with Text Query

Start searching! In this example, we will be taking some form of knowledge in a library with embeddings and bring the pieces together with a model. Learn how to put together the right RAG strategy with a thoughtful retrieval and querying strategy combined with the right model to do the job.

We will use LLMWare’s smallest (but mighty), laptop friendly BLING 1B parameter model from Hugging Face to run contract analysis. Learn how we automatically extract names of parties from contracts and other key information with text query.

6. RAG with Semantic Query

Start semantic searching! Also known as natural language querying, this is where we reap the benefits of embeddings and vector databases. You will be able to query your knowledge base using natural language to ask questions like “what is the executive’s base salary?” to derive answers from complex employment contracts.

Learn the fundamentals of semantic searches in this easy-to-follow example.

7. RAG with Multi-Step Query

It’s Graduation Day if you have been following along with our other examples! We will quickly recap through all that you have learned in the previous videos and learn how to use a quantized DRAGON-Yi-6b-GGUF model from Hugging Face on a laptop.

Perform multi-step hybrid queries to get the responses you need.

Learn also how to perform evidence comparison and how to save all the output as JSON or CSV files for future datasets or auditing.

After following this series, you will be ready to leap to the next launching pad for creating the most sophisticated enterprise-ready AI workflows with LLMWare.

Get started on your AI journey today!