GitHub - pyladiesams/intro-RAG-elastic-apr2024: Learn what is RAG and how it can increase trustability of LLM apps (original) (raw)

An introduction to RAG with Elastic

Presentation: Introduction to RAG

Workshop description

During the workshop you will learn what is Retrieval Augmented Generation, how it can increase trustability of LLM models and how to set up a RAG pipeline using Elastic.

Requirements 🧰

For this workshop, you will need:

• Python 3.6 or later
• An Elastic deployment
  • We'll be using Elastic Cloud for this example (available with a free trial)
• OpenAI account. Where do I find my OpenAI API Key?

Usage

• Clone the repository
• Start your favorite IDE and navigate to the solutions folder
• Run pip install -r requirements.txt
• Launch Jupyter App jupyter notebook

Notebooks

Question answering

In the question-answering.ipynb notebook you'll learn how to:

• Retrieve sample workplace documents from a given URL.
• Set up an Elasticsearch client.
• Chunk documents into 800-character passages with an overlap of 400 characters using the CharacterTextSplitter from langchain.
• Use OpenAIEmbeddings from langchain to create embeddings for the content.
• Retrieve embeddings for the chunked passages using OpenAI.
• Persist the passage documents along with their embeddings into Elasticsearch.
• Set up a question-answering system using OpenAI and ElasticKnnSearch from langchain to retrieve answers along with their source documents.

Chatbot

In the chatbot.ipynb notebook you'll learn how to:

• Retrieve sample workplace documents from a given URL.
• Set up an Elasticsearch client.
• Chunk documents into 800-character passages with an overlap of 400 characters using the CharacterTextSplitter from langchain.
• Use OpenAIEmbeddings from langchain to create embeddings for the content.
• Retrieve embeddings for the chunked passages using OpenAI.
• Run hybrid search in Elasticsearch to find documents that answers asked questions.
• Maintain conversational memory for follow-up questions.

Video record

Re-watch this YouTube stream

Credits

This workshop was set up by @pyladiesams and @ahavrius