We're going to be building a very simple one shot Q&A chatbot that will help you get to grips with Retrieval Augmented Generation (RAG), Vector Databases, Text Embeddings and LLMs.
During the session we'll load the contents of a book into a Vector Database and construct a Chatbot that can:
- receive a user question
- retrieve relevant content from the stored book to answer the question
- use an LLM to generate the answer based on that content.
All in 90 minutes!
It helps to be set-up before the session.
Download this repository to your device. Open the root folder for the project in whatever IDE you usually use.
We'll need OpenAI to create Embeddings and complete LLM calls to generate answers.
Go to https://openai.com/api and log in or sign up.
If you sign up you'll get free credit for the API.
If you signed up sometime before, you'll just need to check you have credit available. The next step will let you know if that's the case. If not, you might want to just put the minimum amount of credit on your account or just collaborate with another group on the day.
See if you can successfully navigate to the Open AI Playground at:
https://platform.openai.com/playground?mode=chat.
Can you enter a chat question in the 'User' field, press submit and get a reply?
Find your OpenAI API Key at https://platform.openai.com/api-keys. You'll likely need to create a new one and copy the key.
We will use dotenv to store the keys, so edit the .env file to include your OPENAI_API_KEY
OPENAI_API_KEY=<<your-api_key>>
We will use Pinecone as our Vector DB.
Go to https://www.pinecone.io, set up a free account and get an API key. This will either be an option on set-up or go to API KEYS menu item on the left hand nav.
Edit the .env file to include you PINECONE_API_KEY
PINECONE_API_KEY=<<your-api_key>>
Follow the instructions using venv or conda to set up the Virtual Environment or do it the way you would normally.
Within the root directory, set up a Python Venv, you will need Python 3.8 or greater. If you are on a lower python version you will need to upgrade. Follow these instructions to set up the Virtual Environment or do it the way you would normally.
Run venv from the terminal
python -m venv venv
Then, on a Mac, run the following from the terminal
source venv/bin/activate
or a PC, run the executable from a terminal
venv/bin/activate
Again, from the terminal, install the packages:
pip install pinecone-client openai python-dotenv
From within the project folder create a new conda environment using the provided environment.yaml file. On the terminal:
conda env create -f environment.yaml
then activate the environment using the command
conda activate rag-workshop
Go and run the code verify_setup.py located in this (the root) directory.
It should all work and if it does, you are ready to start following Exercise 1 to 4 found in the directories in this project.
Enjoy!