This a Streamlit app that leverages cool new AI things to answer questions. It searches the internet for resources and collects information from multiple sources into one coherent answer. It uses a Deep Lake vector store as a knowledge base and LangChain to interact with GPT-3 for searching the internet and the knowledge base to answer questions.
This bot consists of two systems. One system searches the internet for resources such as documentation, blog posts, and tutorials relevant to your question then stores them in a Deep Lake vector store. This vector store becomes a knowledge base for your work and life. This system uses GPT-3.5-Turbo to refine the searches and better understand which resources are relevant to you.
The other system queries the vector store for documents relevant to your question and uses GPT-3 to extract information from each of those documents and combines the information into one well-informed answer.
Use Poetry to install the dependencies. Poetry installation instructions here.
Clone this repo:
git clone https://github.com/mcleonard/docbot.git
Then with Poetry installed:
cd docbot
poetry install
You'll need Python 3.10 or greater. You'll also need an OpenAI API key. Set it as an enviroment variable, or define it in a .env file.
export OPENAI_API_KEY=<your key here>
or in a file called .env in the same directory as the app.py script:
OPENAI_API_KEY=<your_key_here>
Run it with Poetry:
poetry run streamlit run app.py