Learning Information

You can add the ability to learn things to your agent. When you do, each new instance of your agent will maintain its own set of vector databases in which to put the information it learns. These vector databases can be used in a number of ways, but the most common is to combine it with a question answering tool.

Each section in this chapter covers a different way you may want your agent to learn information and how to accomplish it:

• Learning facts from user conversations
• Learning from PDFs
• Learning from YouTube videos

What you can build with this chapter

Together with the Question Answering chapter, this chapter will let you build:

• An agent that answers questions about your PDFs or Books
• An agent that answers questions about your YouTube channel
• An agent that answers questions about facts you've uploaded via API
• An agent that remembers specific things for its human users and can answer questions about them later

You can combine these skills with personalities to:

• Acts like a character from a book, with knowledge from having learned the book
• Acts like your favorite podcast host, with the ability to answer questions about specific interviews from having learned the YouTube videos
• Adopt a socratic tone, acting as a tutor who gives some, but not all, of an answer
• Acts as a friend who remembers your preferences and brings them up when chatting

You can combine these skills with image generation to:

• Generate imagined pictures from books or PDFs
• Fetch and send Google Image Search results relevant to an answer