This is a great example of building an AI agent that is trained on a companies corpus of proprietary data - in this case a library of software user guides and processes in the form of PDF vectors and return the right information complete with citations and links to the user based on their question. This creates an efficiency for staff looking for information and being able to simply ask an AI agent for the information rather than having to search through the intranet or file directories to find the right user guide for the information they need.
Using API keys this example shows Chat GPT embedded inside Autodesk Revit via the Dynamo computational programming interface. Imaging having google search built into end user software so they can ask contextual questions whilst working without having to navigate to a web browser and search through multiple results.
This example shows a simple custom GPT created to search and detail the best coffee based on the user's question. Looking for the best coffee on Clarence street in the Sydney CBD? Easy. What about pet friendly cafe in downtown New York? Sure. Coffee Scout scours the internet to provide up to date recommendations for coffee shops and lists them based on highest google reviews