This looks interesting. If I’m understanding, it takes the raw html elements of a page and saves them into a RAG? Is this because feeding the whole page into an LLM prompt would be bigger than the content window?
Also love that it has both backend source code and a service available
It can work both with html or even the user data in json/array in React/Angular kind of applications.
Feeding the whole page doesnot always make sense, it can have repetitive menu, and layout stuff. It give developer control on what they want to feed to the prompt. End of the day everything have a cost attached, this will help optimise cost to the last mile.
It deeply integration with inAPP, and have native Functions and Backend APIs. It can perform InApp UI/UX workflows based on user query.
A lot of AI assistants are just chatbots which are working in completely isolated thread.
In code various functions are registered with their function and parameter definition. Based on text/voice based query, the framework uses OpenAI function calling to invoke the right function.
Looks great. Had an issue getting it to work though. iPhone, microphone and speech recognition enabled. Says it adds to the list, but the list not changing. Any ideas?
Also love that it has both backend source code and a service available