Hey Wappler Gang…
My app is huge, complex and ever changing… and I’d like to train some AI thing to be my sidekick when it comes to customer support.
I need it to learn all I know about the nuances of how to use the app, to learn new features as I add them, and to forget about old ways of doing things when I make a change.
I’d love some input on how to go about this, as I’m an AI newbie!
Maybe this is a cool way… using a Google Agent via their Vertex AI…
The AI, as in the example, must have training data. If you have existing support docs, help files, Q&A's, those would serve, like in the video of the 2025 budget, to train the AI's language model. Google's AI can also learn from videos on YouTube (but I think this is purely based on transcription currently). So if you have training videos on how to use the app, that may be an additional option to training the model.
I think what you're really asking for, is a way to allow the AI to look at all of your code, and watch you code new features, and be able to answer questions typical of a user based on its knowledge of your code. That would be extremely cool...using your code as the training data rather than help docs.
Google's gemini does have some awareness of Wappler, but that's only because of the docs and community posts available it can train on. It seems to really rely on the underlying technologies to answer questions, like Bootstrap, HTML, CSS, JS, Node, PHP, Docker, DB's, etc. The effectiveness of the AI "agent" is only as good as the input its trained from. That's been true ever since machine learning became a thing.
I'm not sure if any of the tools will be able to understand the context of your code by itself, but tools like Cassidy can help with customer support based on info in knowledge bases. Unfortunately it only integrates with a handful of platforms at the moment. For more complex workflows and integrations tools like LangGraph and crewAI may be more appropriate.
This made me laugh...