As a business idea: If you take an LLM and fine-tune it for Wappler, it will be great at solving Wappler problems. The problem with ChatGPT is that when I start a new conversation, it forgets a lot of what it knows. As a result, you could sell a service based on such an LLM. If I reach that level, I’ll think about starting a business like this—if you don’t mind 
We are already doing this, the AI assistant in Wappler knows every component very well, we trained it with all the required knowledge and we do very strict code validation, when it is generating code.
1 Like
So it seems that in Wappler, you have to use a pre-trained LLM that isn't specific to Wappler. I understand that you insert a lot of different context between my request and Claude, but that's not the same as fine-tuning the model itself. Plus, the context requires a lot of tokens every time. Or did you actually build your own custom LLM?