1. User message -> Botpress processes the intent and enriches the conversation with data -> Sends data to OpenAI -> OpenAI generates a response and handles the history and RAG -> Botpress retrieves the history and manages underlying variables and Business intelligence.
 2. Alternatively: User message -> OpenAI assistants API -> Calls a function hosted in Botpress -> Botpress sends data back to OpenAI -> OpenAI replies to the user -> Botpress processes the conversation history and transforms it into structured data and handles business intelligence.

 Combining OpenAI’s function calling with Botpress intents could be extremely powerful. While this might slightly shift how Botpress is used for some users, it doesn’t mean users necessarily can’t continue using the platform as usual. It also doesn’t prevent Botpress from processing AI requests within the platform (AI task, AI intents and other cards). Instead, OpenAI would handle creating final responses and maintaining conversational continuity, while Botpress excels at integrating LLMs with business workflows.
 
I understand this might not be an easy implementation, and I haven’t provided a complete solution. However, I believe this approach could be a game changer for conversation design and alleviate much of the current workload. It could significantly enhance the intelligence of bot responses and user experience. Although it might seem the impact isn’t much, tweaking the AI conversation cards to provide users with better and more intelligent responses is where I currently spend most of my time inside Botpress, and I’m still not completely satisfied with the results. On the other hand, it was so quick and easy just setting up an assistant. What are your thoughts on this?