Day 7: Improved App UX and Persistance with Agent Task Decomposition

Now that you've restructured your AI app through Task Decomposition, you have much better control but as it currently stands the UX is not great.

In Day 7, we will make our apps internal workings (the steps it takes) visible to the user, ensuring our telemetry is insightful, and persist conversations so that they can be properly continued.

The first step is to show the Agent's throughts to the user. We'll modify our getNextAction function to accomplish this so that when the llm has completed a step it can notify the user with proper information.

One bug you might have noticed up to this point is that if you send multiple messages in a chat, the LLM doesn't remember previous conversation. We'll fix that by providing chat history through SystemContext.

And finally we will persist conversations, including agent steps, to our database so agents and users can take advantage of past conversations.

By the end of Day 7, your agent will not only be more powerful internally but also more transparent to the user and able to maintain conversational context and persist its state.