Dynamic prompt suggesstions

Hi Nathan V,

Thank you for posting your question in the Microsoft Q&A forum. 

Your approach of generating dynamic prompt suggestions based on previous user questions is valid and aligns with Microsoft’s recommended pattern called Suggested Actions, as described in the documentation you referenced:

Currently, there are very few articles or samples that demonstrate dynamic suggested actions, but here is a high-level approach you can consider:

• Although suggested actions are declared in JSON under the suggestedActions property, you can leverage your LLM to read the conversation context, generate suggested prompt text, and then build up to three complete action objects for the JSON payload. Finally, send these as part of the bot’s Activity response. Sample JSON
• (Optional but recommended): Since bots are stateless by default, consider implementing Conversation State to maintain context across turns. This will help your LLM generate more accurate suggestions, especially when user responses are short or ambiguous. You can find more details on state management here:
  • Save user and conversation data
  • Concept: State management

I hope this helps.

If the answer is helpful, please click "Accept Answer" and kindly upvote it. If you have extra questions about this answer, please click "Comment". 

Note: Please follow the steps in our documentation to enable e-mail notifications if you want to receive the related email notification for this thread.