Hello,
I'm working on a back to basics agent creation where two functionalities are being put on display:
1. An agent who can read, draft, and send emails through Outlook to assist users in Inbox curation.
2. An agent who can capture user input and generated responses for the purpose of troubleshooting an erroring topic.
My problem is primarily related to the second functionality, but has bled into the first as well.
The instructions I'm using for my agent are simple:
"You will help users answer their questions. This includes assisting them with email facilitation. When a question related to emails is asked, you need to either summarize previously received emails for the user, and draft a potential response to the summarized email, or if you are helping the user send an email, you need to check with them about the contents of the email and who it is being sent to before you send it.
All other questions are valid as well, please maintain a professional tone when responding to user questions."
The topic that is managing functionality number 2 is simple as well:
Topic Name: Question facilitation
Trigger description: When a user asks a question unrelated to reading, summarizing, drafting, or sending emails, utilize this topic.
Inside of the topic there is a generative answer node with input of System.Activity.Text. This node is set to only use general knowledge for answer generation, and will write it's response to a string variable named 'botResponse'
Underneath this generative answer node, there is a condition that evaluates if 'botResponse' is blank. If it is not, a message is sent in the chat window that says "Success" and if it is blank, a message is sent that says "Failure"
When I ask a question of my custom agent: "What color is the sky" or "How could I start to learn to fish" my generative node returns a system message stating that there is no knowledge available to answer this question. Then I receive a "Failure" message per my condition. Once my topic has ended, the agent then posts a completely acceptable message in the chat responding to the question that has been asked.
I could use some help in understanding why my generative answer node fails while the agent itself seems capable of answering the question. The current behavior prevents me from capturing the response into a variable and utilizing it in future generative output/tools.