-
I have created a sales_agent_worker and instructed it to ask a set of 15 questions .. THE QUESTION SET
CODE SNIPPET OF AGENT
Steps to reproduceNo response Screenshots and logsAdditional Information
|
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
You can try sequential chats (https://microsoft.github.io/autogen/docs/tutorial/conversation-patterns/#sequential-chats). Your scenario looks extremely sequentially decomposable. Create one user-proxy and assistant agent chat for each question and a custom summary prompt for the last one to generate a structured JSON output that captures all users' inputs. Also, another comment: system message instruction to terminate the chat needs to specify a termination word, e.g., "TERMINATE". Just saying "exit the chat" won't do anything. Take a look at the tutorial chapter on termination https://microsoft.github.io/autogen/docs/tutorial/chat-termination |
Beta Was this translation helpful? Give feedback.
-
https://microsoft.github.io/autogen/docs/tutorial/conversation-patterns/#sequential-chats This helped. Also, a good prompt with GPT model 4 turbo will do the work. Thanks |
Beta Was this translation helpful? Give feedback.
https://microsoft.github.io/autogen/docs/tutorial/conversation-patterns/#sequential-chats
This helped. Also, a good prompt with GPT model 4 turbo will do the work. Thanks