AI Agent: Follow-up Questions output is incorrect
Summary:
I’ve been testing the Follow-up Questions feature, but it doesn’t seem to recognize all the conditions. The script below is what I used, but it always processes only the first condition, even when it hasn’t been met.
Use the following system context variable exactly: $param.system_context.chat_history
Perform these checks in order. All matching must be case-insensitive. Do not print, log, or output anything except the exact JSON specified below when a condition matches. If no condition matches, output nothing (empty response).
- CONDITION A — "list of invoice numbers with hold"
If the chat history contains evidence of a list of invoice numbers that are shown as having holds, output exactly this JSON (no extra whitespace, no markdown, no comments):
0
