I'm encountering an issue with a Forge app integrated into Confluence Cloud via a ROVO agent.
The app is designed to search content from a .txt file hosted on an external server (it will be using Joomla API later on, after this first stage is working). The Forge function has been updated and successfully deployed, and when tested independently (e.g., via forge tunnel), it correctly fetches and parses the file, returning relevant results based on the query.
However, when the ROVO agent is used in Confluence, it consistently returns a generic success message like: "Your Joomla content search has been executed successfully. There were no links provided with this action..."
This happens even when the query clearly matches content in the file, and the function logs confirm that results are being found and returned in the expected format.
What I've Tried:
Verified the function is deployed and logs show the correct version is running.
Confirmed the function returns a body object with a results array and a message string.
Added debug logs and version markers to ensure the correct function is being invoked.
Checked the manifest to confirm the ROVO agent is linked to the correct action and function.
Could you please help me understand:
Why the ROVO agent might not be interpreting or displaying the results returned by the function?
Whether there are any formatting or structural requirements for the results array that the agent expects?
If there's any caching or propagation delay that might cause the agent to use an outdated version of the function?
Any guidance or documentation on how ROVO agents parse and display Forge function responses would be greatly appreciated.
Thanks!