You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We have migrated our existing AutoGen app to Agent Framework using the version 1.0.0b260127 (Python). This migration went quite smoothly, and we got the app working on Agent Framework.
However, when we upgraded to RC2 (1.0.0rc2), the functionality of the app did break. After investigations we realized that there is a fundamental change in broadcasting tool outputs in GroupChat. In other words, tool outputs are not broadcasted anymore for downstream agents. This breaking change has been introduced in PR #3712 without any big headlines.
Broadcasting tool outputs to downstream agents would be beneficial in many real world scenarios. For example, if agent A runs a tool with a vast output and then agent B analyses it. Now the only workaround for this is to prompt the agent A to include the tool output in its response causing more token consumption and performance loss.
Please consider making tool output broadcasting a configurable feature in workflows. It could be even configurable per agent or executor.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
We have migrated our existing AutoGen app to Agent Framework using the version 1.0.0b260127 (Python). This migration went quite smoothly, and we got the app working on Agent Framework.
However, when we upgraded to RC2 (1.0.0rc2), the functionality of the app did break. After investigations we realized that there is a fundamental change in broadcasting tool outputs in GroupChat. In other words, tool outputs are not broadcasted anymore for downstream agents. This breaking change has been introduced in PR #3712 without any big headlines.
Broadcasting tool outputs to downstream agents would be beneficial in many real world scenarios. For example, if agent A runs a tool with a vast output and then agent B analyses it. Now the only workaround for this is to prompt the agent A to include the tool output in its response causing more token consumption and performance loss.
Please consider making tool output broadcasting a configurable feature in workflows. It could be even configurable per agent or executor.
Beta Was this translation helpful? Give feedback.
All reactions