Agent graph judgement Fail

[Gimlettt]

It is quite common when 3 agent graph judgement all fails, with very basic query like, "summarize this video" or "what happened in the video". How can i avoid this?

[xvrrr]

It is quite common when 3 agent graph judgement all fails, with very basic query like, "summarize this video" or "what happened in the video". How can i avoid this?

Hi, sometimes, overly simplistic requirement descriptions can cause agent graphs to fail. For the video overview function (generating a overview video), you could enter: "I'd like to create a news summary video based on existing footage. I have a collection of news video clips that I want to combine into a cohesive news segment with professional narration that matches my specific style requirements. The video should maintain a consistent news presentation format with proper voice-over that accurately summarizes the key points from the source material."

For "summarize this video" or "what happened in the video," try entering: "I want to have a video Q&A." If successfully activating the agent graph, keep ask the question "summarize this video" in the continuous dialogue window.

From a research perspective, agent graphs still have room for improvement. We will upload a workflow focusing on functional usage later, skip agent graph.