Trying out the new Visual Editor, and the LLM does not seem to understand the context of the changes I’m asking it to make.
Steps to Reproduce
Select an element, use the CSS inspector to change something, then hit apply.
Expected Behavior
I would anticipate the LLM to go away and start making the changes, like shown in the new marketing video. However no matter how many ways I try, the LLM just asks for clarity on what I’m asking it to do, despite the fact I can visually see the change I’ve made on the screen, and through the new workflow instructed the change in the chat box via the new visual editor.
It looks like this is a known issue with the Visual Editor in early 2.2 builds. After Apply, changes from design mode don’t flow correctly into the agent’s context - it only sees an aggregated message like “14 changes” instead of the specific diffs.
Could you please share:
The Request ID from the chat where the agent misses the changes (chat menu in the top-right > Copy Request ID)
A screenshot of what appears in the chat after you click Apply in the Visual Editor
Which model you’re using
The Visual Editor is a very fresh feature, and the team is actively gathering feedback for improvements. I’ll pass this to the team.
