DuckAi/ChatTab: Add New Image from + menu - #9773
Conversation
|
Warning This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
This stack of pull requests is managed by Graphite. Learn more about stacking. |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 5f564fc. Configure here.
Fixes a conflicting-overloads compile error in duckchat-impl tests. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…uild The force-image-generation flag was consumed on every showDuckAI, even when showNativeInput early-returns because the widget is already attached. That let an already-open Duck.ai tab (whose widget is shown) clear the one-shot before the new tab could apply it, so the target tab never preselected image generation. Gate the consume on isNativeInputShown() so the flag is only spent when the input will actually be built. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
malmstein
left a comment
malmstein
left a comment
There was a problem hiding this comment.
Works as expected. I have two comments about the keyboard and the persistent flag stored in the Activity. None are blocker but worth discussing it.
We could leverage the native input state to store those flags if necessary, as opposed to keep flags at the activity / fragment level. In the end we are reading those from the input manager which is able to access the state.
At the bare minimum, we could have a config repository for these things that can be interfaced, injected and tested.
| // finishes before the enter morph. | ||
| pendingEnterOwnsAnimating = enterStarted | ||
|
|
||
| if (forceImageGeneration) { |
There was a problem hiding this comment.
is there a reason we don't do that for all modes? There's a bug report about the keyboard https://app.asana.com/1/137249556945/project/1213881875984009/task/1218357434084408?focus=true I wonder if we can fix it here?
There was a problem hiding this comment.
Can do this as a follow up 👍
| // One-shot carried from an "open Duck.ai for image generation" launch. The Duck.ai tab this | ||
| // launch creates doesn't exist yet, so the flag is held here and consumed by that tab's fragment | ||
| // the first time it shows the native input (see BrowserTabFragment.consumeDuckAiForceImageGeneration). | ||
| private var pendingDuckChatForceImageGeneration: Boolean = false |
There was a problem hiding this comment.
should this live in the native input state instead? not a blocker, but wonder if we can avoid adding flags to represent state outside of the native input
2 participants
Task/Issue URL: https://app.asana.com/1/137249556945/project/1212087397361015/task/1218350178023317?focus=true
Tech Design URL (if applicable):
API Proposals URL(s) (if applicable): https://app.asana.com/1/137249556945/project/1201462763415876/task/1218365860196482?focus=true
Description
Steps to test this PR
nativeDuckAiSidebar enabled — entry visibility
nativeDuckAiSidebar enabled, currently selected model already supports image generation
nativeDuckAiSidebar enabled, selected model does NOT support image generation but another accessible model does
nativeDuckAiSidebar disabled (via FF inventory)
Note
Medium Risk
Changes global Duck.ai model selection and new-tab launch plumbing; behavior is gated on model capabilities but affects core chat entry flows.
Overview
Wires the Duck.ai + menu New Image action to open a fresh Duck.ai tab with image generation ready to use, instead of leaving a TODO.
DuckChat.openDuckChatImageGenerationpicks an accessible image-capable model when the current one cannot generate images, then launches a new session viaBrowserNavwith a one-shotforceImageGenerationintent extra.BrowserActivityholds that flag until the new tab’s fragment consumes it on first native input show; the native input stack preselects the image-generation tool and auto-focuses the field when the flag is set. If no suitable model exists, it still opens a normal new chat without forcing the tool.BrowserTabViewModel.openNewImageDuckChatroutes analytics entry points for NTP vs in-page Duck.ai. Tests cover model-switch behavior, navigation flags, widget configuration, and menu handlers.Reviewed by Cursor Bugbot for commit 8534163. Bugbot is set up for automated code reviews on this repo. Configure here.