Conversation
…Enabled is respected The Deep Think checkbox previously defaulted to false, which always overrode MIDSCENE_MODEL_REASONING_ENABLED from env config. Changed to a tri-state Radio (Auto/On/Off) where Auto means the env variable controls reasoning behavior.
Deploying midscene with
|
| Latest commit: |
4d6215c
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://39304175.midscene.pages.dev |
| Branch Preview URL: | https://fix-visualizer-deep-think-tr.midscene.pages.dev |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4d6215cc16
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| deepThink: boolean | 'unset'; | ||
| setDeepThink: (deepThink: boolean | 'unset') => void; |
There was a problem hiding this comment.
Rebuild @midscene/visualizer dist for the new tri-state API
@midscene/visualizer is consumed through its published dist/* entrypoints (packages/visualizer/package.json points main/module/types there), but this commit only updates src/*. The checked-in build artifacts are still on the old boolean implementation (packages/visualizer/dist/lib/component/config-selector/index.js still renders a checkbox, and packages/visualizer/dist/types/store/store.d.ts still declares deepThink: boolean), so apps importing @midscene/visualizer such as packages/playground-app/src/PlaygroundApp.tsx and apps/report/src/components/playground/index.tsx will keep the pre-fix behavior until the package is rebuilt and the generated files are committed.
Useful? React with 👍 / 👎.
Summary
false, which always overrodeMIDSCENE_MODEL_REASONING_ENABLEDfrom env config — users who set it totruestill sawthinking=falsein logs.Auto/On/Off):deepThink, soMIDSCENE_MODEL_REASONING_ENABLEDenv variable controls reasoning behavior. A tooltip on "Auto" explains this.deepThink=true, force-enables reasoning.deepThink=false, force-disables reasoning.booleantoboolean | 'unset', with proper localStorage persistence and restoration.Test plan
MIDSCENE_MODEL_REASONING_ENABLED=true→ reasoning should be enabled (checkai-call.log)MIDSCENE_MODEL_REASONING_ENABLED=false→ reasoning should be disabledaiActactions should never forwarddeepThink