I’ve been developing an app for ChatGPT with Apps SDK and sometimes the agent decides to perform multiple tool calls that each can render its widget in a single chat turn.
This works OK on the web (both desktop and mobile), but fails quite randomly inside the native mobile app.
Expected behavior:
- Multiple widgets are correctly rendered.
Actual behavior:
- 1 o 2 widgets are rendered, rest are not for no reason at all (no Sentry errors, widget code is directly not being initialized)