Hi everyone,
My name is Nick, and I’m a product manager on the editor team. Today we’re releasing Bubble Version 33, which updates how the API Connector detects errors on streaming API calls.
What changed:
When an API endpoint returns a streaming response, sometimes, the server will send an event with “event”: “error” to signal that something went wrong mid-stream. This means an API call can start returning data successfully and then fail partway through.
In previous versions, if the initial start of the api call was successful, we wouldn’t capture any mid stream event errors, which means if there was an error mid stream, and you have “capture errors” set, the api call would still be marked as fully successful.
With this update, the API call will throw an error if the stream receives an event with “event”: "error”. Text from the stream can still populate up until the error event is received. Non-streaming API calls are not affected by this change.
What this means for you:
If you’re using streaming API calls (common with LLM providers like OpenAI and Anthropic), your error-handling workflows will now correctly detect mid-stream failures. If you have workflows that check _returned_an_error, they will now fire in cases that were previously missed.
For most apps, this is a strictly better experience — errors that were silently swallowed will now be caught. That said, if you’ve built workarounds for this behavior, it’s worth testing your streaming workflows after upgrading to make sure everything works as expected.
You can upgrade to version 33 in Settings > Versions.
As always, we welcome your feedback. Thanks for building on Bubble!
Nick
