-
Notifications
You must be signed in to change notification settings - Fork 28
Description
Can we access your project?
- I give permission for members of the FlutterFlow team to access and test my project for the sole purpose of investigating this issue.
Current Behavior
- In Test Mode, attempt to make a streaming API call
- The call fails, giving a CORS error in the console.
This happens regardless of whether "Make Private" is on, "Use Proxy for Test" is on, or "Use Proxy for Test/Run"
This happens even with "with Credentials" turned off
This appears to be unique to streaming calls. If I make the exact same call and turn off the streaming response, it works just fine.
It does work on device.
Expected Behavior
I expect these calls to work. They used to work until about 1.5 months ago when I first opened a ticket with your support folks. This used to work without any special configuration - no Make Private, Use Proxy, or anything - I just turned on the streaming switch, and it worked.
Steps to Reproduce
- Create a page
- Create an API Call to a streaming API (I did this with OpenAI and a custom in-house streaming API)
- Ensure the streaming switch is enabled
- Create a button on the page to trigger the call
- Start up test mode
- Press the button and watch the console for the CORS error
Reproducible from Blank
- The steps to reproduce above start from a blank project.
Bug Report Code (Required)
ITESz/Hqx89goeYD1s76Y/lQhQAkKUg4WLg7m9cbFBMaC7HkEIkAOur+VG9uQu25dFVlMGaJgmc0wfTXv/PhJ/kCYQWraINoz6VpEw3xQl2RaJiyPaqSfmtfP8pYJhWayru7sR5+XdBvZXAjw1r/cq3qNleeY8aSfxBlZ7vfcPo=
Visual documentation
Environment
- FlutterFlow version: 6.6.12
- Platform: Web
- Browser name and version: Chrome/Mac: 145.0.7632.160
- Operating system and version affected: SameAdditional Information
I have two projects where I need to fine-tune streaming AI API calls, and it's MUCH easier to do this on test mode where I can refresh and test new prompts quickly, so this ticket is a very big deal for me. I also cannot easily test outside of Flutterflow (via Postman, etc.) because the calls happen within a series of calls/functions that utilize some of the more advanced OpenAI features like function calls and vector stores.
I have already stalled both projects for over a month as I worked with your support team to try to resolve the issue before they gave up, so I would very much appreciate some help on this. With your support folks (Leo), we attempted to use Make Private as well as some of the other Advanced Settings for the API call, but none seemed to affect this behavior.