Updates: fixed on 20 Nov, thanks to the OpenAI team.
Hi, the ChatGPT App was working early today, but now always shows “Error in message stream” message with a retry button.
Console log:
Failed to load resource: the server responded with a status of 500 ()
k4027n9ylcc9lece.js:31 Uncaught (in promise) RequestError: Internal Server Error
at Zj (k4027n9ylcc9lece.js:31:79771)
at async k8.fetch (k4027n9ylcc9lece.js:31:80071)
I am getting an “Error in message stream” issue when trying to invoke a custom connector using the Apps SDK in developer mode. It does not appear to be the same issue you are seeing but I thought I would at least mention this for you and others as the same error is occurring.
Mine was working yesterday. I hadn’t tested it today until recently and I also haven’t changed any code so mine should be working as well.
Commenting on this as well to bump this! Thought I was going crazy for a bit. Same exact issue. Was working perfectly fine 3 hours ago. Now 1/10 requests will work. Anything trying to render widgets will result in error in stream
I’m getting the same error. No prompts are working with our custom MCP. It always shows this error: ‘Error in message stream.’ Can anyone from ChatGPT assist with this problem?
I have the same issue, no code changed, worked perfectly yesterday. I am running my MCP over sse… I tried adding other MCP servers to see if it would generate the same error. Stripe worked fine but Cloudflare had the same error.
Yes same issue since morning. I saw they added oauth client id and secret optional params while adding custom connectors . and since that update it started breaking
Would be great if Open ai would add gpt apps connectivity test cases to each new deployment. Even a simple Hello World app test would suffice. Something of this sort happens multiple times per week and it’s difficult to meet tight deadlines. Demos get pushed back, etc. Frustrating. Reporting the same issue here.
