Keep getting this error. Here’s my JSON. Can someone help?
{
"schema_version": "v1",
"name_for_human": "Movies API (no auth)",
"name_for_model": "movies",
"description_for_human": "API for retrieving relevant movies or series and their poster image URLs from the streaming service Tubi.",
"description_for_model": "API for retrieving relevant movies or series and their poster image URLs from the streaming service Tubi.",
"auth": {
"type": "none"
},
"api": {
"type": "openapi",
"url": "https://0539-216-250-210-0.ngrok-free.app/openapi.yaml",
"is_user_authenticated": false
},
"logo_url": "https://0539-216-250-210-0.ngrok-free.app/logo.png",
"contact_email": "support@example.com",
"legal_info_url": "http://www.example.com/legal"
}
Counting the commas and curly braces - they seem to be ok.
Another cause could be a hidden character at the beginning of the file, a common example is the BOM (byte order mark), which can be added by some editors.
I guess this error is due to ngrok, I have tried localhost, then using another production server. All are not giving any error.
So there is bug with ngrok
I ran the code above thorough a JSON validator and it looks fine. My guess there might be the response from server delivering the file is missing the correct Content-Type header or similar. Another though is to check that OpenAI reaches your file at all and is not fetching a 404 not found page.
Same here, my json is also correct, using the ngrok website externally to check if the ai-plugin.json is visible or not. It was visible but this error is still occuring.
I have the same issue. Could it be that when OpenAI fetches the ai-plugin.json file through ngrok, the ngrok browser warning site is shown to ChatGPT instead of the manifest file? I think this would only be fixable if OpenAI adds the ‘ngrok-skip-browser-warning’ Header with any value in the call.
Related on StackOverflow: questions/73017353/how-to-bypass-ngrok-browser-warning (I can’t post links)