Unknown error occurred during API call with chatgpt plugin developer tool

My local plugin was able to be installed but none of the API calls are success.

I tested my API on railway deployed url and that works, so I assume this only happens to my local API? Honestly right now it doesn’t have a great developer experience as it seems like I need to tweak my environment a bit for the local plugin installation to work and I haven’t yet figured out why the request failed (with no errors or clues)

2 Likes

Same question for localhost: Example Doesn't Seem To Work?

I haven’t tried railway, but I tried AWS EC2 with nginx => python server and that couldn’t fetch the ai-plugin.json file given the IP address (I could access with CURL or from the browser) http://18.205.244.229/.well-known/ai-plugin.json. I am putting “18.205.244.229”

and it can’t find the manifest file, I don’t even get a request to nginx (access log empty)

1 Like

Hey folks, sorry for the delay here, looks like there is a bug preventing local host dev right now, please stay tuned!

4 Likes

Thank you! Can you comment on my problem with EC2 as well? Also can the team release some barebones developer tools or just much better errors?

Oh now I see in the console “Only HTTPS schemes are supported:” but that is not the error you get in the UI which is confusing

1 Like

Ah, I should have checked this thread a couple of hours ago. Was not having any luck at all trying to find manifest on my local host dev today. Hopefully this will be sorted soon.

I’ve tried many methods, including hosting on Azure and GoDaddy, but it doesn’t recognize the well-known folder, and I can’t access the manifest inside. Does anyone know the best practice for this or have any recommendations for hosting providers?

1 Like

http is not supported, see the javascript console which took me too long to figure out. I’m not affiliated with vercel, but I did use them and created a template repo: GitHub - kennethgoodman/gpt4_plugin_test that you can use to just deploy with the vercel command

2 Likes

Thank you so much! What it the plugin_hostname exact?

1 Like

That gets updated in the api.py file based on the server that is running the code

1 Like

I am almost there :slight_smile:

1 Like

Looks like this is still not working for localhost. Going to try deploying it on Vercel.

1 Like

From Vercel, not getting an SSL error, but still not working.


Getting this error now:
I error https://chat.openai.com/backend-api/aip/p Error creating plugin: {“message”:“Internal server error from https://chatgpt-plugin1.vercel.app/.well-known/ai-plugin.json”}
(

use 127.0.0.1 instead of localhost

1 Like

Thanks, I tried that. It still can’t find ai-plugin.json. I also tried several different ports.

@arevolutionofone can you share the Python code you are working with? Check out this repo: GitHub - openai/plugins-quickstart: Get a ChatGPT Plugin up and running in under 5 minutes! it has the code to get started in a few minutes or less.

1 Like

the issue is on line 46, you nee to have the correct path to that file, which is not right at the moment.

1 Like

is the http issue fixed? I still can’t use http on localhost (or 127.0.0.1, 0.0.0.0) but can use my remote https url.

1 Like

Thx. I changed the path in line 46 to this:

FileNotFoundError [Errno 2] No such file or directory: '/.well-known/ai-plugin.json'
File /Users/larry/Dev/OpenAI/chatgpt-plugin1/index.py, line 45, in
43	@app.get("/.well-known/ai-plugin.json")
44	async def plugin_manifest():
45	host = request.headers["Host"]
46	with open("/.well-known/ai-plugin.json") as f:
47	text = f.read()
48	# This is a trick we do to populate the PLUGIN_HOSTNAME constant in the manifest
49	text = text.replace("PLUGIN_HOSTNAME", f"http://{host}")
50	return quart.Response(text, mimetype="text/json")
host	127.0.0.1:8000

I’m still getting the same error.

Ok, I got plugins-quickstart running, thanks!