Open AI error: "InvalidRequestError: Resource not found". Please help to fix.

AIoTMR Architect 40

Dear Team

Could you please help us fixing this error? Code has been taken from sample given and key has been tried as key1 and key2 both separately. Same error is found.

python result: openai.error.InvalidRequestError: Resource not found

#Note: The openai-python library support for Azure OpenAI is in preview.
import os
import openai
openai.api_type = "azure"
openai.api_base = "https://explore123.openai.azure.com/"
openai.api_version = "2022-12-01"
#openai.api_key = os.getenv("66bcffe0ed6c440dba76940d04a7bb03")
openai.api_key = "<key>"

response = openai.Completion.create(
  engine="text-davinci-002",
  prompt="Classify the following news article into 1 of the following categories: categories: [Business, Tech, Politics, Sport, Entertainment]\n\nnews article: Donna Steffensen Is Cooking Up a New Kind of Perfection. The Internet’s most beloved cooking guru has a buzzy new book and a fresh new perspective:\n\nClassified category: Business",
  temperature=0,
  max_tokens=60,
  top_p=1,
  frequency_penalty=0,
  presence_penalty=0,
  best_of=1,
  stop=None)

print(response)

Result:

Traceback (most recent call last):
  File "test1.py", line 19, in <module>
    stop=None)
  File "/home/user/.local/lib/python3.6/site-packages/openai/api_resources/completion.py", line 31, in create
    return super().create(*args, **kwargs)
  File "/home/user/.local/lib/python3.6/site-packages/openai/api_resources/abstract/engine_api_resource.py", line 67, in create
    "post", url, params, headers, stream=stream
  File "/home/user/.local/lib/python3.6/site-packages/openai/api_requestor.py", line 129, in request
    resp = self.interpret_response(rbody, rcode, rheaders, stream=stream)
  File "/home/user/.local/lib/python3.6/site-packages/openai/api_requestor.py", line 348, in interpret_response
    return self.interpret_response_line(rbody, rcode, rheaders, stream)
  File "/home/user/.local/lib/python3.6/site-packages/openai/api_requestor.py", line 368, in interpret_response_line
    rbody, rcode, resp.data, rheaders, stream_error=stream_error
openai.error.InvalidRequestError: Resource not found

======================================================================

Curl Result:  -> no error shown but how can i see the result?

curl https://explore123.openai.azure.com/openai/deployments/text-davinci-002/completions?api-version=2022-12-01   -H "Content-Type: application/json"   -H "api-key: 66bcffe0ed6c440dba76940d04a7bb03"   -d '{
  "prompt": "Classify the following news article into 1 of the following categories: categories: [Business, Tech, Politics, Sport, Entertainment]\n\nnews article: Donna Steffensen Is Cooking Up a New Kind of Perfection. The Internet’s most beloved cooking guru has a buzzy new book and a fresh new perspective:\n\nClassified category: Business",
  "max_tokens": 60,
  "temperature": 0,
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "top_p": 1,
  "best_of": 1,
  "stop": null
}'
{"id":"cmpl-6rLFaLc8zTYo8oHbjA5WNm5Tp6i7k","object":"text_completion","created":1678171090,"model":"text-davinci-002","choices":[{"text":"","index":0,"finish_reason":"stop","logprobs":null}],"usage":{"prompt_tokens":73,"total_tokens":73}}



Regards
Champak

Ramr-msft 17,731 Reputation points

2023-03-07T14:15:09.55+00:00

@AIoTMR Architect In your code, I see that you have set the openai.api_base to "https://explore123.openai.azure.com/". Please make sure that this is the correct endpoint URL for your OpenAI account. Also, please verify that the API key you are using is valid and has the necessary permissions to access the resource.

Here is the sample for completion.

https://learn.microsoft.com/en-us/azure/cognitive-services/openai/reference
William Harding 15 Reputation points

2023-03-10T00:01:20.1266667+00:00
I'm having the same issue. New instance. InvalidRequestError: Resource not found

params = {"api_key":"xxx", "api_base":"https://exodestai.openai.azure.com/", "api_type":"azure", "api_version":"2022-12-01"} openai.api_key = params['api_key'] openai.api_base = params['api_base'] openai.api_type = params['api_type'] openai.api_version = params['api_version']

Just to be sure I checked in the portal:

and I checked it in the Tool:

And, just to be sure I put that end point into the browser:

The resource exists, it clearly has that end point, but it looks like it's still listed as 'not found'? Thoughts?
Jack 15 Reputation points

2023-03-10T09:01:20.9+00:00

Same problem with me.
Marlos Bosso 0 Reputation points Microsoft Employee

2023-03-11T14:56:57.3533333+00:00

my previous answer doesn't work for most of the cases. So I deleted it.

Please, use the endpoint shown in the Azure Portal.
William Harding 15 Reputation points

2023-03-13T19:01:18.5533333+00:00

Question about your post: is <|im_start|> part of the query syntax? Is there documentation around this? As far as I had seen the docs only take simple text strings. If there is inline text and substring operations I'd like to learn more.
Jimmy Bandell 0 Reputation points

2023-03-15T07:08:40.8033333+00:00

I have the same issiue. I cant resolve it.
Jimmy Bandell 0 Reputation points

2023-03-15T07:26:50.66+00:00

I have the same issue and I cant resolve it with the answer here :)
Eve Pardi 6 Reputation points MVP

2023-04-12T16:32:21.3866667+00:00

Same problem I face with VS Code and Python 3.9. It worked a few weeks ago but now I get the same exact issue. I made a new deployment before it failed to work! Not sure if this info helps.. I couldn't figure yet out what the problem is.

Bin Lin 0

Same issue here from VS Code and python 3.9, my workaround is using the REST call: ....

    url = base_url + "/openai/deployments/" + deployment_name + "/chat/completions?api-version=2023-03-15-preview"
    payload = {"messages":messages}

    r = requests.post(url,
        headers={
            "api-key": api_key,
            "Content-Type": "application/json"
        },
        json = payload
        )
    response = json.loads(r.text)

works to me.

Dylan K 5

I am having the same issue on an existing deployment that was previously working. I double checked every variable and they are all correct. This was working last week and is not working today.

import os
import requests
import json
import openai

openai.api_key = "XXX"
openai.api_base = "XXX"
openai.api_version = '2022-12-01'
deployment_name = 'XXX'

def get_response(start_phrase):
    print(start_phrase)    
    response = openai.Completion.create(engine=deployment_name, prompt=start_phrase, max_tokens=200)
    text = response['choices'][0]['text'].replace('\n', '').replace(' .', '.').strip()
    print(text)
    
    return response

start_phrase = 'PROMPT'
response = get_response(start_phrase)

SZETO TAI SHUN 0 Reputation points

2023-04-19T18:30:55.8733333+00:00

Same isssue here. The code runs well in my first attempt. After that, I save and reload the notebook to re-run, then come up with the issue! Of course I have checked very carefully the correctness of endpoint and key. Not feeling confident in Azure stability.
Neergaard, Cedric 0 Reputation points

2023-04-27T08:17:31.5366667+00:00
We've been running into the same problem, and tried the different solutions provided in this thread to no avail. But, we found (a weird) fix to it:

When running the chat completions from our flask server or a separate python test script, we ran into the issue with InvalidRequestError. When turning on openai.log = "debug", we saw that the api_version sent in the request is set to None, even though we knowingly set it correctly to 2022-12-01 in the code.

However, if we simply don't import the openai-package in our code (server-side and in the test script), the request goes through and works... Seems like a weird fix, but it works for some reason. If we import the openai-package and don't use it in the scripts (i.e. a redundant import), it won't work.

NOTE!

We're running openai version = 0.27.4

Downgrading to openai version = 0.27.0 doesn't work

Sending requests using cURL also works (api-version = 2022-12-01)

All the configurations have been tested to work correctly so shouldn't be anything wrong with that
Hjörtur Líndal Stefánsson 5 Reputation points

2023-05-16T14:45:48.6266667+00:00

I had the same issue (using the openai python package). Setting the api_version to "2023-03-15-preview" did the trick for me.

9 answers

Michael B 0 Microsoft Employee

I'll clean up the docs to make this more clear this week.

openai.api_base ="Your endpoint address"

This corresponds to the unique endpoint address for your Azure OpenAI resource where the gpt-35-model is deployed. You can find this by going to Keys and Endpoint for your resource.

User's image

 engine="This is the name you assigned to the deployment when you deployed the gpt-35-turbo model"

Marlos Bosso 0 Reputation points Microsoft Employee

2023-03-13T18:07:39.6266667+00:00

Great, thanks

William Harding 15

I'm assuming that its the deployment name: User's image

User's image

And the resource is this. I'm using the same variables that is in the documentation.

Just to check, I tried with the GPT-35-turbo:

User's image

Here is the full code:

import os
import requests
import json
import openai

params = {"api_key":"xxx",
          "api_base":"https://exodestai.openai.azure.com/",
          "api_type":"azure",
          "api_version":"2022-12-01"}

openai.api_key = os.getenv('OPENAI_APIKEY')
openai.api_base =  params['api_base']
openai.api_type = params['api_type']
openai.api_version = params['api_version']

deployment_name='exo-gpt-35'

# # Send a completion call to generate an answer
start_phrase = 'Write a tagline for an ice cream shop. '
response = openai.Completion.create(engine=deployment_name, prompt=start_phrase, max_tokens=10)

The end delivers the same InvalidRequestError: Resource not found . Can you help me with what I'm not seeing?

William Harding 15 Reputation points

2023-03-13T19:19:28.0366667+00:00

I'm still getting InvalidRequestError: Resource not found can you help me understand what is different about my process? What am I missing?
William Harding 15 Reputation points

2023-03-13T19:36:43.6433333+00:00

Not sure if it helps, but when I try this with:

openai.api_base = "https://exodestai.api.cognitive.microsoft.com/"

per @Marlos Bosso above, I got a different error:

APIConnectionError: Unexpected error communicating with OpenAI. If this problem persists, let us know at support@openai.com. (Network error: ConnectionError: HTTPSConnectionPool(host='exodestai.api.cognitive.microsoft.com', port=443): Max retries exceeded with url: //v1/engines/exo-gpt-35/completions (Caused by NewConnectionError(': Failed to establish a new connection: [Errno 11001] getaddrinfo failed')) (url=https://exodestai.api.cognitive.microsoft.com//v1/engines/exo-gpt-35/completions))

Could it be that there is a different API url?

William Harding 15

Just to check to see if it was my deployment, I deleted the resources and created new ones.

{
    "$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
    "contentVersion": "1.0.0.0",
    "parameters": {
        "accounts_ExoDestAI_name": {
            "type": "String"
        },
        "deployed_model_name": {
            "type": "String"
        },
        "tags": {
            "type": "object",
            "defaultValue": {}
        },
        "region": {
            "type":"string",
            "defaultValue": "West US 2"
        }
        
    },
    "variables": {},
    "resources": [
        { 
            "type": "Microsoft.CognitiveServices/accounts",
            "apiVersion": "2022-12-01",
            "name": "[parameters('accounts_ExoDestAI_name')]",
            "location": "[parameters('region')]",
            "tags": "[parameters('tags')]",
            "sku": {
                "name": "S0"
            },
            "kind": "OpenAI",
            "properties": {
                "customSubDomainName": "[parameters('accounts_ExoDestAI_name')]",
                "publicNetworkAccess": "Enabled"
            }
        },
        {
            "type": "Microsoft.CognitiveServices/accounts/deployments",
            "apiVersion": "2022-12-01",
            "name": "[concat(parameters('accounts_ExoDestAI_name'), '/', parameters('deployed_model_name'))]",
            "dependsOn": [
                "[resourceId('Microsoft.CognitiveServices/accounts', parameters('accounts_ExoDestAI_name'))]"
            ],
            "properties": {
                "model": {
                    "format": "OpenAI",
                    "name": "gpt-35-turbo",
                    "version": "0301"
                },
                "scaleSettings": {
                    "scaleType": "Standard"
                }
            }
        }
    ]
}

I confirmed that it built the resources in the same way, and got all of the keys and values right. Still no dice. I think that the domain in the docs might be wrong.

Shreshtha Agrahari 0 Reputation points

2023-04-15T05:18:24.8666667+00:00

I am getting the same error, azure openAI chat completion SDK works locally but when deployed to Azure function app gives 404- Resource not found error. How to solve this issue?
Please sign in to rate this answer.

0 comments No comments
Sign in to comment
Gianluca D'Ardia 5 Reputation points Microsoft Employee

2023-04-15T12:49:39.64+00:00
I got same issue: worked on local but not on deployed env. I solved providing right module version. Checked version on my local env with

pip list | more

In my local env I had 0.27.4 but in deployed env I got 0.10.0 changed requirements.txt in code repo (python flask app) with

openai==0.27.4

and everything worked.
Please sign in to rate this answer.
Shreshtha Agrahari 0 Reputation points

2023-04-17T13:54:06.5933333+00:00

checked the version in deployed env, it is the same. Still getting same error
Sign in to comment
Shawn Mittal 35 Reputation points Microsoft Employee

2023-04-26T15:45:12.0633333+00:00

Hi folks, looks like there seems to be an issue with the openai==0.27.4 version for python. Try installing openai==0.27.0. This fix worked for me on local. Not sure if it'll work once deployed on Azure Functions or in an App Service, but I intend to try it out.
Please sign in to rate this answer.

0 comments No comments
Sign in to comment

Share via

Open AI error: "InvalidRequestError: Resource not found". Please help to fix.

9 answers