Hello. I have deployed a model in Azure AI, but when trying to use it in cursor, I’ve got an error no_model_name.
I am sure the endpoint is the OpenAI endpoint, the name is a deployment name not a model name, and the API key is correct.
There are some related issues like Python: azure ai service model invoke fails with no_model_name error code in sk python · Issue #10339 · microsoft/semantic-kernel. Is there any insight why it is happening?
Thanks!
For Azure, I believe you need an Azure-specifc end-point, provided to you by Azure in your deployment instance.
The link below might help: How to switch between OpenAI and Azure OpenAI Service endpoints with Python - Azure OpenAI Service | Microsoft Learn
Thanks, but I am using the OpenAI endpoint provided by Azure AI.
.It matches the format of the placeholder value in the base URL input.
Also, the error is a meaningful business error no_model_error, which I believe should mean the URL is right but parameters are not.
Thanks for confirming.
When I connected Azure, I remember is being a massive pain because they gave me three different API keys and a different endpoint URLs.
Just to ensure the details are correct, could you confirm your connection is working outside of Cursor, with the same API details?
You should be able to get an answer running the juypter notebook below (just ensure you replace variables or have them set in your .env file):
import os
from openai import AzureOpenAI
from typing import Dict, Union
from dotenv import load_dotenv
# Load environment variables
load_dotenv(override=True)
endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")
deployment = os.getenv("AZURE_OPENAI_DEPLOYMENT_ID", "gpt-4o")
subscription_key = os.getenv("AZURE_OPENAI_API_KEY")
api_version=os.getenv("AZURE_OPENAI_API_VERSION", "2024-05-01-preview")
def test_azure_openai(prompt: str = "What is the capital of France?") -> Dict[str, Union[str, None]]:
"""
Test Azure OpenAI connectivity using a deployed model.
:param prompt: Input question to test the model
:type prompt: str
:return: Dictionary containing response or error message
:rtype: Dict[str, Union[str, None]]
"""
try:
client = AzureOpenAI(
api_key=subscription_key,
api_version=api_version,
azure_endpoint=endpoint,
)
# Get the exact deployment name
deployment_name = os.getenv("AZURE_OPENAI_DEPLOYMENT_ID", "gpt-4o") # Ensure this is correct
print(deployment_name)
# Print the inputs
print(f"Deployment Name: {deployment_name}")
print(f"Prompt: {prompt}")
response = client.chat.completions.create(
model=deployment_name, # Use the exact deployment name
messages=[{"role": "user", "content": prompt}],
temperature=0.7,
max_tokens=100,
)
return {"response": response.choices[0].message.content, "error": None}
except Exception as e:
return {"response": None, "error": str(e)}
# Run it
if __name__ == "__main__":
result = test_azure_openai()
if result["error"]:
print(f"Error: {result['error']}")
else:
print(f"Response: {result['response']}")
Note in this case I assumed the deployment_name was gpt-4o.
Thanks for the code snippets! I’ve made some attempts on this code, and it seems to be an issue on Azure side.
The original model I was trying to run with was DeepSeek-R1, which when run with this script, returned an error 400 - {'error': {'code': 'unknown_model', 'message': 'Unknown model: ds-r1', 'details': 'Unknown model: ds-r1'}}. It seems to be a similar error cursor got.
I tried the script with anPhi-4 model deployment but the error became 500 - {'error': {'code': 'InternalServerError', 'message': 'Backend returned unexpected response. Please contact Microsoft for help.'}}. It was clearly an Azure issue.
It was finally working with an o1-mini deployment, but the temperature and max_token parameters must be removed for this deployment.
It seems not all models in Azure AI are working fine with cursor.
Thanks again for this script!