Error: Operation returned an invalid status 'Bad Request' on client.complete

I am using below code to get data from "Mistral-Nemo" model hosted as SaaS in Azure AI Foundry. It gives me below error:

Error: Operation returned an invalid status 'Bad Request'
Traceback (most recent call last):
  File "c:\Users\guptswapmax\TestFalcon\TestFalcon.py", line 32, in <module>
        response = client.complete(
                   ~~~~~~~~~~~~~~~^
        messages=[
        ^^^^^^^^^^
    ...<5 lines>...
        model=model_name
        ^^^^^^^^^^^^^^^^
    )
    ^
  File "c:\Users\abc\TestFalcon\.venv\Lib\site-packages\azure\ai\inference\_patch.py", line 738, in complete
    raise HttpResponseError(response=response)
azure.core.exceptions.HttpResponseError: Operation returned an invalid status 'Bad Request'

I have looked into the permissions and make sure my SP has required permissions and it also has "Coginitive Support User" and "Cognitive User Contributor" role and trying to implement using Service Principal

When i use the same code with Azure API Key and AzureKeyCredential it works. https://github.com/MicrosoftDocs/azure-ai-docs/blob/main/articles/ai-foundry/model-inference/includes/use-chat-completions/csharp.md

import os
from azure.ai.inference import ChatCompletionsClient
from azure.ai.inference.models import SystemMessage, UserMessage
from azure.identity import DefaultAzureCredential, ClientSecretCredential
import logging
from azure.core.pipeline.policies import HttpLoggingPolicy

endpoint = "https://abcdomain.services.ai.azure.com/models"
model_name  = "Mistral-Nemo"

Client_Id = "xxxxxxxxxxxxxxxxxxxxxxxx"
Client_Secret = "xxxxxxxxxxxxxxxxxxx"
Tenant_Id = "xxxxxxxxxxxxxxxxxxxxxxxxx"    

cred = ClientSecretCredential(
    tenant_id=Tenant_Id,
    client_id=Client_Id,
    client_secret=Client_Secret    
)

client = ChatCompletionsClient(
        endpoint=endpoint,
        credential=cred,
        credential_scopes=["https://cognitiveservices.azure.com/.default"]
    )

try:
   result = client.complete(
    messages=[
        SystemMessage(content="You are a helpful assistant."),
        UserMessage(content="How many languages are in the world?"),
    ],
    temperature=0.8,
    top_p=0.1,
    max_tokens=2048,
    stream=True,
    model=model_name
)
except Exception as e:
    print("Error:", e)

File "c:\Users\abc\TestFalcon.venv\Lib\site-packages\azure\ai\inference_patch.py", line 738, in complete raise HttpResponseError(response=response) azure.core.exceptions.HttpResponseError: Operation returned an invalid status 'Bad Request'

The error HttpResponseError: Operation returned an invalid status 'Bad Request' usually indicates that the request payload is malformed or that there is a mismatch between the model's expectations and the API client usage.

Here is the document that can be deployed based upon their models.

In my environment I created app and added necessary RBAC role to the azure AI service.

Portal:

You can use the below code which will used to get data from your model using python.

Code:

import requests
from azure.identity import ClientSecretCredential

# Get token
tenant_id = "xxxxx"
client_id = "xxxx"
client_secret = "xxxxxx"

credential = ClientSecretCredential(tenant_id, client_id, client_secret)
token = credential.get_token("https://cognitiveservices.azure.com/.default").token

endpoint = "https://xxxxxxx.services.ai.azure.com"
api_version = "2024-12-01-preview"
model = "xxxxxx"

url = f"{endpoint}/openai/deployments/{model}/chat/completions?api-version={api_version}"

headers = {
    "Authorization": f"Bearer {token}",
    "Content-Type": "application/json"
}

data = {
    "messages": [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "How many languages are in the world?"}
    ],
    "temperature":0.8,
    "top_p":0.1,
    "max_tokens":2048,
    "stream":True
}


response = requests.post(url, headers=headers, json=data)
response_json = response.json()
content = response_json["choices"][0]["message"]["content"]
print(content)

Output:

Estimates vary depending on how you count distinct “languages” versus dialects, but the most frequently cited figure is that there are roughly 7,000 living human languages in the world today.  Key points:  
1. The Ethnologue, a widely used catalog published by SIL International, lists about 7,100 living languages (7,151 in its 2023 edition).  
2. UNESCO and other bodies often give a rounded figure of “6,000–7,000,” reflecting different classification criteria.  
3. Nearly 40 percent of those languages are considered endangered, meaning they risk falling out of use as their speakers shift to more dominant tongues.  
4. In addition to the living languages, several thousand more have gone extinct or survive only in written records.  
So while there’s no single “official” count, it’s safe to say the number of living languages worldwide hovers around seven thousand.

Reference:

azure-sdk-for-python/sdk/ai/azure-ai-inference/samples at azure-ai-inference_1.0.0b9 · Azure/azure-sdk-for-python