Azure OpenAI function call error: Missing parameter 'name': messages with role 'function' must have a 'name'.

Question

Azure OpenAI function call error: Missing parameter 'name': messages with role 'function' must have a 'name'.

Ulysses Wu 10

The following request works on OpenAI API, but fails on Azure OpenAI with this error:

data: {"code":500,"message":"stream error: (error, status code: 400, message: Missing parameter 'name': messages with role 'function' must have a 'name'.)","model":"","choices":null}

    {        
        "model": "gpt-4-0613",
        "messages": [
            {
                "role": "system",
                "content": ""
            },
            {
                "role": "user",
                "content": "What is the weather like in Boston?"
            },
            {
                "role": "assistant",
                "content": "What is the weather like in Boston?",
                "function_call":{"name":"get_current_weather","arguments":"{\n  \u0022location\u0022: \u0022Boston, MA\u0022\n}"}
            },
            {
                "name": "get_current_weather",
                "role": "function",
                "content": "20 C"
            }
        ],
        "temperature": 0.001,
        "top_p": 1,
        "n": 1,
        "stream": true,
        "max_tokens": null,
        "presence_penalty": 0,
        "frequency_penalty": 0,
        "functions": [
     {
       "name": "get_current_weather",
       "description": "Get the current weather in a given location",
       "parameters": {
         "type": "object",
         "properties": {
           "location": {
             "type": "string",
             "description": "The city and state, e.g. San Francisco, CA"
           },
           "unit": {
             "type": "string",
             "enum": ["celsius", "fahrenheit"]
           }
         },
         "required": ["location"]
       }
     }
   ],
        "function_call": "auto"
    }

Saurabh Sharma 23,846 Reputation points Microsoft Employee Moderator

2023-09-12T21:39:54.7266667+00:00

Hi @Ulysses Wu ,

Thanks for using Microsoft Q&A!!

I believe you are getting this error as you are explicitly stating "function_call" property as "auto" in the request. As mentioned in the documentation over here, if you have already added "functions" property then by default the function_call property will be set to "auto". In case you are adding function_call parameter then please specify the specific function name to force the api to call that function else use "none" to prevent the model to call any functions.

Please let me know if you see any other issues.

Thanks

Saurabh
Ulysses Wu 10 Reputation points

2023-09-13T03:48:18.8333333+00:00

Thanks for reply, but I'm getting exactly the same error without "function_call": "auto" .

However, the OpenAI API always work as expected.
Saurabh Sharma 23,846 Reputation points Microsoft Employee Moderator

2023-09-13T20:40:19.8233333+00:00
Hi Ulysses Wu,

Can you please provide your code which you are trying to use along with the function_call so that I can further investigate. I have tried this using the functions sample and it worked fine without any errors.

Here is how exactly the code works -

If you use auto against function_call in ChatCompletion.Create code, then the model decides what function to call.

response = openai.ChatCompletion.create( engine="gpt-35-turbo-0613", messages=messages, functions=functions, function_call="auto" ) print(response["choices"][0]['message'])

If you use none in the create call then, the model doesn't call any function

response = openai.ChatCompletion.create( engine="gpt-35-turbo-0613", messages=messages, functions=functions, function_call="none" ) print(response["choices"][0]['message'])

If you want to force model to call any specific function then you need to provide function name in the function_call by passing the name parameter like below -

response = openai.ChatCompletion.create( engine="gpt-35-turbo-0613", messages=messages, functions=functions, function_call={"name": "get_current_weather"} ) print(response["choices"][0]['message'])

Result:

Thanks

Saurabh
Ulysses Wu 10 Reputation points

2023-09-14T06:49:59.46+00:00

[duplicated post, removed]

Ulysses Wu 10

The problem happens when the last massage is a "function" message (I have attached the request in my first post but I didn't see you try the similar step... I hope I was not getting replied by GPT). That is:

You asked something with a function list - fine

GPT replied with a function_call request - fine (you are here)

You replied with a function result

GPT should reply based on your function result, but no

import os
import openai

#openai.api_key = ""
openai.api_version = "2023-07-01-preview"
openai.api_type = "azure"
#openai.api_base = ""


messages= [
    {"role": "user", "content": "Find beachfront hotels in San Diego for less than $300 a month with free breakfast."},
#     {
#   "role": "assistant",
#   "content": None,
#   "function_call": {
#     "name": "search_hotels",
#     "arguments": "{\n  \"features\": \"beachfront, free breakfast\",\n  \"location\": \"San Diego\",\n  \"max_price\": 300\n}"
#         }  
#     },
#     {"role": "function", "content": "Hotel California", "name": "search_hotels"}
]

functions= [  
    {
        "name": "search_hotels",
        "description": "Retrieves hotels from the search index based on the parameters provided",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {
                    "type": "string",
                    "description": "The location of the hotel (i.e. Seattle, WA)"
                },
                "max_price": {
                    "type": "number",
                    "description": "The maximum price for the hotel"
                },
                "features": {
                    "type": "string",
                    "description": "A comma separated list of features (i.e. beachfront, free wifi, etc.)"
                }
            },
            "required": ["location"]
        }
    }
]  

response = openai.ChatCompletion.create(
    engine="gpt-4",
    messages=messages,
    functions=functions,
    function_call="auto", 
)

response_message = response['choices'][0]['message']
# print(response_message)

# Check if the model wants to call a function
if response_message.get("function_call"):

    # Call the function. The JSON response may not always be valid so make sure to handle errors
    function_name = response_message["function_call"]["name"]    
    function_response = "Hotel California"

    # Add the assistant response and function response to the messages
    messages.append( # adding assistant response to messages
        {
            "role": response_message["role"],
            "function_call": {
                "name": function_name,
                "arguments": response_message["function_call"]["arguments"],
            },
            "content": None
        }
    )
    messages.append( # adding function response to messages
        {
            "role": "function",
            "name": function_name,
            "content": function_response,
        }
    ) 

    # Call the API again to get the final response from the model
    second_response = openai.ChatCompletion.create(
            messages=messages,
            engine="gpt-4" 
            # optionally, you could provide functions in the second call as well
        )
    
    print(second_response["choices"][0]["message"])

expect: (this is how OpenAI replies)

{
  "role": "assistant",
  "content": "I found a hotel that matches your criteria:\n\nHotel: Hotel California\nLocation: San Diego\nFeatures: Beachfront, Free breakfast\nPrice: Under $300 per month\n\nPlease note that prices are subject to change and it's always a good idea to check the current rate before making a reservation."
}

actual: (this is how Azure OpenAI replies)

data: {"code":500,"message":"stream error: (error, status code: 400, message: Missing parameter 'name': messages with role 'function' must have a 'name'.)","model":"","choices":null}

Saurabh Sharma 23,846 Reputation points Microsoft Employee Moderator

2023-09-20T22:22:17.9133333+00:00

@Ulysses Wu Thanks for sharing. I am looking into this and update you back at the earliest.

Thanks

Saurabh

1 answer

Your answer

Saurabh Sharma 23,846 Reputation points Microsoft Employee Moderator

2023-09-12T21:39:54.7266667+00:00

Hi @Ulysses Wu ,

Thanks for using Microsoft Q&A!!

I believe you are getting this error as you are explicitly stating "function_call" property as "auto" in the request. As mentioned in the documentation over here, if you have already added "functions" property then by default the function_call property will be set to "auto". In case you are adding function_call parameter then please specify the specific function name to force the api to call that function else use "none" to prevent the model to call any functions.

Please let me know if you see any other issues.

Thanks

Saurabh
Ulysses Wu 10 Reputation points

2023-09-13T03:48:18.8333333+00:00

Thanks for reply, but I'm getting exactly the same error without "function_call": "auto" .

However, the OpenAI API always work as expected.
Saurabh Sharma 23,846 Reputation points Microsoft Employee Moderator

2023-09-13T20:40:19.8233333+00:00

Hi Ulysses Wu,

Can you please provide your code which you are trying to use along with the function_call so that I can further investigate. I have tried this using the functions sample and it worked fine without any errors.

Here is how exactly the code works -

If you use auto against function_call in ChatCompletion.Create code, then the model decides what function to call.

response = openai.ChatCompletion.create( engine="gpt-35-turbo-0613", messages=messages, functions=functions, function_call="auto" ) print(response["choices"][0]['message'])

If you use none in the create call then, the model doesn't call any function

response = openai.ChatCompletion.create( engine="gpt-35-turbo-0613", messages=messages, functions=functions, function_call="none" ) print(response["choices"][0]['message'])

If you want to force model to call any specific function then you need to provide function name in the function_call by passing the name parameter like below -

response = openai.ChatCompletion.create( engine="gpt-35-turbo-0613", messages=messages, functions=functions, function_call={"name": "get_current_weather"} ) print(response["choices"][0]['message'])

Result:

Thanks

Saurabh
Ulysses Wu 10 Reputation points

2023-09-14T06:49:59.46+00:00

[duplicated post, removed]
Saurabh Sharma 23,846 Reputation points Microsoft Employee Moderator

2023-09-20T22:22:17.9133333+00:00

@Ulysses Wu Thanks for sharing. I am looking into this and update you back at the earliest.

Thanks

Saurabh

Answer 1

Saurabh Sharma 23,846 Microsoft Employee Moderator

@Ulysses Wu

I am getting results as you are expecting from Azure Open AI. I have tried this with GPT-35-Turbo and GPT-4 models. (See screenshots below)

User's image

If you are still getting this error then, I suggest you to please open an Azure Support ticket so that an engineer could look into your environment. Please let me know in case you have any limitations opening a support case.

Thank you,

Saurabh

Share via

Azure OpenAI function call error: Missing parameter 'name': messages with role 'function' must have a 'name'.

1 answer

Your answer