# Consistent internal server error (status 500) responses for completions using functions

**URL:** <https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314>\
**Category:** Bugs\
**Created:** [July 19, 2023, 6:12pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314 "2023-07-19T18:12:23Z")\
**Posts on this page:** 15\
**Page:** 1

<div class="post-metadata">

**Author:** ![Mathyn](https://avatars.discourse-cdn.com/v4/letter/m/ce7236/32.png) [@Mathyn](https://community.openai.com/u/Mathyn)\
**Post date:** [July 19, 2023, 6:12pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/1 "2023-07-19T18:12:23Z")

</div>

For some reason I am getting consistent internal server errors with the request shown below. I do not know what I am doing wrong. Anyone got any idea?

**What I try to do**

I do a POST call to to the V1 completions endpoint

I give the following POST body

```auto
{
    "model": "gpt-3.5-turbo-16k-0613",
    "messages": [
        {
            "role": "system",
            "content": "You are an assistant called Maia. You are always open for a pleasant conversation. You love using emojies."
        },
        {
            "role": "user",
            "content": "I could actually use a bit of help understanding this Wikipedia page [LINK REDACTED] specifically I want to know more about public works and such he did. Can you help?",
            "name": "John"
        },
        {
            "role": "assistant",
            "content": "During his reign, Al-Walid I financed impressive public works, including:\n\n🕌 The construction of the Great Mosque of Damascus\n🕌 The construction of the al-Aqsa Mosque in Jerusalem\n🕌 The construction of the Prophet's Mosque in Medina\n\nHe also built way-stations and dug wells along the roads in Syria, installed street lighting in the cities, and invested in land reclamation projects. Additionally, he built numerous palaces, fortresses, and other buildings throughout the Umayyad Empire.",
            "name": "Maia"
        },
        {
            "role": "user",
            "content": "Ah thank you!",
            "name": "John"
        }
    ],
    "functions": [
        {
            "name": "QueryWebPage",
            "description": "Queries the giving web page for the requested information.",
            "parameters": {
                "type": "object",
                "properties": {
                    "url": {
                        "type": "string"
                    },
                    "question": {
                        "type": "string",
                        "description": "A human readable sentence describing what information you would like to retrieve from the web page."
                    }
                },
                "required": [
                    "url",
                    "question"
                ]
            }
        },
        {
            "name": "DoWebSearch",
            "description": "Performs a web search and returns a list of URLs fitting the search query.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {
                        "type": "string",
                        "description": "The search query."
                    }
                },
                "required": [
                    "query"
                ]
            }
        }
    ],
    "temperature": 0.5
}

```

**What happens**

I get a 500 internal server error with the following message: “The server had an error processing your request. Sorry about that! You can retry your request, or contact us through our help center at [help.openai.com](http://help.openai.com) if you keep seeing this error.”

**What I would expect**

The request would not fail and I would retrieve a completion response back.

**Observations**

1. Try and remove the last message and the call does not fail
2. Try and remove the functions and the call does not fail

---

<div class="post-metadata">

**Author:** ![Foxalabs](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/foxalabs/32/738355_2.png) [@Foxalabs](https://community.openai.com/u/Foxalabs)\
**Post date:** [July 19, 2023, 6:33pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/2 "2023-07-19T18:33:30Z")

</div>

Welcome to the forum!

Are you calling `/v1/completions` or `/v1/chat/completions`? as to use the model in your post, you would need `/v1/chat/completions`

---

<div class="post-metadata">

**Author:** ![Mathyn](https://avatars.discourse-cdn.com/v4/letter/m/ce7236/32.png) [@Mathyn](https://community.openai.com/u/Mathyn)\
**Post date:** [July 19, 2023, 7:08pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/3 "2023-07-19T19:08:50Z")

</div>

Hi and thank you for your reply!

I am calling `/v1/chat/completions`.

Just to clarify: by making small tweaks to the POST body (like removing the last message) the endpoint does work and I do get a valid response.

---

<div class="post-metadata">

**Author:** ![Foxalabs](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/foxalabs/32/738355_2.png) [@Foxalabs](https://community.openai.com/u/Foxalabs)\
**Post date:** [July 19, 2023, 7:14pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/4 "2023-07-19T19:14:20Z")

</div>

So if the message body has

```auto
        {
            "role": "user",
            "content": "Ah thank you!",
            "name": "John"
        }

```

in it it 500’s and if not, it’s ok?

---

<div class="post-metadata">

**Author:** ![Mathyn](https://avatars.discourse-cdn.com/v4/letter/m/ce7236/32.png) [@Mathyn](https://community.openai.com/u/Mathyn)\
**Post date:** [July 19, 2023, 7:57pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/5 "2023-07-19T19:57:08Z")

</div>

Yes, correct. I just checked again and consistently with this last message I get an error and without this message things work fine.

However removing the functions from the POST body also ‘fixes’ the issue.

---

<div class="post-metadata">

**Author:** ![Foxalabs](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/foxalabs/32/738355_2.png) [@Foxalabs](https://community.openai.com/u/Foxalabs)\
**Post date:** [July 19, 2023, 8:01pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/6 "2023-07-19T20:01:59Z")

</div>

Interesting, I think I’d be tempted to recreate the body and function call with the OpenAI python library and see if that works, and if it does… then I think I’d take a look at the message that got sent out, intercept the socket and log it, something like that and run a Diff engine over the two sets of text.

If that ALSO fails, then I think it’s bug report time.

---

<div class="post-metadata">

**Author:** ![Mathyn](https://avatars.discourse-cdn.com/v4/letter/m/ce7236/32.png) [@Mathyn](https://community.openai.com/u/Mathyn)\
**Post date:** [July 19, 2023, 8:18pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/7 "2023-07-19T20:18:26Z")

</div>

I so far only verified this using Postman so I can communicate directly with the API and eliminate any libraries I use from messing with the request.

Anything I can do to help? I’m afraid your plan of attack is a bit out of my own area of expertise.

---

<div class="post-metadata">

**Author:** ![Foxalabs](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/foxalabs/32/738355_2.png) [@Foxalabs](https://community.openai.com/u/Foxalabs)\
**Post date:** [July 19, 2023, 8:33pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/8 "2023-07-19T20:33:25Z")

</div>

Ahh, understood, I’ve not gone through the function calling post syntax myself yet, so I don’t know if there are any discrepancies in the post.

There is a walk through and examples here of using the python library to make the API calls (see below) What I’m trying to narrow down is, if it’s a bug in the API, which I doubt, or if it’s a bug in the post syntax.

> <https://github.com/openai/openai-cookbook/blob/main/examples/How_to_call_functions_with_chat_models.ipynb>

---

<div class="post-metadata">

**Author:** ![supershaneski](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/supershaneski/32/71779_2.png) [@supershaneski](https://community.openai.com/u/supershaneski)\
**Post date:** [July 20, 2023, 12:15am UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/9 "2023-07-20T00:15:40Z")

</div>

I ran your code and it causes error like you observed.

But when I removed **name** property in message entries, it work as expected.

You probably do not want to remove name property. My workaround is to include it in content.

```javascript
    { role: 'system', content: 'You are an assistant called Maia. You are always open for a pleasant conversation. You love using emojies.' },
            { role: 'user', content: 'John: I could actually use a bit of help understanding this Wikipedia page https://en.wikipedia.org/wiki/Al-Walid_I specifically I want to know more about public works and such he did. Can you help?' },
            { role: 'assistant', content: "Maia: During his reign, Al-Walid I financed impressive public works, including:\n\n🕌 The construction of the Great Mosque of Damascus\n🕌 The construction of the al-Aqsa Mosque in Jerusalem\n🕌 The construction of the Prophet's Mosque in Medina\n\nHe also built way-stations and dug wells along the roads in Syria, installed street lighting in the cities, and invested in land reclamation projects. Additionally, he built numerous palaces, fortresses, and other buildings throughout the Umayyad Empire." },
            { role: 'user', content: 'John: Ah thank you!' }

```

and got response as

```javascript
{
  role: 'assistant',
  content: "Maia: You're welcome, John! If you have any more questions or need further assistance, feel free to ask. 😊"
}

```

---

<div class="post-metadata">

**Author:** ![pabloromeo](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/pabloromeo/32/159814_2.png) [@pabloromeo](https://community.openai.com/u/pabloromeo)\
**Post date:** [July 20, 2023, 12:52am UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/10 "2023-07-20T00:52:48Z")

</div>

I hit this same bug time ago, but also found nowhere to report it.  
It’s due to using the “name” property. I was using it in the same way, for multiuser chat, and it consistently failed as soon as functions came into play.  
Unfortunately I only found a workaround, which is to remove the name property and include user’s name within the content instead.

---

<div class="post-metadata">

**Author:** ![\_j](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/_j/32/766292_2.png) [@\_j](https://community.openai.com/u/_j)\
**Post date:** [July 20, 2023, 3:02am UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/11 "2023-07-20T03:02:52Z")

</div>

I can get both custom name and function to go.

```auto
 --- POST ---
b'{"model": "gpt-3.5-turbo-0613", "max_tokens": 100, "temperature": 0.3, "messages": [{"role": "system", "content": "You are an assistant called Maia.\\nYou are always open for a pleasant conversation.\\nYou love using emoji."}, {"role": "user", "name": "John", "content": "Banana color?"}, {"role": "assistant", "name": "Maya", "content": "Yellow."}, {"role": "user", "name": "John", "content": "One word that rhymes with tear?"}, {"role": "assistant", "name": "Maya", "content": " How about \\"Bare\\". Or maybe \\"near\\"."}, {"role": "user", "name": "John", "content": "What are the top news stories of 2023?"}, {"role": "function", "name": "QueryWebPage", "content": "No additional info"}], "functions": [{"name": "QueryWebPage", "description": "Queries the giving web page for the requested information.", "parameters": {"type": "object", "properties": {"url": {"type": "string"}, "question": {"type": "string", "description": "A human readable sentence describing what information you would like to retrieve from the web page."}}, "required": ["url", "question"]}}, {"name": "DoWebSearch", "description": "Performs a web search and returns a list of URLs fitting the search query.", "parameters": {"type": "object", "properties": {"query": {"type": "string", "description": "The search query."}}, "required": ["query"]}}], "function_call": "auto"}'
 - end POST - 
api_requestor.py status
<Response [200]>
-----done---------

```

and with the function-calling intact as described:

```auto
  "model": "gpt-3.5-turbo-0613",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": null,
        "function_call": {
          "name": "QueryWebPage",
          "arguments": "{\n \"url\": \"https://www.google.com/search?q=top+news+stories+of+2023\",\n \"question\": \"Top news stories of 2023\"\n}"
        }
      },
      "finish_reason": "function_call"
    }
  ],

```

I spent a good deal of time making a very expository template that is better than openai examples of function calling. I plan to up the quality more for my own tutorial in a bit, but here is the code replicating that in the first post.

```auto

import openai

openai.api_key = "sk-xxx" # mine

# Names must not have spaces
user_name="John"
ai_name="Maya"

functions_are_included=True
fake_a_function_return=True
my_ai_model="gpt-3.5-turbo-0613"
prompt="""What are the top news stories of 2023?"""

allmessages=[
    {
    "role": "system",
    "content": """You are an assistant called Maia.
You are always open for a pleasant conversation.
You love using emoji."""
    },
    {
    "role": "user",
    "name": user_name,
    "content": """Banana color?"""
    },
    {
    "role": "assistant",
    "name": ai_name,
    "content": """Yellow."""
    },
    {
    "role": "user",
    "name": user_name,
    "content": """One word that rhymes with tear?"""
    },
    {
    "role": "assistant",
    "name": ai_name,
    "content": """ How about "Bare". Or maybe "near"."""
    },
    {
    "role": "user",
    "name": user_name,
    "content": prompt,
    },
]

fake_function_return=[
    {
    "role": "function",
    "name": "QueryWebPage",
    "content": "No additional info"
    }
]
if fake_a_function_return:
    allmessages.extend(fake_function_return)

allfunctions=[
{
    "name": "QueryWebPage",
    "description": "Queries the giving web page for the requested information.",
    "parameters": {
        "type": "object",
        "properties": {
            "url": {
                "type": "string"
            },
            "question": {
                "type": "string",
                "description": "A human readable sentence describing what information you would like to retrieve from the web page."
            }
        },
        "required": [
            "url",
            "question"
        ]
    }
},
{
    "name": "DoWebSearch",
    "description": "Performs a web search and returns a list of URLs fitting the search query.",
    "parameters": {
        "type": "object",
        "properties": {
            "query": {
                "type": "string",
                "description": "The search query."
            }
        },
        "required": [
            "query"
        ]
    }
}
]

def api_playground(all_messages,
                   all_functions,
                   functions_enabled=False,
                   model="gpt-3.5-turbo",
                   max_tokens=100,
                   temperature=0.3
                   ):
    
    api_params = {
    "model": model,
    "max_tokens": max_tokens,
    "temperature": temperature,
    "messages": all_messages
    }
    if functions_enabled:
        api_params["functions"] = all_functions
        api_params["function_call"] = "auto"

    try:
        api_response = openai.ChatCompletion.create(**api_params)
        return api_response
    # many more openai-specific error handlers can go here
    except Exception as err:
        error_message = f"API Error: {str(err)}"
        return error_message # not the right way

api_out = api_playground(allmessages, allfunctions, functions_are_included, my_ai_model)
print('--API Playground--')
print(api_out)
# print(api_out["choices"][0]["message"])
print('---------------')

```

And sure, tell me what I did wrong…

---

<div class="post-metadata">

**Author:** ![Mathyn](https://avatars.discourse-cdn.com/v4/letter/m/ce7236/32.png) [@Mathyn](https://community.openai.com/u/Mathyn)\
**Post date:** [July 20, 2023, 6:45am UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/14 "2023-07-20T06:45:54Z")

</div>

I’ll do a double check of the official functions documentation just to make sure I did not make a mistake @Foxalabs

However even if I made a mistake in the function syntax the API should probably return a better error code (for example a 400 bad request, if I make an obvious mistake in the function body it does return this).

Also looking at the other replies (like supershaneski and pabloromeo mentioned) it seems to be related to the `name` property. I can confirm if I remove this property everything works fine.

I don’t want to jump to conclusions but this does sound an awful lot like a bug. Unless there is some documentation somewhere which explicitly states you cannot use the `name` property in combination with functions.

Finally @_j I’ll take a look at your example and see if I can use it, thanks!

---

<div class="post-metadata">

**Author:** ![Foxalabs](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/foxalabs/32/738355_2.png) [@Foxalabs](https://community.openai.com/u/Foxalabs)\
**Post date:** [July 20, 2023, 7:13am UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/15 "2023-07-20T07:13:54Z")

</div>

I think it’s worth a post in the [Bugs](https://community.openai.com/c/api/bugs/30) category

---

<div class="post-metadata">

**Author:** ![pabloromeo](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/pabloromeo/32/159814_2.png) [@pabloromeo](https://community.openai.com/u/pabloromeo)\
**Post date:** [December 29, 2023, 1:41pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/16 "2023-12-29T13:41:34Z")

</div>

It appears that OpenAI did something about this, since it no longer causes a 500 internal error.  
HOWEVER, it still causes problems. What I’m seeing now, is that using “name” for conversation participants does function correctly, but it confuses function calling. Every once in a while, the model will try to call a function with the name of a User, a function that’s completely hallucinated.  
It appears that OpenAI’s training on that “name” property makes it think it should only be used for function names. Quite unfortunate. Is anybody else seeing that same behavior?

---

<div class="post-metadata">

**Author:** ![ronabarrett](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/ronabarrett/32/265685_2.png) [@ronabarrett](https://community.openai.com/u/ronabarrett)\
**Post date:** [December 31, 2023, 5:40pm UTC](https://community.openai.com/t/consistent-internal-server-error-status-500-responses-for-completions-using-functions/304314/17 "2023-12-31T17:40:24Z")

</div>

I’m continually hitting 500 internal errors as I run a autogen groupchat with a memgpt agent.