Functions: Arguments are absent when called via ChatCompletion.acreate

[valaises]

Describe the bug

Hi! If you create a chat using acreate, provide functions, up-to-date model, the arguments field is always empty regardless, reproduces every single time. Altho, if you are using simple synchronous create on the same setup, it works perfectly.

To Reproduce

ChatCompletion.acreate

gen = await openai.ChatCompletion.acreate(
    model='gpt-3.5-turbo-0613',
    messages=[{"role": "user", "content": "What's the weather like in Boston?"}],
    stream=True,
    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'
)

  resp = gen.__anext__()
  delta = resp.choices[0].delta
  print(f'DELTA:\n{delta}')

OUTPUT (ARGUMENTS ARE ABSENT)

DELTA:
{
  "role": "assistant",
  "content": null,
  "function_call": {
    "name": "get_current_weather",
    "arguments": ""
  }
}

ChatCompletion.create

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo-0613",
    messages=[{"role": "user", "content": "What's the weather like in Boston?"}],
    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",
)

message = response["choices"][0]["message"]
print(message)

OUTPUT (ARGUMENTS ARE PRESENT)

{
  "role": "assistant",
  "content": null,
  "function_call": {
    "name": "get_current_weather",
    "arguments": "{\n  \"location\": \"Boston\"\n}"
  }
}

Code snippets

No response

OS

Ubuntu 20.04.5 LTS

Python version

Python 3.9.15

Library version

openai-python v0.27.8

[n3d1117]

Hey @valaises, just chiming in since I also bumped into this a few days ago. When using stream=True you need to wait for all the chunks to be sent and combine all the deltas. This works for me:

gen = await openai.ChatCompletion.acreate(
  model='gpt-3.5-turbo-0613',
  messages=[{"role": "user", "content": "What's the weather like in Boston?"}],
  stream=True,
  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'
)

function_name = ''
arguments = ''
async for item in gen:
  delta = item.choices[0].delta
  if delta.get('function_call'):
      function_name += delta.function_call.get('name', '')
      arguments += str(delta.function_call.get('arguments', ''))

print(f'function_name: {function_name}\narguments: {json.loads(arguments)}')

Output:

function_name: get_current_weather
arguments: {'location': 'Boston, MA'}

[valaises]

Hi @n3d1117, thanks for your help and solution! It helped 👍 👍 👍