OpenAI completion logging fails when using function-calling

Summary of problem

If you have OpenAI chat completion logging enabled, and your code is using the recent ChatGPT or GPT 4 function-calling features, ddtrace will fail to log completions and there will be a lot of warnings like this:

Failed to set text tag 'openai.response.choices.0.message.content', 1 additional messages skipped
...
TypeError: not expecting type '<class 'NoneType'>'
raise TypeError("not expecting type '%s'" % type(s))
File "/venv/lib/python3.11/site-packages/six.py", line 955, in ensure_text
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
self._meta[key] = ensure_text(value, errors="replace")
File "/venv/lib/python3.11/site-packages/ddtrace/span.py", line 383, in set_tag_str
Traceback (most recent call last):

I believe this is because the code here effectively assumes that the content member on a returned message completion can never be null, but if you check the API docs here, that's actually not the case. It seems it can be null if instead function_call has been set.

Which version of dd-trace-py are you using?

1.20.3

Which version of pip are you using?

23.1.2

Which libraries and their versions are you using?

`pip freeze`

aiohttp==3.8.5 aiosignal==1.3.1 alembic==1.12.0 amqp==5.1.1 anthropic==0.3.11 anyio==3.7.1 astroid==2.15.8 async-timeout==4.0.3 asyncpg==0.27.0 attrs==23.1.0 Authlib==1.2.1 azure-core==1.29.4 azure-storage-blob==12.18.2 beautifulsoup4==4.12.2 billiard==4.1.0 black==22.12.0 brotlipy==0.7.0 bytecode==0.15.0 cachetools==5.3.1 cattrs==23.1.2 celery==5.3.1 certifi==2023.7.22 cffi==1.16.0 charset-normalizer==3.3.0 click==8.1.7 click-didyoumean==0.3.0 click-plugins==1.1.1 click-repl==0.3.0 colorama==0.4.6 coverage==7.3.2 cryptography==41.0.4 databases==0.7.0 dataclasses-json==0.5.9 ddsketch==2.0.4 ddtrace==1.20.3 Deprecated==1.2.14 deprecation==2.1.0 dill==0.3.7 distro==1.8.0 docker==6.1.3 envier==0.4.0 fastapi==0.95.2 fastapi-pagination==0.12.10 filelock==3.12.4 flake8==6.1.0 flower==2.0.1 frozenlist==1.4.0 fsspec==2023.9.2 google-api-core==2.12.0 google-api-python-client==2.102.0 google-auth==2.23.2 google-auth-httplib2==0.1.1 googleapis-common-protos==1.60.0 greenlet==2.0.2 gunicorn==20.1.0 h11==0.14.0 halo==0.0.31 httpcore==0.17.3 httplib2==0.22.0 httptools==0.6.0 httpx==0.24.1 huggingface-hub==0.16.4 humanize==4.8.0 idna==3.4 importlib-metadata==6.8.0 iniconfig==2.0.0 isodate==0.6.1 isort==5.12.0 itsdangerous==2.1.2 json-logic-qubit==0.9.1 jsonschema==4.19.1 jsonschema-specifications==2023.7.1 kombu==5.3.2 langchain==0.0.287 langsmith==0.0.42 lazy-object-proxy==1.9.0 Levenshtein==0.22.0 log-symbols==0.0.14 Mako==1.2.4 MarkupSafe==2.1.3 marshmallow==3.20.1 marshmallow-enum==1.5.1 mccabe==0.7.0 multidict==6.0.4 mypy-extensions==1.0.0 number-parser==0.3.2 numexpr==2.8.7 numpy==1.25.2 openai==0.27.10 opentelemetry-api==1.20.0 orjson==3.9.7 packaging==23.2 pandas==1.5.3 parsy==1.4.0 pathspec==0.11.2 Pillow==10.0.1 platformdirs==3.11.0 pluggy==1.3.0 prometheus-client==0.17.1 prompt-toolkit==3.0.39 protobuf==4.24.4 psycopg2-binary==2.9.9 pyasn1==0.5.0 pyasn1-modules==0.3.0 pycodestyle==2.11.0 pycountry==22.3.5 pycparser==2.21 pydantic==1.10.13 pyflakes==3.1.0 pylint==2.17.7 pyparsing==3.1.1 PyPDF2==2.12.1 pytest==7.4.2 pytest-asyncio==0.20.3 pytest-cov==4.1.0 pytest-mock==3.11.1 pytest-recording==0.12.2 python-dateutil==2.8.2 python-dotenv==1.0.0 python-Levenshtein==0.22.0 python-multipart==0.0.6 pytz==2023.3.post1 PyYAML==6.0.1 rapidfuzz==3.3.1 redis==5.0.1 referencing==0.30.2 regex==2022.10.31 reportlab==3.6.13 requests==2.31.0 -e git+ssh://git@github.com/orbitalwitness/re-concierge.git@f8825f942fecd1974465ed43a63380a23bc16b04#egg=risks&subdirectory=api rpds-py==0.10.3 rsa==4.9 sentry-sdk==1.31.0 six==1.16.0 slack-sdk==3.23.0 sniffio==1.3.0 soupsieve==2.5 spinners==0.0.24 SQLAlchemy==1.4.49 SQLAlchemy-Utils==0.39.0 starlette==0.27.0 structlog==22.3.0 syrupy==4.5.0 tenacity==8.2.3 termcolor==2.3.0 testcontainers==3.7.1 testcontainers-core==0.0.1rc1 testcontainers-postgres==0.0.1rc1 testcontainers-redis==0.0.1rc1 thefuzz==0.19.0 tiktoken==0.4.0 tokenizers==0.14.0 tomlkit==0.12.1 tornado==6.3.3 tqdm==4.66.1 typer==0.7.0 types-pyOpenSSL==23.2.0.2 types-redis==4.6.0.7 typing-inspect==0.9.0 typing_extensions==4.8.0 tzdata==2023.3 uritemplate==4.1.1 urllib3==2.0.6 uvicorn==0.21.1 uvloop==0.17.0 validators==0.22.0 vcrpy==4.3.1 vine==5.0.0 watchdog==3.0.0 watchfiles==0.20.0 wcwidth==0.2.8 weaviate-client==3.24.2 websocket-client==1.6.3 websockets==11.0.3 wrapt==1.15.0 xmltodict==0.13.0 yarl==1.9.2 zipp==3.17.0

How can we reproduce your problem?

Insrument a Python application using dd-trace with openai library installed
Set DD_OPENAI_LOGS_ENABLED=1 to enable completion logging and adjust sample rate if needed
Make a chat completion call using the function_call API
dd-trace will log a warning, the request won't be logged

What is the result that you get?

Stack trace above in the summary

What is the result that you expected?

Completion logged, no warnings

DataDog / dd-trace-py