chore(openai): fix responses reasoning token metric tracking support for streaming

.riot/requirements/1911f94.txt

@@ -0,0 +1,51 @@
+#
+# This file is autogenerated by pip-compile with Python 3.10
+# by the following command:
+#
+#    pip-compile --allow-unsafe --no-annotate .riot/requirements/1911f94.in
+#
+annotated-types==0.7.0
+anyio==4.9.0
+attrs==25.3.0
+certifi==2025.7.14
+charset-normalizer==3.4.2
+coverage[toml]==7.9.2
+distro==1.9.0
+exceptiongroup==1.3.0
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+hypothesis==6.45.0
+idna==3.10
+iniconfig==2.1.0
+jiter==0.10.0
+mock==5.2.0
+multidict==6.6.3
+openai==1.66.0
+opentracing==2.4.0
+packaging==25.0
+pillow==11.3.0
+pluggy==1.6.0
+propcache==0.3.2
+pydantic==2.11.7
+pydantic-core==2.33.2
+pygments==2.19.2
+pytest==8.4.1
+pytest-asyncio==0.21.1
+pytest-cov==6.2.1
+pytest-mock==3.14.1
+pytest-randomly==3.16.0
+pyyaml==6.0.2
+regex==2024.11.6
+requests==2.32.4
+sniffio==1.3.1
+sortedcontainers==2.4.0
+tiktoken==0.9.0
+tomli==2.2.1
+tqdm==4.67.1
+typing-extensions==4.14.1
+typing-inspection==0.4.1
+urllib3==1.26.20
+vcrpy==7.0.0
+wrapt==1.17.2
+yarl==1.20.1

.riot/requirements/1a0657d.txt

@@ -0,0 +1,49 @@
+#
+# This file is autogenerated by pip-compile with Python 3.13
+# by the following command:
+#
+#    pip-compile --allow-unsafe --no-annotate .riot/requirements/1a0657d.in
+#
+annotated-types==0.7.0
+anyio==4.9.0
+attrs==25.3.0
+certifi==2025.7.14
+charset-normalizer==3.4.2
+coverage[toml]==7.9.2
+distro==1.9.0
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+hypothesis==6.45.0
+idna==3.10
+iniconfig==2.1.0
+jiter==0.10.0
+mock==5.2.0
+multidict==6.6.3
+openai==1.66.0
+opentracing==2.4.0
+packaging==25.0
+pillow==11.3.0
+pluggy==1.6.0
+propcache==0.3.2
+pydantic==2.11.7
+pydantic-core==2.33.2
+pygments==2.19.2
+pytest==8.4.1
+pytest-asyncio==0.21.1
+pytest-cov==6.2.1
+pytest-mock==3.14.1
+pytest-randomly==3.16.0
+pyyaml==6.0.2
+regex==2024.11.6
+requests==2.32.4
+sniffio==1.3.1
+sortedcontainers==2.4.0
+tiktoken==0.9.0
+tqdm==4.67.1
+typing-extensions==4.14.1
+typing-inspection==0.4.1
+urllib3==1.26.20
+vcrpy==7.0.0
+wrapt==1.17.2
+yarl==1.20.1

.riot/requirements/85ff44d.txt

@@ -0,0 +1,49 @@
+#
+# This file is autogenerated by pip-compile with Python 3.12
+# by the following command:
+#
+#    pip-compile --allow-unsafe --no-annotate .riot/requirements/85ff44d.in
+#
+annotated-types==0.7.0
+anyio==4.9.0
+attrs==25.3.0
+certifi==2025.7.14
+charset-normalizer==3.4.2
+coverage[toml]==7.9.2
+distro==1.9.0
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+hypothesis==6.45.0
+idna==3.10
+iniconfig==2.1.0
+jiter==0.10.0
+mock==5.2.0
+multidict==6.6.3
+openai==1.66.0
+opentracing==2.4.0
+packaging==25.0
+pillow==11.3.0
+pluggy==1.6.0
+propcache==0.3.2
+pydantic==2.11.7
+pydantic-core==2.33.2
+pygments==2.19.2
+pytest==8.4.1
+pytest-asyncio==0.21.1
+pytest-cov==6.2.1
+pytest-mock==3.14.1
+pytest-randomly==3.16.0
+pyyaml==6.0.2
+regex==2024.11.6
+requests==2.32.4
+sniffio==1.3.1
+sortedcontainers==2.4.0
+tiktoken==0.9.0
+tqdm==4.67.1
+typing-extensions==4.14.1
+typing-inspection==0.4.1
+urllib3==1.26.20
+vcrpy==7.0.0
+wrapt==1.17.2
+yarl==1.20.1

.riot/requirements/b68c552.txt

@@ -0,0 +1,51 @@
+#
+# This file is autogenerated by pip-compile with Python 3.8
+# by the following command:
+#
+#    pip-compile --allow-unsafe --no-annotate .riot/requirements/b68c552.in
+#
+annotated-types==0.7.0
+anyio==4.5.2
+attrs==25.3.0
+certifi==2025.7.14
+charset-normalizer==3.4.2
+coverage[toml]==7.6.1
+distro==1.9.0
+exceptiongroup==1.3.0
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+hypothesis==6.45.0
+idna==3.10
+importlib-metadata==8.5.0
+iniconfig==2.1.0
+jiter==0.9.1
+mock==5.2.0
+multidict==6.1.0
+openai==1.66.0
+opentracing==2.4.0
+packaging==25.0
+pillow==10.4.0
+pluggy==1.5.0
+propcache==0.2.0
+pydantic==2.10.6
+pydantic-core==2.27.2
+pytest==8.3.5
+pytest-asyncio==0.21.1
+pytest-cov==5.0.0
+pytest-mock==3.14.1
+pytest-randomly==3.15.0
+pyyaml==6.0.2
+regex==2024.11.6
+requests==2.32.4
+sniffio==1.3.1
+sortedcontainers==2.4.0
+tiktoken==0.7.0
+tomli==2.2.1
+tqdm==4.67.1
+typing-extensions==4.13.2
+urllib3==1.26.20
+vcrpy==6.0.2
+wrapt==1.17.2
+yarl==1.15.2
+zipp==3.20.2

.riot/requirements/c5399d2.txt

@@ -0,0 +1,49 @@
+#
+# This file is autogenerated by pip-compile with Python 3.11
+# by the following command:
+#
+#    pip-compile --allow-unsafe --no-annotate .riot/requirements/c5399d2.in
+#
+annotated-types==0.7.0
+anyio==4.9.0
+attrs==25.3.0
+certifi==2025.7.14
+charset-normalizer==3.4.2
+coverage[toml]==7.9.2
+distro==1.9.0
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+hypothesis==6.45.0
+idna==3.10
+iniconfig==2.1.0
+jiter==0.10.0
+mock==5.2.0
+multidict==6.6.3
+openai==1.66.0
+opentracing==2.4.0
+packaging==25.0
+pillow==11.3.0
+pluggy==1.6.0
+propcache==0.3.2
+pydantic==2.11.7
+pydantic-core==2.33.2
+pygments==2.19.2
+pytest==8.4.1
+pytest-asyncio==0.21.1
+pytest-cov==6.2.1
+pytest-mock==3.14.1
+pytest-randomly==3.16.0
+pyyaml==6.0.2
+regex==2024.11.6
+requests==2.32.4
+sniffio==1.3.1
+sortedcontainers==2.4.0
+tiktoken==0.9.0
+tqdm==4.67.1
+typing-extensions==4.14.1
+typing-inspection==0.4.1
+urllib3==1.26.20
+vcrpy==7.0.0
+wrapt==1.17.2
+yarl==1.20.1

.riot/requirements/e9c67e1.txt

@@ -0,0 +1,53 @@
+#
+# This file is autogenerated by pip-compile with Python 3.9
+# by the following command:
+#
+#    pip-compile --allow-unsafe --no-annotate .riot/requirements/e9c67e1.in
+#
+annotated-types==0.7.0
+anyio==4.9.0
+attrs==25.3.0
+certifi==2025.7.14
+charset-normalizer==3.4.2
+coverage[toml]==7.9.2
+distro==1.9.0
+exceptiongroup==1.3.0
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+hypothesis==6.45.0
+idna==3.10
+importlib-metadata==8.7.0
+iniconfig==2.1.0
+jiter==0.10.0
+mock==5.2.0
+multidict==6.6.3
+openai==1.66.0
+opentracing==2.4.0
+packaging==25.0
+pillow==11.3.0
+pluggy==1.6.0
+propcache==0.3.2
+pydantic==2.11.7
+pydantic-core==2.33.2
+pygments==2.19.2
+pytest==8.4.1
+pytest-asyncio==0.21.1
+pytest-cov==6.2.1
+pytest-mock==3.14.1
+pytest-randomly==3.16.0
+pyyaml==6.0.2
+regex==2024.11.6
+requests==2.32.4
+sniffio==1.3.1
+sortedcontainers==2.4.0
+tiktoken==0.9.0
+tomli==2.2.1
+tqdm==4.67.1
+typing-extensions==4.14.1
+typing-inspection==0.4.1
+urllib3==1.26.20
+vcrpy==7.0.0
+wrapt==1.17.2
+yarl==1.20.1
+zipp==3.23.0

ddtrace/llmobs/_integrations/openai.py

@@ -153,9 +153,9 @@ def _extract_llmobs_metrics_tags(
 
         # in the streamed responses case, `resp` is a list with `usage` being stored in the first element
         if resp and isinstance(resp, list) and _get_attr(resp[0], "usage", None):
-            token_usage = resp[0].get("usage", {})
-        elif resp and getattr(resp, "usage", None):
-            token_usage = resp.usage
+            token_usage = _get_attr(resp[0], "usage", {})
+        elif resp and _get_attr(resp, "usage", None):
+            token_usage = _get_attr(resp, "usage", {})
         if token_usage is not None:
             prompt_tokens = _get_attr(token_usage, "prompt_tokens", 0)
             completion_tokens = _get_attr(token_usage, "completion_tokens", 0)

riotfile.py

@@ -2628,7 +2628,7 @@ def select_pys(min_version: str = MIN_PYTHON_VERSION, max_version: str = MAX_PYT
                 Venv(
                     pys=select_pys(min_version="3.8"),
                     pkgs={
-                        "openai": [latest, "~=1.76.2"],
+                        "openai": [latest, "~=1.76.2", "==1.66.0"],
                         "tiktoken": latest,
                         "pillow": latest,
                     },

tests/contrib/openai/test_openai_llmobs.py

@@ -1737,43 +1737,12 @@ def test_responses_reasoning_stream(self, openai, ddtrace_global_config, mock_ll
         for event in stream:
             pass
 
-        span = mock_tracer.pop_traces()[0][0]
-        assert mock_llmobs_writer.enqueue.call_count == 1
-        mock_llmobs_writer.enqueue.assert_called_with(
-            _expected_llmobs_llm_span_event(
-                span,
-                model_name="o4-mini-2025-04-16",
-                model_provider="openai",
-                input_messages=[{"content": "If one plus a number is 10, what is the number?", "role": "user"}],
-                output_messages=[
-                    {"role": "reasoning", "content": mock.ANY},
-                    {"role": "assistant", "content": "The number is 9, since 1 + x = 10 ⇒ x = 10 − 1 = 9."},
-                ],
-                metadata={
-                    "reasoning": {"effort": "medium", "summary": "detailed"},
-                    "stream": True,
-                    "temperature": 1.0,
-                    "top_p": 1.0,
-                    "tools": [],
-                    "tool_choice": "auto",
-                    "truncation": "disabled",
-                    "text": {"format": {"type": "text"}},
-                    "reasoning_tokens": 128,
-                },
-                token_metrics={
-                    "input_tokens": mock.ANY,
-                    "output_tokens": mock.ANY,
-                    "total_tokens": mock.ANY,
-                    "cache_read_input_tokens": mock.ANY,
-                },
-                tags={"ml_app": "<ml-app-name>", "service": "tests.contrib.openai"},
-            )
-        )
-
         # special assertion on rough reasoning content
         span_event = mock_llmobs_writer.enqueue.call_args[0][0]
         reasoning_content = json.loads(span_event["meta"]["output"]["messages"][0]["content"])
+        assistant_content = span_event["meta"]["output"]["messages"][1]["content"]
         assert reasoning_content["summary"] is not None
+        assert assistant_content == "The number is 9, since 1 + x = 10 ⇒ x = 10 − 1 = 9."
 
     @pytest.mark.skipif(
         parse_version(openai_module.version.VERSION) < (1, 66), reason="Response options only available openai >= 1.66"