Add instrumentation for chat.completions.parse() (structured outputs) (#18)

* Add instrumentation for chat.completions.parse() structured outputs

Wrap Completions.parse and AsyncCompletions.parse with the same
telemetry wrappers used for create(). ParsedChatCompletion extends
ChatCompletion so the existing wrapper logic handles it correctly.

A version guard (_is_parse_supported) skips wrapping on openai < 1.40.0
where parse() does not exist.

response_format handling in both the legacy and experimental paths now
recognises a bare Python type (Pydantic model class) and maps it to the
appropriate output type attribute (json on experimental, json_schema on
legacy).

Includes sync and async tests with VCR cassettes, skipped on older SDK
versions.

* add 404 error-path tests for structured outputs

Adds sync and async test_*_structured_output_404 covering the parse()
error path with a 404 cassette and an assertion that the span records
error.type = "NotFoundError".

Mirrors test_chat_completion_404 / test_async_chat_completion_404 in
structure.

Author: Nik-Reddy

instrumentation/opentelemetry-instrumentation-openai-v2/.changelog/18.added

@@ -0,0 +1 @@
+Add instrumentation for `chat.completions.parse()` structured outputs

instrumentation/opentelemetry-instrumentation-openai-v2/src/opentelemetry/instrumentation/openai_v2/__init__.py

@@ -97,9 +97,25 @@
 )
 
 
+def _is_parse_supported():
+    """Check if the parse() method is available on the Completions class.
+
+    The parse() method for structured outputs was added in openai >= 1.40.0.
+    """
+    try:
+        from openai.resources.chat.completions import (  # pylint: disable=import-outside-toplevel  # noqa: PLC0415
+            Completions,
+        )
+
+        return hasattr(Completions, "parse")
+    except ImportError:
+        return False
+
+
 class OpenAIInstrumentor(BaseInstrumentor):
     def __init__(self):
         self._meter = None
+        self._parse_supported = False
 
     def instrumentation_dependencies(self) -> Collection[str]:
         return _instruments
@@ -181,6 +197,36 @@ def _instrument(self, **kwargs):
             ),
         )
 
+        # parse() wraps create() internally in the OpenAI SDK and returns a
+        # ParsedChatCompletion. The telemetry-relevant fields (model, usage,
+        # choices, finish_reason) are identical to ChatCompletion, so the
+        # existing create() wrappers handle it correctly.
+        self._parse_supported = _is_parse_supported()
+        if self._parse_supported:
+            wrap_function_wrapper(
+                "openai.resources.chat.completions",
+                "Completions.parse",
+                (
+                    chat_completions_create_v_new(handler)
+                    if latest_experimental_enabled
+                    else chat_completions_create_v_old(
+                        tracer, logger, instruments, is_content_enabled()
+                    )
+                ),
+            )
+
+            wrap_function_wrapper(
+                "openai.resources.chat.completions",
+                "AsyncCompletions.parse",
+                (
+                    async_chat_completions_create_v_new(handler)
+                    if latest_experimental_enabled
+                    else async_chat_completions_create_v_old(
+                        tracer, logger, instruments, is_content_enabled()
+                    )
+                ),
+            )
+
         responses_module = _get_responses_module()
         # Responses instrumentation is intentionally limited to the latest
         # experimental semconv path. Unlike chat completions, we do not carry
@@ -201,6 +247,9 @@ def _uninstrument(self, **kwargs):
         unwrap(openai.resources.chat.completions.AsyncCompletions, "create")
         unwrap(openai.resources.embeddings.Embeddings, "create")
         unwrap(openai.resources.embeddings.AsyncEmbeddings, "create")
+        if self._parse_supported:
+            unwrap(openai.resources.chat.completions.Completions, "parse")
+            unwrap(openai.resources.chat.completions.AsyncCompletions, "parse")
         responses_module = _get_responses_module()
         if responses_module is not None:
             unwrap(responses_module.Responses, "create")

instrumentation/opentelemetry-instrumentation-openai-v2/src/opentelemetry/instrumentation/openai_v2/utils.py

@@ -280,8 +280,18 @@ def get_llm_request_attributes(
             else GenAIAttributes.GEN_AI_OPENAI_REQUEST_RESPONSE_FORMAT
         )
         if (response_format := kwargs.get("response_format")) is not None:
-            # response_format may be string or object with a string in the `type` key
-            if isinstance(response_format, Mapping):
+            # response_format may be string, object with a string in the `type` key,
+            # or a type (e.g. Pydantic model class used with parse())
+            if isinstance(response_format, type):
+                if latest_experimental_enabled:
+                    attributes[request_response_format_attr_key] = (
+                        GenAIAttributes.GenAiOutputTypeValues.JSON.value
+                    )
+                else:
+                    attributes[request_response_format_attr_key] = (
+                        GenAIAttributes.GenAiOpenaiRequestResponseFormatValues.JSON_SCHEMA.value
+                    )
+            elif isinstance(response_format, Mapping):
                 if (
                     response_format_type := response_format.get("type")
                 ) is not None:
@@ -369,8 +379,13 @@ def create_chat_invocation(
     if (
         response_format := get_value(kwargs.get("response_format"))
     ) is not None:
-        # response_format may be string or object with a string in the `type` key
-        if isinstance(response_format, Mapping):
+        # response_format may be string, object with a string in the `type` key,
+        # or a type (e.g. Pydantic model class used with parse())
+        if isinstance(response_format, type):
+            invocation.attributes[GenAIAttributes.GEN_AI_OUTPUT_TYPE] = (
+                GenAIAttributes.GenAiOutputTypeValues.JSON.value
+            )
+        elif isinstance(response_format, Mapping):
             if (
                 response_format_type := get_value(response_format.get("type"))
             ) is not None:

instrumentation/opentelemetry-instrumentation-openai-v2/tests/cassettes/test_async_structured_output_404.yaml

@@ -0,0 +1,90 @@
+interactions:
+- request:
+    body: |-
+      {
+        "messages": [
+          {
+            "role": "user",
+            "content": "Extract the event information from: Team Meeting on 2024-01-15 with Alice and Bob"
+          }
+        ],
+        "model": "this-model-does-not-exist",
+        "response_format": {
+          "type": "json_schema",
+          "json_schema": {
+            "name": "CalendarEvent",
+            "strict": true,
+            "schema": {
+              "type": "object",
+              "properties": {
+                "name": {"type": "string"},
+                "date": {"type": "string"},
+                "participants": {"items": {"type": "string"}, "type": "array"}
+              },
+              "required": ["name", "date", "participants"],
+              "additionalProperties": false
+            }
+          }
+        }
+      }
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      authorization:
+      - Bearer test_openai_api_key
+      connection:
+      - keep-alive
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - AsyncOpenAI/Python 1.54.3
+      x-stainless-async:
+      - async:asyncio
+      x-stainless-lang:
+      - python
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: |-
+        {
+          "error": {
+            "message": "The model `this-model-does-not-exist` does not exist or you do not have access to it.",
+            "type": "invalid_request_error",
+            "param": null,
+            "code": "model_not_found"
+          }
+        }
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      Connection:
+      - keep-alive
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Mon, 11 Nov 2024 23:43:52 GMT
+      Server:
+      - cloudflare
+      Set-Cookie: test_set_cookie
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      openai-organization: test_openai_org_id
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_structured_output_404_async
+    status:
+      code: 404
+      message: Not Found
+version: 1

instrumentation/opentelemetry-instrumentation-openai-v2/tests/cassettes/test_async_structured_output_no_content.yaml

@@ -0,0 +1,116 @@
+interactions:
+- request:
+    body: |-
+      {
+        "messages": [
+          {
+            "role": "user",
+            "content": "Extract the event information from: Team Meeting on 2024-01-15 with Alice and Bob"
+          }
+        ],
+        "model": "gpt-4o-mini",
+        "response_format": {
+          "type": "json_schema",
+          "json_schema": {
+            "name": "CalendarEvent",
+            "strict": true,
+            "schema": {
+              "type": "object",
+              "properties": {
+                "name": {"type": "string"},
+                "date": {"type": "string"},
+                "participants": {"items": {"type": "string"}, "type": "array"}
+              },
+              "required": ["name", "date", "participants"],
+              "additionalProperties": false
+            }
+          }
+        }
+      }
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      authorization:
+      - Bearer test_openai_api_key
+      connection:
+      - keep-alive
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.54.3
+      x-stainless-async:
+      - async:asyncio
+      x-stainless-lang:
+      - python
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: |-
+        {
+          "id": "chatcmpl-structured-test-004",
+          "object": "chat.completion",
+          "created": 1731368630,
+          "model": "gpt-4o-mini-2024-07-18",
+          "choices": [
+            {
+              "index": 0,
+              "message": {
+                "role": "assistant",
+                "content": "{\"name\": \"Team Meeting\", \"date\": \"2024-01-15\", \"participants\": [\"Alice\", \"Bob\"]}",
+                "refusal": null
+              },
+              "logprobs": null,
+              "finish_reason": "stop"
+            }
+          ],
+          "usage": {
+            "prompt_tokens": 50,
+            "completion_tokens": 30,
+            "total_tokens": 80,
+            "prompt_tokens_details": {
+              "cached_tokens": 0,
+              "audio_tokens": 0
+            },
+            "completion_tokens_details": {
+              "reasoning_tokens": 0,
+              "audio_tokens": 0,
+              "accepted_prediction_tokens": 0,
+              "rejected_prediction_tokens": 0
+            }
+          },
+          "system_fingerprint": "fp_0ba0d124f1"
+        }
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      Connection:
+      - keep-alive
+      Content-Type:
+      - application/json
+      Date:
+      - Mon, 11 Nov 2024 23:43:50 GMT
+      Server:
+      - cloudflare
+      Set-Cookie: test_set_cookie
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      content-length:
+      - '800'
+      openai-organization: test_openai_org_id
+      openai-processing-ms:
+      - '350'
+      openai-version:
+      - '2020-10-01'
+      x-request-id:
+      - req_structured_test_004
+    status:
+      code: 200
+      message: OK
+version: 1