fix: remove attributes where the value is NOT_GIVEN in openai-v2 instrumentation for chat.completions.create operations

Author: benkawecki

instrumentation-genai/opentelemetry-instrumentation-openai-v2/CHANGELOG.md

@@ -9,6 +9,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - Record prompt and completion events regardless of span sampling decision.
   ([#3226](https://github.com/open-telemetry/opentelemetry-python-contrib/pull/3226))
+- Filter out attributes with the value of NOT_GIVEN for chat.completions.create.
+  ([#3760](https://github.com/open-telemetry/opentelemetry-python-contrib/pull/3760))
 
 ## Version 2.1b0 (2025-01-18)
 
@@ -30,4 +32,4 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   ([#2925](https://github.com/open-telemetry/opentelemetry-python-contrib/pull/2925))
 
 - Initial OpenAI instrumentation
-  ([#2759](https://github.com/open-telemetry/opentelemetry-python-contrib/pull/2759))
+  ([#2759](https://github.com/open-telemetry/opentelemetry-python-contrib/pull/2759))

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

@@ -225,8 +225,10 @@ def get_llm_request_attributes(
         service_tier if service_tier != "auto" else None
     )
 
-    # filter out None values
-    return {k: v for k, v in attributes.items() if v is not None}
+    # filter out None values and NOT_GIVEN values
+    return {
+        k: v for k, v in attributes.items() if v is not None and v != NOT_GIVEN
+    }
 
 
 def handle_span_exception(span, error):

instrumentation-genai/opentelemetry-instrumentation-openai-v2/tests/test_chat_completions.py

@@ -13,10 +13,11 @@
 # limitations under the License.
 # pylint: disable=too-many-locals
 
+import logging
 from typing import Optional
 
 import pytest
-from openai import APIConnectionError, NotFoundError, OpenAI
+from openai import NOT_GIVEN, APIConnectionError, NotFoundError, OpenAI
 from openai.resources.chat.completions import ChatCompletion
 
 from opentelemetry.sdk.trace import ReadableSpan
@@ -43,7 +44,9 @@ def test_chat_completion_with_content(
     messages_value = [{"role": "user", "content": "Say this is a test"}]
 
     response = openai_client.chat.completions.create(
-        messages=messages_value, model=llm_model_value, stream=False
+        messages=messages_value,
+        model=llm_model_value,
+        stream=False,
     )
 
     spans = span_exporter.get_finished_spans()
@@ -68,6 +71,30 @@ def test_chat_completion_with_content(
     assert_message_in_logs(logs[1], "gen_ai.choice", choice_event, spans[0])
 
 
+@pytest.mark.vcr()
+def test_chat_completion_handles_not_given(
+    span_exporter, log_exporter, openai_client, instrument_no_content, caplog
+):
+    caplog.set_level(logging.WARNING)
+    llm_model_value = "gpt-4o-mini"
+    messages_value = [{"role": "user", "content": "Say this is a test"}]
+
+    response = openai_client.chat.completions.create(
+        messages=messages_value,
+        model=llm_model_value,
+        stream=False,
+        top_p=NOT_GIVEN,
+    )
+
+    spans = span_exporter.get_finished_spans()
+    assert_completion_attributes(spans[0], llm_model_value, response)
+
+    logs = log_exporter.get_finished_logs()
+    assert len(logs) == 2
+
+    assert_no_invalid_type_warning(caplog)
+
+
 @pytest.mark.vcr()
 def test_chat_completion_no_content(
     span_exporter, log_exporter, openai_client, instrument_no_content
@@ -947,3 +974,7 @@ def get_current_weather_tool_definition():
             },
         },
     }
+
+
+def assert_no_invalid_type_warning(caplog):
+    assert "Invalid type" not in caplog.text