OpenAI pipeline name and streaming attribute (#4628)

Rename attributes for AI pipeline name and the streaming boolean.

---------

Co-authored-by: Ivana Kellyer <[email protected]>

Author: antonpirker

sentry_sdk/ai/monitoring.py

@@ -40,7 +40,7 @@ def sync_wrapped(*args, **kwargs):
                 for k, v in kwargs.pop("sentry_data", {}).items():
                     span.set_data(k, v)
                 if curr_pipeline:
-                    span.set_data(SPANDATA.AI_PIPELINE_NAME, curr_pipeline)
+                    span.set_data(SPANDATA.GEN_AI_PIPELINE_NAME, curr_pipeline)
                     return f(*args, **kwargs)
                 else:
                     _ai_pipeline_name.set(description)
@@ -69,7 +69,7 @@ async def async_wrapped(*args, **kwargs):
                 for k, v in kwargs.pop("sentry_data", {}).items():
                     span.set_data(k, v)
                 if curr_pipeline:
-                    span.set_data(SPANDATA.AI_PIPELINE_NAME, curr_pipeline)
+                    span.set_data(SPANDATA.GEN_AI_PIPELINE_NAME, curr_pipeline)
                     return await f(*args, **kwargs)
                 else:
                     _ai_pipeline_name.set(description)
@@ -108,7 +108,7 @@ def record_token_usage(
     # TODO: move pipeline name elsewhere
     ai_pipeline_name = get_ai_pipeline_name()
     if ai_pipeline_name:
-        span.set_data(SPANDATA.AI_PIPELINE_NAME, ai_pipeline_name)
+        span.set_data(SPANDATA.GEN_AI_PIPELINE_NAME, ai_pipeline_name)
 
     if input_tokens is not None:
         span.set_data(SPANDATA.GEN_AI_USAGE_INPUT_TOKENS, input_tokens)

sentry_sdk/consts.py

@@ -169,6 +169,7 @@ class SPANDATA:
     AI_PIPELINE_NAME = "ai.pipeline.name"
     """
     Name of the AI pipeline or chain being executed.
+    DEPRECATED: Use GEN_AI_PIPELINE_NAME instead.
     Example: "qa-pipeline"
     """
 
@@ -229,6 +230,7 @@ class SPANDATA:
     AI_STREAMING = "ai.streaming"
     """
     Whether or not the AI model call's response was streamed back asynchronously
+    DEPRECATED: Use GEN_AI_RESPONSE_STREAMING instead.
     Example: true
     """
 
@@ -372,12 +374,24 @@ class SPANDATA:
     Example: "chat"
     """
 
+    GEN_AI_PIPELINE_NAME = "gen_ai.pipeline.name"
+    """
+    Name of the AI pipeline or chain being executed.
+    Example: "qa-pipeline"
+    """
+
     GEN_AI_RESPONSE_MODEL = "gen_ai.response.model"
     """
     Exact model identifier used to generate the response
     Example: gpt-4o-mini-2024-07-18
     """
 
+    GEN_AI_RESPONSE_STREAMING = "gen_ai.response.streaming"
+    """
+    Whether or not the AI model call's response was streamed back asynchronously
+    Example: true
+    """
+
     GEN_AI_RESPONSE_TEXT = "gen_ai.response.text"
     """
     The model's response text messages.

sentry_sdk/integrations/openai.py

@@ -183,7 +183,7 @@ def _set_input_data(span, kwargs, operation, integration):
     # Input attributes: Optional
     kwargs_keys_to_attributes = {
         "model": SPANDATA.GEN_AI_REQUEST_MODEL,
-        "stream": SPANDATA.AI_STREAMING,
+        "stream": SPANDATA.GEN_AI_RESPONSE_STREAMING,
         "max_tokens": SPANDATA.GEN_AI_REQUEST_MAX_TOKENS,
         "presence_penalty": SPANDATA.GEN_AI_REQUEST_PRESENCE_PENALTY,
         "frequency_penalty": SPANDATA.GEN_AI_REQUEST_FREQUENCY_PENALTY,

tests/integrations/openai/test_openai.py

@@ -1159,11 +1159,11 @@ async def test_ai_client_span_streaming_responses_async_api(
     assert spans[0]["op"] == "gen_ai.responses"
     assert spans[0]["origin"] == "auto.ai.openai"
     assert spans[0]["data"] == {
-        "ai.streaming": True,
         "gen_ai.operation.name": "responses",
         "gen_ai.request.messages": "How do I check if a Python object is an instance of a class?",
         "gen_ai.request.model": "gpt-4o",
         "gen_ai.response.model": "response-model-id",
+        "gen_ai.response.streaming": True,
         "gen_ai.system": "openai",
         "gen_ai.usage.input_tokens": 20,
         "gen_ai.usage.input_tokens.cached": 5,