Merge branch 'master' into ivana/bump-max-value-length-limit

Author: sentrivana

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
     """
 
@@ -372,6 +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.
@@ -411,7 +431,7 @@ class SPANDATA:
     GEN_AI_REQUEST_MODEL = "gen_ai.request.model"
     """
     The model identifier being used for the request.
-    Example: "gpt-4-turbo-preview"
+    Example: "gpt-4-turbo"
     """
 
     GEN_AI_REQUEST_PRESENCE_PENALTY = "gen_ai.request.presence_penalty"
@@ -649,9 +669,11 @@ class OP:
     FUNCTION_AWS = "function.aws"
     FUNCTION_GCP = "function.gcp"
     GEN_AI_CHAT = "gen_ai.chat"
+    GEN_AI_EMBEDDINGS = "gen_ai.embeddings"
     GEN_AI_EXECUTE_TOOL = "gen_ai.execute_tool"
     GEN_AI_HANDOFF = "gen_ai.handoff"
     GEN_AI_INVOKE_AGENT = "gen_ai.invoke_agent"
+    GEN_AI_RESPONSES = "gen_ai.responses"
     GRAPHQL_EXECUTE = "graphql.execute"
     GRAPHQL_MUTATION = "graphql.mutation"
     GRAPHQL_PARSE = "graphql.parse"
@@ -674,8 +696,6 @@ class OP:
     MIDDLEWARE_STARLITE = "middleware.starlite"
     MIDDLEWARE_STARLITE_RECEIVE = "middleware.starlite.receive"
     MIDDLEWARE_STARLITE_SEND = "middleware.starlite.send"
-    OPENAI_CHAT_COMPLETIONS_CREATE = "ai.chat_completions.create.openai"
-    OPENAI_EMBEDDINGS_CREATE = "ai.embeddings.create.openai"
     HUGGINGFACE_HUB_CHAT_COMPLETIONS_CREATE = (
         "ai.chat_completions.create.huggingface_hub"
     )

sentry_sdk/integrations/celery/__init__.py

@@ -391,6 +391,7 @@ def _inner(*args, **kwargs):
                         )
 
                 if latency is not None:
+                    latency *= 1000  # milliseconds
                     span.set_data(SPANDATA.MESSAGING_MESSAGE_RECEIVE_LATENCY, latency)
 
                 with capture_internal_exceptions():