chore: updated semantic conventions on Generative AI spans (#319)

Author: poshinchen

src/strands/event_loop/event_loop.py

@@ -102,7 +102,7 @@ def event_loop_cycle(
     # Create tracer span for this event loop cycle
     tracer = get_tracer()
     cycle_span = tracer.start_event_loop_cycle_span(
-        event_loop_kwargs=kwargs, parent_span=event_loop_parent_span, messages=messages
+        event_loop_kwargs=kwargs, messages=messages, parent_span=event_loop_parent_span
     )
     kwargs["event_loop_cycle_span"] = cycle_span
 
@@ -124,8 +124,8 @@ def event_loop_cycle(
     for attempt in range(MAX_ATTEMPTS):
         model_id = model.config.get("model_id") if hasattr(model, "config") else None
         model_invoke_span = tracer.start_model_invoke_span(
-            parent_span=cycle_span,
             messages=messages,
+            parent_span=cycle_span,
             model_id=model_id,
         )
 
@@ -140,7 +140,7 @@ def event_loop_cycle(
             kwargs.setdefault("request_state", {})
 
             if model_invoke_span:
-                tracer.end_model_invoke_span(model_invoke_span, message, usage)
+                tracer.end_model_invoke_span(model_invoke_span, message, usage, stop_reason)
             break  # Success! Break out of retry loop
 
         except ContextWindowOverflowException as e:

src/strands/telemetry/tracer.py

@@ -14,7 +14,7 @@
 
 from ..agent.agent_result import AgentResult
 from ..types.content import Message, Messages
-from ..types.streaming import Usage
+from ..types.streaming import StopReason, Usage
 from ..types.tools import ToolResult, ToolUse
 from ..types.traces import AttributeValue
 
@@ -196,20 +196,31 @@ def end_span_with_error(self, span: Span, error_message: str, exception: Optiona
         error = exception or Exception(error_message)
         self._end_span(span, error=error)
 
+    def _add_event(self, span: Optional[Span], event_name: str, event_attributes: Dict[str, AttributeValue]) -> None:
+        """Add an event with attributes to a span.
+
+        Args:
+            span: The span to add the event to
+            event_name: Name of the event
+            event_attributes: Dictionary of attributes to set on the event
+        """
+        if not span:
+            return
+
+        span.add_event(event_name, attributes=event_attributes)
+
     def start_model_invoke_span(
         self,
+        messages: Messages,
         parent_span: Optional[Span] = None,
-        agent_name: str = "Strands Agent",
-        messages: Optional[Messages] = None,
         model_id: Optional[str] = None,
         **kwargs: Any,
     ) -> Optional[Span]:
         """Start a new span for a model invocation.
 
         Args:
+            messages: Messages being sent to the model.
             parent_span: Optional parent span to link this span to.
-            agent_name: Name of the agent making the model call.
-            messages: Optional messages being sent to the model.
             model_id: Optional identifier for the model being invoked.
             **kwargs: Additional attributes to add to the span.
 
@@ -219,8 +230,6 @@ def start_model_invoke_span(
         attributes: Dict[str, AttributeValue] = {
             "gen_ai.system": "strands-agents",
             "gen_ai.operation.name": "chat",
-            "gen_ai.agent.name": agent_name,
-            "gen_ai.prompt": serialize(messages),
         }
 
         if model_id:
@@ -229,28 +238,41 @@ def start_model_invoke_span(
         # Add additional kwargs as attributes
         attributes.update({k: v for k, v in kwargs.items() if isinstance(v, (str, int, float, bool))})
 
-        return self._start_span("Model invoke", parent_span, attributes, span_kind=trace_api.SpanKind.CLIENT)
+        span = self._start_span("Model invoke", parent_span, attributes, span_kind=trace_api.SpanKind.CLIENT)
+        for message in messages:
+            self._add_event(
+                span,
+                f"gen_ai.{message['role']}.message",
+                {"content": serialize(message["content"])},
+            )
+        return span
 
     def end_model_invoke_span(
-        self, span: Span, message: Message, usage: Usage, error: Optional[Exception] = None
+        self, span: Span, message: Message, usage: Usage, stop_reason: StopReason, error: Optional[Exception] = None
     ) -> None:
         """End a model invocation span with results and metrics.
 
         Args:
             span: The span to end.
             message: The message response from the model.
             usage: Token usage information from the model call.
+            stop_reason (StopReason): The reason the model stopped generating.
             error: Optional exception if the model call failed.
         """
         attributes: Dict[str, AttributeValue] = {
-            "gen_ai.completion": serialize(message["content"]),
             "gen_ai.usage.prompt_tokens": usage["inputTokens"],
             "gen_ai.usage.input_tokens": usage["inputTokens"],
             "gen_ai.usage.completion_tokens": usage["outputTokens"],
             "gen_ai.usage.output_tokens": usage["outputTokens"],
             "gen_ai.usage.total_tokens": usage["totalTokens"],
         }
 
+        self._add_event(
+            span,
+            "gen_ai.choice",
+            event_attributes={"finish_reason": str(stop_reason), "message": serialize(message["content"])},
+        )
+
         self._end_span(span, attributes, error)
 
     def start_tool_call_span(self, tool: ToolUse, parent_span: Optional[Span] = None, **kwargs: Any) -> Optional[Span]:
@@ -265,18 +287,29 @@ def start_tool_call_span(self, tool: ToolUse, parent_span: Optional[Span] = None
             The created span, or None if tracing is not enabled.
         """
         attributes: Dict[str, AttributeValue] = {
-            "gen_ai.prompt": serialize(tool),
+            "gen_ai.operation.name": "execute_tool",
             "gen_ai.system": "strands-agents",
-            "tool.name": tool["name"],
-            "tool.id": tool["toolUseId"],
-            "tool.parameters": serialize(tool["input"]),
+            "gen_ai.tool.name": tool["name"],
+            "gen_ai.tool.call.id": tool["toolUseId"],
         }
 
         # Add additional kwargs as attributes
         attributes.update(kwargs)
 
         span_name = f"Tool: {tool['name']}"
-        return self._start_span(span_name, parent_span, attributes, span_kind=trace_api.SpanKind.INTERNAL)
+        span = self._start_span(span_name, parent_span, attributes, span_kind=trace_api.SpanKind.INTERNAL)
+
+        self._add_event(
+            span,
+            "gen_ai.tool.message",
+            event_attributes={
+                "role": "tool",
+                "content": serialize(tool["input"]),
+                "id": tool["toolUseId"],
+            },
+        )
+
+        return span
 
     def end_tool_call_span(
         self, span: Span, tool_result: Optional[ToolResult], error: Optional[Exception] = None
@@ -293,30 +326,36 @@ def end_tool_call_span(
             status = tool_result.get("status")
             status_str = str(status) if status is not None else ""
 
-            tool_result_content_json = serialize(tool_result.get("content"))
             attributes.update(
                 {
-                    "tool.result": tool_result_content_json,
-                    "gen_ai.completion": tool_result_content_json,
                     "tool.status": status_str,
                 }
             )
 
+            self._add_event(
+                span,
+                "gen_ai.choice",
+                event_attributes={
+                    "message": serialize(tool_result.get("content")),
+                    "id": tool_result.get("toolUseId", ""),
+                },
+            )
+
         self._end_span(span, attributes, error)
 
     def start_event_loop_cycle_span(
         self,
         event_loop_kwargs: Any,
+        messages: Messages,
         parent_span: Optional[Span] = None,
-        messages: Optional[Messages] = None,
         **kwargs: Any,
     ) -> Optional[Span]:
         """Start a new span for an event loop cycle.
 
         Args:
             event_loop_kwargs: Arguments for the event loop cycle.
             parent_span: Optional parent span to link this span to.
-            messages: Optional messages being processed in this cycle.
+            messages:  Messages being processed in this cycle.
             **kwargs: Additional attributes to add to the span.
 
         Returns:
@@ -326,7 +365,6 @@ def start_event_loop_cycle_span(
         parent_span = parent_span if parent_span else event_loop_kwargs.get("event_loop_parent_span")
 
         attributes: Dict[str, AttributeValue] = {
-            "gen_ai.prompt": serialize(messages),
             "event_loop.cycle_id": event_loop_cycle_id,
         }
 
@@ -337,7 +375,15 @@ def start_event_loop_cycle_span(
         attributes.update({k: v for k, v in kwargs.items() if isinstance(v, (str, int, float, bool))})
 
         span_name = f"Cycle {event_loop_cycle_id}"
-        return self._start_span(span_name, parent_span, attributes, span_kind=trace_api.SpanKind.INTERNAL)
+        span = self._start_span(span_name, parent_span, attributes)
+        for message in messages or []:
+            self._add_event(
+                span,
+                f"gen_ai.{message['role']}.message",
+                {"content": serialize(message["content"])},
+            )
+
+        return span
 
     def end_event_loop_cycle_span(
         self,
@@ -354,13 +400,12 @@ def end_event_loop_cycle_span(
             tool_result_message: Optional tool result message if a tool was called.
             error: Optional exception if the cycle failed.
         """
-        attributes: Dict[str, AttributeValue] = {
-            "gen_ai.completion": serialize(message["content"]),
-        }
+        attributes: Dict[str, AttributeValue] = {}
+        event_attributes: Dict[str, AttributeValue] = {"message": serialize(message["content"])}
 
         if tool_result_message:
-            attributes["tool.result"] = serialize(tool_result_message["content"])
-
+            event_attributes["tool.result"] = serialize(tool_result_message["content"])
+        self._add_event(span, "gen_ai.choice", event_attributes=event_attributes)
         self._end_span(span, attributes, error)
 
     def start_agent_span(
@@ -387,17 +432,15 @@ def start_agent_span(
         """
         attributes: Dict[str, AttributeValue] = {
             "gen_ai.system": "strands-agents",
-            "agent.name": agent_name,
             "gen_ai.agent.name": agent_name,
-            "gen_ai.prompt": prompt,
+            "gen_ai.operation.name": "invoke_agent",
         }
 
         if model_id:
             attributes["gen_ai.request.model"] = model_id
 
         if tools:
             tools_json = serialize(tools)
-            attributes["agent.tools"] = tools_json
             attributes["gen_ai.agent.tools"] = tools_json
 
         # Add custom trace attributes if provided
@@ -407,7 +450,18 @@ def start_agent_span(
         # Add additional kwargs as attributes
         attributes.update({k: v for k, v in kwargs.items() if isinstance(v, (str, int, float, bool))})
 
-        return self._start_span(agent_name, attributes=attributes, span_kind=trace_api.SpanKind.CLIENT)
+        span = self._start_span(
+            f"invoke_agent {agent_name}", attributes=attributes, span_kind=trace_api.SpanKind.CLIENT
+        )
+        self._add_event(
+            span,
+            "gen_ai.user.message",
+            event_attributes={
+                "content": prompt,
+            },
+        )
+
+        return span
 
     def end_agent_span(
         self,
@@ -426,10 +480,10 @@ def end_agent_span(
         attributes: Dict[str, AttributeValue] = {}
 
         if response:
-            attributes.update(
-                {
-                    "gen_ai.completion": str(response),
-                }
+            self._add_event(
+                span,
+                "gen_ai.choice",
+                event_attributes={"message": str(response), "finish_reason": str(response.stop_reason)},
             )
 
             if hasattr(response, "metrics") and hasattr(response.metrics, "accumulated_usage"):