openlit 1.34.31__py3-none-any.whl → 1.34.32__py3-none-any.whl

openlit/__init__.py

@@ -268,7 +268,9 @@ def init(
             logger.error(
                 "OpenLIT metrics setup failed. Metrics will not be available: %s", err
             )
-            return
+            # Set metrics_dict to None and disable metrics instead of returning early
+            metrics_dict = None
+            disable_metrics = True
 
         if (
             os.getenv("OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT", "").lower

openlit/instrumentation/ag2/__init__.py

@@ -5,10 +5,21 @@ import importlib.metadata
 from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
 from wrapt import wrap_function_wrapper
 
-from openlit.instrumentation.ag2.ag2 import conversable_agent, agent_run
+from openlit.instrumentation.ag2.ag2 import (
+    conversable_agent,
+    agent_run,
+    agent_generate_reply,
+    agent_receive,
+    agent_send,
+    groupchat_manager_run_chat,
+    groupchat_select_speaker,
+)
 from openlit.instrumentation.ag2.async_ag2 import (
     async_conversable_agent,
     async_agent_run,
+    async_agent_generate_reply,
+    async_agent_receive,
+    async_agent_send,
 )
 
 _instruments = ("ag2 >= 0.3.2",)
@@ -64,6 +75,86 @@ class AG2Instrumentor(BaseInstrumentor):
             ),
         )
 
+        # sync agent generate_reply
+        wrap_function_wrapper(
+            "autogen.agentchat.conversable_agent",
+            "ConversableAgent.generate_reply",
+            agent_generate_reply(
+                version,
+                environment,
+                application_name,
+                tracer,
+                pricing_info,
+                capture_message_content,
+                metrics,
+                disable_metrics,
+            ),
+        )
+
+        # sync agent receive
+        wrap_function_wrapper(
+            "autogen.agentchat.conversable_agent",
+            "ConversableAgent.receive",
+            agent_receive(
+                version,
+                environment,
+                application_name,
+                tracer,
+                pricing_info,
+                capture_message_content,
+                metrics,
+                disable_metrics,
+            ),
+        )
+
+        # sync agent send
+        wrap_function_wrapper(
+            "autogen.agentchat.conversable_agent",
+            "ConversableAgent.send",
+            agent_send(
+                version,
+                environment,
+                application_name,
+                tracer,
+                pricing_info,
+                capture_message_content,
+                metrics,
+                disable_metrics,
+            ),
+        )
+
+        # sync groupchat manager run_chat
+        wrap_function_wrapper(
+            "autogen.agentchat.groupchat",
+            "GroupChatManager.run_chat",
+            groupchat_manager_run_chat(
+                version,
+                environment,
+                application_name,
+                tracer,
+                pricing_info,
+                capture_message_content,
+                metrics,
+                disable_metrics,
+            ),
+        )
+
+        # sync groupchat select_speaker
+        wrap_function_wrapper(
+            "autogen.agentchat.groupchat",
+            "GroupChat.select_speaker",
+            groupchat_select_speaker(
+                version,
+                environment,
+                application_name,
+                tracer,
+                pricing_info,
+                capture_message_content,
+                metrics,
+                disable_metrics,
+            ),
+        )
+
     def _uninstrument(self, **kwargs):
         # Proper uninstrumentation logic to revert patched methods
         pass

openlit/instrumentation/ag2/ag2.py

@@ -8,10 +8,45 @@ from openlit.__helpers import handle_exception, set_server_address_and_port
 from openlit.instrumentation.ag2.utils import (
     process_agent_creation,
     process_agent_run,
+    process_agent_generate_reply,
+    process_agent_receive,
+    process_agent_send,
+    process_groupchat_operation,
+    process_speaker_selection,
 )
 from openlit.semcov import SemanticConvention
 
 
+def extract_agent_name(instance, fallback="unknown_agent"):
+    """
+    Extract agent name from AG2 instance with intelligent fallbacks.
+
+    Args:
+        instance: AG2 instance (Agent, GroupChat, etc.)
+        fallback: Default name if no name can be extracted
+
+    Returns:
+        str: Agent name or meaningful fallback
+    """
+    # Try to get the name attribute first
+    agent_name = getattr(instance, "name", None)
+    if agent_name:
+        return agent_name
+
+    # Try to get from class name and make it meaningful
+    class_name = getattr(instance, "__class__", type(instance)).__name__.lower()
+
+    # Map common AG2 class names to meaningful names
+    class_name_map = {
+        "conversableagent": "conversable_agent",
+        "groupchat": "group_chat",
+        "groupchatmanager": "group_chat_manager",
+        "agent": "agent",
+    }
+
+    return class_name_map.get(class_name, fallback)
+
+
 def conversable_agent(
     version,
     environment,
@@ -34,7 +69,7 @@ def conversable_agent(
         server_address, server_port = set_server_address_and_port(
             instance, "127.0.0.1", 80
         )
-        agent_name = kwargs.get("name", "NOT_FOUND")
+        agent_name = kwargs.get("name", "unknown_agent")
         llm_config = kwargs.get("llm_config", {})
         system_message = kwargs.get("system_message", "")
 
@@ -96,12 +131,12 @@ def agent_run(
         )
 
         # Extract agent name from instance
-        agent_name = getattr(instance, "name", "NOT_FOUND")
+        agent_name = extract_agent_name(instance)
 
         # Extract model from instance llm_config
-        request_model = "gpt-4o"
+        request_model = "unknown"
         if hasattr(instance, "llm_config") and isinstance(instance.llm_config, dict):
-            request_model = instance.llm_config.get("model", "gpt-4o")
+            request_model = instance.llm_config.get("model", "unknown")
 
         span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_EXECUTE_AGENT_TASK} {agent_name}"
 
@@ -133,3 +168,389 @@ def agent_run(
             return response
 
     return wrapper
+
+
+def agent_generate_reply(
+    version,
+    environment,
+    application_name,
+    tracer,
+    pricing_info,
+    capture_message_content,
+    metrics,
+    disable_metrics,
+):
+    """
+    Generates a telemetry wrapper for AG2 ConversableAgent.generate_reply.
+    """
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the AG2 ConversableAgent.generate_reply call.
+        """
+
+        server_address, server_port = set_server_address_and_port(
+            instance, "127.0.0.1", 80
+        )
+
+        # Extract agent name from instance
+        agent_name = extract_agent_name(instance)
+
+        # Extract model from instance llm_config
+        request_model = "unknown"
+        if hasattr(instance, "llm_config") and isinstance(instance.llm_config, dict):
+            request_model = instance.llm_config.get("model", "unknown")
+
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_EXECUTE_AGENT_TASK} {agent_name}"
+
+        with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+            start_time = time.time()
+            response = wrapped(*args, **kwargs)
+
+            try:
+                response = process_agent_generate_reply(
+                    response=response,
+                    agent_name=agent_name,
+                    request_model=request_model,
+                    messages=args[0] if args else kwargs.get("messages", []),
+                    sender=args[1] if len(args) > 1 else kwargs.get("sender", None),
+                    pricing_info=pricing_info,
+                    server_port=server_port,
+                    server_address=server_address,
+                    environment=environment,
+                    application_name=application_name,
+                    metrics=metrics,
+                    start_time=start_time,
+                    span=span,
+                    capture_message_content=capture_message_content,
+                    disable_metrics=disable_metrics,
+                    version=version,
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
+
+    return wrapper
+
+
+def agent_receive(
+    version,
+    environment,
+    application_name,
+    tracer,
+    pricing_info,
+    capture_message_content,
+    metrics,
+    disable_metrics,
+):
+    """
+    Generates a telemetry wrapper for AG2 ConversableAgent.receive.
+    """
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the AG2 ConversableAgent.receive call.
+        """
+
+        server_address, server_port = set_server_address_and_port(
+            instance, "127.0.0.1", 80
+        )
+
+        # Extract agent name from instance
+        agent_name = extract_agent_name(instance)
+
+        # Extract sender information
+        sender = args[0] if args else kwargs.get("sender", None)
+        sender_name = getattr(sender, "name", "Unknown") if sender else "Unknown"
+
+        # Extract message
+        message = args[1] if len(args) > 1 else kwargs.get("message", "")
+
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_EXECUTE_AGENT_TASK} {agent_name}"
+
+        with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+            start_time = time.time()
+            response = wrapped(*args, **kwargs)
+
+            try:
+                process_agent_receive(
+                    message=message,
+                    agent_name=agent_name,
+                    sender_name=sender_name,
+                    agent_instance=instance,
+                    pricing_info=pricing_info,
+                    server_port=server_port,
+                    server_address=server_address,
+                    environment=environment,
+                    application_name=application_name,
+                    metrics=metrics,
+                    start_time=start_time,
+                    span=span,
+                    capture_message_content=capture_message_content,
+                    disable_metrics=disable_metrics,
+                    version=version,
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
+
+    return wrapper
+
+
+def agent_send(
+    version,
+    environment,
+    application_name,
+    tracer,
+    pricing_info,
+    capture_message_content,
+    metrics,
+    disable_metrics,
+):
+    """
+    Generates a telemetry wrapper for AG2 ConversableAgent.send.
+    """
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the AG2 ConversableAgent.send call.
+        """
+
+        server_address, server_port = set_server_address_and_port(
+            instance, "127.0.0.1", 80
+        )
+
+        # Extract agent name from instance
+        agent_name = extract_agent_name(instance)
+
+        # Extract recipient information
+        recipient = args[0] if args else kwargs.get("recipient", None)
+        recipient_name = (
+            getattr(recipient, "name", "Unknown") if recipient else "Unknown"
+        )
+
+        # Extract message
+        message = args[1] if len(args) > 1 else kwargs.get("message", "")
+
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_EXECUTE_AGENT_TASK} {agent_name}"
+
+        with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+            start_time = time.time()
+            response = wrapped(*args, **kwargs)
+
+            try:
+                process_agent_send(
+                    message=message,
+                    agent_name=agent_name,
+                    recipient_name=recipient_name,
+                    agent_instance=instance,
+                    pricing_info=pricing_info,
+                    server_port=server_port,
+                    server_address=server_address,
+                    environment=environment,
+                    application_name=application_name,
+                    metrics=metrics,
+                    start_time=start_time,
+                    span=span,
+                    capture_message_content=capture_message_content,
+                    disable_metrics=disable_metrics,
+                    version=version,
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
+
+    return wrapper
+
+
+def groupchat_manager_run_chat(
+    version,
+    environment,
+    application_name,
+    tracer,
+    pricing_info,
+    capture_message_content,
+    metrics,
+    disable_metrics,
+):
+    """
+    Generates a telemetry wrapper for AG2 GroupChatManager.run_chat.
+    """
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the AG2 GroupChatManager.run_chat call.
+        """
+
+        server_address, server_port = set_server_address_and_port(
+            instance, "127.0.0.1", 80
+        )
+
+        # Extract groupchat information
+        groupchat = getattr(instance, "groupchat", None)
+        if groupchat:
+            participants = [agent.name for agent in groupchat.agents]
+            group_name = f"GroupChat_{len(participants)}_agents"
+        else:
+            participants = []
+            group_name = "UnknownGroupChat"
+
+        # Extract model information from GroupChatManager
+        request_model = "unknown"  # Default fallback
+        if hasattr(instance, "llm_config") and isinstance(instance.llm_config, dict):
+            request_model = instance.llm_config.get("model", "unknown")
+
+        # Try to get more specific model from groupchat
+        if groupchat and hasattr(groupchat, "select_speaker_auto_llm_config"):
+            llm_config = groupchat.select_speaker_auto_llm_config
+            if isinstance(llm_config, dict):
+                request_model = llm_config.get("model", request_model)
+            elif hasattr(llm_config, "model"):
+                request_model = llm_config.model
+
+        # Extract sender information
+        sender = kwargs.get("sender", None)
+
+        # Extract messages
+        messages = args[0] if args else kwargs.get("messages", [])
+
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_FRAMEWORK} {group_name}"
+
+        with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+            start_time = time.time()
+            response = wrapped(*args, **kwargs)
+
+            try:
+                process_groupchat_operation(
+                    group_name=group_name,
+                    participants=participants,
+                    messages=messages,
+                    sender=sender,
+                    max_turns=None,  # Not available in new API
+                    request_model=request_model,
+                    pricing_info=pricing_info,
+                    server_port=server_port,
+                    server_address=server_address,
+                    environment=environment,
+                    application_name=application_name,
+                    metrics=metrics,
+                    start_time=start_time,
+                    span=span,
+                    capture_message_content=capture_message_content,
+                    disable_metrics=disable_metrics,
+                    version=version,
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
+
+    return wrapper
+
+
+def groupchat_select_speaker(
+    version,
+    environment,
+    application_name,
+    tracer,
+    pricing_info,
+    capture_message_content,
+    metrics,
+    disable_metrics,
+):
+    """
+    Generates a telemetry wrapper for AG2 GroupChat.select_speaker.
+    """
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the AG2 GroupChat.select_speaker call.
+        """
+
+        server_address, server_port = set_server_address_and_port(
+            instance, "127.0.0.1", 80
+        )
+
+        # Extract speaker information
+        last_speaker = args[0] if args else kwargs.get("last_speaker", None)
+        selector = args[1] if len(args) > 1 else kwargs.get("selector", None)
+
+        last_speaker_name = (
+            getattr(last_speaker, "name", "Unknown") if last_speaker else "Unknown"
+        )
+
+        # Extract agents list
+        agents = getattr(instance, "agents", [])
+
+        # Extract model information from GroupChat instance
+        request_model = "unknown"  # Default fallback
+        # Check for speaker selection specific config
+        if hasattr(instance, "select_speaker_auto_llm_config"):
+            llm_config = instance.select_speaker_auto_llm_config
+            if isinstance(llm_config, dict):
+                request_model = llm_config.get("model", "unknown")
+            elif hasattr(llm_config, "model"):
+                request_model = llm_config.model
+
+        # Try to get model from selector if available
+        if (
+            selector
+            and hasattr(selector, "llm_config")
+            and isinstance(selector.llm_config, dict)
+        ):
+            request_model = selector.llm_config.get("model", request_model)
+
+        # Try to get model from agents if still unknown
+        if request_model == "unknown" and agents:
+            for agent in agents:
+                if hasattr(agent, "llm_config") and isinstance(agent.llm_config, dict):
+                    model = agent.llm_config.get("model")
+                    if model:
+                        request_model = model
+                        break
+
+        span_name = (
+            f"{SemanticConvention.GEN_AI_OPERATION_TYPE_AGENT} speaker_selection"
+        )
+
+        with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+            start_time = time.time()
+            response = wrapped(*args, **kwargs)
+
+            try:
+                selected_speaker_name = (
+                    getattr(response, "name", "Unknown") if response else "Unknown"
+                )
+
+                process_speaker_selection(
+                    last_speaker=last_speaker_name,
+                    selected_speaker=selected_speaker_name,
+                    selector=selector,
+                    agents=agents,
+                    request_model=request_model,
+                    pricing_info=pricing_info,
+                    server_port=server_port,
+                    server_address=server_address,
+                    environment=environment,
+                    application_name=application_name,
+                    metrics=metrics,
+                    start_time=start_time,
+                    span=span,
+                    capture_message_content=capture_message_content,
+                    disable_metrics=disable_metrics,
+                    version=version,
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
+
+    return wrapper