openlit 1.34.3__py3-none-any.whl → 1.34.4__py3-none-any.whl

openlit/instrumentation/ai21/__init__.py

@@ -13,53 +13,52 @@ from openlit.instrumentation.ai21.async_ai21 import (
     async_chat, async_chat_rag
 )
 
-_instruments = ('ai21 >= 3.0.0',)
+_instruments = ("ai21 >= 3.0.0",)
 
 class AI21Instrumentor(BaseInstrumentor):
     """
-    An instrumentor for AI21's client library.
+    An instrumentor for AI21 client library.
     """
 
     def instrumentation_dependencies(self) -> Collection[str]:
         return _instruments
 
     def _instrument(self, **kwargs):
-        application_name = kwargs.get('application_name', 'default')
-        environment = kwargs.get('environment', 'default')
-        tracer = kwargs.get('tracer')
-        event_provider = kwargs.get('event_provider')
-        metrics = kwargs.get('metrics_dict')
-        pricing_info = kwargs.get('pricing_info', {})
-        capture_message_content = kwargs.get('capture_message_content', False)
-        disable_metrics = kwargs.get('disable_metrics')
-        version = importlib.metadata.version('ai21')
+        application_name = kwargs.get("application_name", "default")
+        environment = kwargs.get("environment", "default")
+        tracer = kwargs.get("tracer")
+        metrics = kwargs.get("metrics_dict")
+        pricing_info = kwargs.get("pricing_info", {})
+        capture_message_content = kwargs.get("capture_message_content", False)
+        disable_metrics = kwargs.get("disable_metrics")
+        version = importlib.metadata.version("ai21")
 
         #sync
         wrap_function_wrapper(
-            'ai21.clients.studio.resources.chat.chat_completions',
-            'ChatCompletions.create',
+            "ai21.clients.studio.resources.chat.chat_completions",
+            "ChatCompletions.create",
             chat(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
         )
         wrap_function_wrapper(
-            'ai21.clients.studio.resources.studio_conversational_rag',
-            'StudioConversationalRag.create',
+            "ai21.clients.studio.resources.studio_conversational_rag",
+            "StudioConversationalRag.create",
             chat_rag(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
         )
 
         #Async
         wrap_function_wrapper(
-            'ai21.clients.studio.resources.chat.async_chat_completions',
-            'AsyncChatCompletions.create',
+            "ai21.clients.studio.resources.chat.async_chat_completions",
+            "AsyncChatCompletions.create",
             async_chat(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
         )
         wrap_function_wrapper(
-            'ai21.clients.studio.resources.studio_conversational_rag',
-            'AsyncStudioConversationalRag.create',
+            "ai21.clients.studio.resources.studio_conversational_rag",
+            "AsyncStudioConversationalRag.create",
             async_chat_rag(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
         )
 
     def _uninstrument(self, **kwargs):

openlit/instrumentation/ai21/ai21.py

@@ -22,7 +22,7 @@ from openlit.semcov import SemanticConvention
 logger = logging.getLogger(__name__)
 
 def chat(version, environment, application_name,
-                     tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics):
+                     tracer, pricing_info, capture_message_content, metrics, disable_metrics):
     """
     Generates a telemetry wrapper for GenAI function call
     """
@@ -46,9 +46,9 @@ def chat(version, environment, application_name,
             self._span = span
             self._span_name = span_name
             # Placeholder for aggregating streaming response
-            self._llmresponse = ''
-            self._response_id = ''
-            self._finish_reason = ''
+            self._llmresponse = ""
+            self._response_id = ""
+            self._finish_reason = ""
             self._input_tokens = 0
             self._output_tokens = 0
             self._choices = []
@@ -92,14 +92,13 @@ def chat(version, environment, application_name,
                             environment=environment,
                             application_name=application_name,
                             metrics=metrics,
-                            event_provider=event_provider,
                             capture_message_content=capture_message_content,
                             disable_metrics=disable_metrics,
                             version=version
                         )
                 except Exception as e:
                     handle_exception(self._span, e)
-                    logger.error('Error in trace creation: %s', e)
+                    logger.error("Error in trace creation: %s", e)
                 raise
 
     def wrapper(wrapped, instance, args, kwargs):
@@ -108,12 +107,12 @@ def chat(version, environment, application_name,
         """
 
         # Check if streaming is enabled for the API call
-        streaming = kwargs.get('stream', False)
+        streaming = kwargs.get("stream", False)
 
-        server_address, server_port = set_server_address_and_port(instance, 'api.ai21.com', 443)
-        request_model = kwargs.get('model', 'jamba-1.5-mini')
+        server_address, server_port = set_server_address_and_port(instance, "api.ai21.com", 443)
+        request_model = kwargs.get("model", "jamba-1.5-mini")
 
-        span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}'
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
         # pylint: disable=no-else-return
         if streaming:
@@ -127,30 +126,34 @@ def chat(version, environment, application_name,
             with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
                 start_time = time.time()
                 response = wrapped(*args, **kwargs)
-                response = process_chat_response(
-                    response=response,
-                    request_model=request_model,
-                    pricing_info=pricing_info,
-                    server_port=server_port,
-                    server_address=server_address,
-                    environment=environment,
-                    application_name=application_name,
-                    metrics=metrics,
-                    event_provider=event_provider,
-                    start_time=start_time,
-                    span=span,
-                    capture_message_content=capture_message_content,
-                    disable_metrics=disable_metrics,
-                    version=version,
-                    **kwargs
-                )
 
-            return response
+                try:
+                    response = process_chat_response(
+                        response=response,
+                        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,
+                        **kwargs
+                    )
+
+                except Exception as e:
+                    handle_exception(span, e)
+
+                return response
 
     return wrapper
 
 def chat_rag(version, environment, application_name,
-                tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics):
+                tracer, pricing_info, capture_message_content, metrics, disable_metrics):
     """
     Generates a telemetry wrapper for GenAI function call
     """
@@ -160,32 +163,36 @@ def chat_rag(version, environment, application_name,
         Wraps the GenAI function call.
         """
 
-        server_address, server_port = set_server_address_and_port(instance, 'api.ai21.com', 443)
-        request_model = kwargs.get('model', 'jamba-1.5-mini')
+        server_address, server_port = set_server_address_and_port(instance, "api.ai21.com", 443)
+        request_model = kwargs.get("model", "jamba-1.5-mini")
 
-        span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}'
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
         with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
             start_time = time.time()
             response = wrapped(*args, **kwargs)
-            response = process_chat_rag_response(
-                response=response,
-                request_model=request_model,
-                pricing_info=pricing_info,
-                server_port=server_port,
-                server_address=server_address,
-                environment=environment,
-                application_name=application_name,
-                metrics=metrics,
-                event_provider=event_provider,
-                start_time=start_time,
-                span=span,
-                capture_message_content=capture_message_content,
-                disable_metrics=disable_metrics,
-                version=version,
-                **kwargs
-            )
-
-        return response
+
+            try:
+                response = process_chat_rag_response(
+                    response=response,
+                    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,
+                    **kwargs
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
 
     return wrapper

openlit/instrumentation/ai21/async_ai21.py

@@ -22,7 +22,7 @@ from openlit.semcov import SemanticConvention
 logger = logging.getLogger(__name__)
 
 def async_chat(version, environment, application_name,
-                     tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics):
+                     tracer, pricing_info, capture_message_content, metrics, disable_metrics):
     """
     Generates a telemetry wrapper for GenAI function call
     """
@@ -46,9 +46,9 @@ def async_chat(version, environment, application_name,
             self._span = span
             self._span_name = span_name
             # Placeholder for aggregating streaming response
-            self._llmresponse = ''
-            self._response_id = ''
-            self._finish_reason = ''
+            self._llmresponse = ""
+            self._response_id = ""
+            self._finish_reason = ""
             self._input_tokens = 0
             self._output_tokens = 0
             self._choices = []
@@ -92,14 +92,13 @@ def async_chat(version, environment, application_name,
                             environment=environment,
                             application_name=application_name,
                             metrics=metrics,
-                            event_provider=event_provider,
                             capture_message_content=capture_message_content,
                             disable_metrics=disable_metrics,
                             version=version
                         )
                 except Exception as e:
                     handle_exception(self._span, e)
-                    logger.error('Error in trace creation: %s', e)
+
                 raise
 
     async def wrapper(wrapped, instance, args, kwargs):
@@ -108,12 +107,12 @@ def async_chat(version, environment, application_name,
         """
 
         # Check if streaming is enabled for the API call
-        streaming = kwargs.get('stream', False)
+        streaming = kwargs.get("stream", False)
 
-        server_address, server_port = set_server_address_and_port(instance, 'api.ai21.com', 443)
-        request_model = kwargs.get('model', 'jamba-1.5-mini')
+        server_address, server_port = set_server_address_and_port(instance, "api.ai21.com", 443)
+        request_model = kwargs.get("model", "jamba-1.5-mini")
 
-        span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}'
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
         # pylint: disable=no-else-return
         if streaming:
@@ -127,30 +126,34 @@ def async_chat(version, environment, application_name,
             with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
                 start_time = time.time()
                 response = await wrapped(*args, **kwargs)
-                response = process_chat_response(
-                    response=response,
-                    request_model=request_model,
-                    pricing_info=pricing_info,
-                    server_port=server_port,
-                    server_address=server_address,
-                    environment=environment,
-                    application_name=application_name,
-                    metrics=metrics,
-                    event_provider=event_provider,
-                    start_time=start_time,
-                    span=span,
-                    capture_message_content=capture_message_content,
-                    disable_metrics=disable_metrics,
-                    version=version,
-                    **kwargs
-                )
 
-            return response
+                try:
+                    response = process_chat_response(
+                        response=response,
+                        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,
+                        **kwargs
+                    )
+
+                except Exception as e:
+                    handle_exception(span, e)
+
+                return response
 
     return wrapper
 
 def async_chat_rag(version, environment, application_name,
-                tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics):
+                tracer, pricing_info, capture_message_content, metrics, disable_metrics):
     """
     Generates a telemetry wrapper for GenAI function call
     """
@@ -160,32 +163,36 @@ def async_chat_rag(version, environment, application_name,
         Wraps the GenAI function call.
         """
 
-        server_address, server_port = set_server_address_and_port(instance, 'api.ai21.com', 443)
-        request_model = kwargs.get('model', 'jamba-1.5-mini')
+        server_address, server_port = set_server_address_and_port(instance, "api.ai21.com", 443)
+        request_model = kwargs.get("model", "jamba-1.5-mini")
 
-        span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}'
+        span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
         with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
             start_time = time.time()
             response = await wrapped(*args, **kwargs)
-            response = process_chat_rag_response(
-                response=response,
-                request_model=request_model,
-                pricing_info=pricing_info,
-                server_port=server_port,
-                server_address=server_address,
-                environment=environment,
-                application_name=application_name,
-                metrics=metrics,
-                event_provider=event_provider,
-                start_time=start_time,
-                span=span,
-                capture_message_content=capture_message_content,
-                disable_metrics=disable_metrics,
-                version=version,
-                **kwargs
-            )
-
-        return response
+
+            try:
+                response = process_chat_rag_response(
+                    response=response,
+                    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,
+                    **kwargs
+                )
+
+            except Exception as e:
+                handle_exception(span, e)
+
+            return response
 
     return wrapper

openlit/instrumentation/ai21/utils.py

@@ -14,9 +14,7 @@ from openlit.__helpers import (
     general_tokens,
     extract_and_format_input,
     get_chat_model_cost,
-    handle_exception,
     create_metrics_attributes,
-    otel_event,
     concatenate_all_contents
 )
 from openlit.semcov import SemanticConvention
@@ -29,36 +27,38 @@ def setup_common_span_attributes(span, request_model, kwargs, tokens,
     """
 
     # Base attributes from SDK and operation settings.
-    span.set_attribute(TELEMETRY_SDK_NAME, 'openlit')
+    span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
     span.set_attribute(SemanticConvention.GEN_AI_OPERATION, SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT)
     span.set_attribute(SemanticConvention.GEN_AI_SYSTEM, SemanticConvention.GEN_AI_SYSTEM_AI21)
     span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MODEL, request_model)
     span.set_attribute(SemanticConvention.SERVER_PORT, server_port)
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_SEED, kwargs.get('seed', ''))
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_FREQUENCY_PENALTY, kwargs.get('frequency_penalty', 0.0))
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, kwargs.get('max_tokens', -1))
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_PRESENCE_PENALTY, kwargs.get('presence_penalty', 0.0))
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_STOP_SEQUENCES, kwargs.get('stop', []))
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, kwargs.get('temperature', 0.4))
-    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_P, kwargs.get('top_p', 1.0))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_SEED, kwargs.get("seed", ""))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_FREQUENCY_PENALTY, kwargs.get("frequency_penalty", 0.0))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, kwargs.get("max_tokens", -1))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_PRESENCE_PENALTY, kwargs.get("presence_penalty", 0.0))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_STOP_SEQUENCES, kwargs.get("stop", []))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, kwargs.get("temperature", 0.4))
+    span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_P, kwargs.get("top_p", 1.0))
 
     # Add token-related attributes if available.
-    if 'finish_reason' in tokens:
-        span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_FINISH_REASON, [tokens['finish_reason']])
-    if 'response_id' in tokens:
-        span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_ID, tokens['response_id'])
-    if 'input_tokens' in tokens:
-        span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS, tokens['input_tokens'])
-    if 'output_tokens' in tokens:
-        span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, tokens['output_tokens'])
-    if 'total_tokens' in tokens:
-        span.set_attribute(SemanticConvention.GEN_AI_USAGE_TOTAL_TOKENS, tokens['total_tokens'])
+    if "finish_reason" in tokens:
+        span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_FINISH_REASON, [tokens["finish_reason"]])
+    if "response_id" in tokens:
+        span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_ID, tokens["response_id"])
+    if "input_tokens" in tokens:
+        span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS, tokens["input_tokens"])
+    if "output_tokens" in tokens:
+        span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, tokens["output_tokens"])
+    if "total_tokens" in tokens:
+        span.set_attribute(SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE, tokens["total_tokens"])
 
     span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_MODEL, request_model)
     span.set_attribute(SemanticConvention.SERVER_ADDRESS, server_address)
+
     # Environment and service identifiers.
     span.set_attribute(DEPLOYMENT_ENVIRONMENT, environment)
     span.set_attribute(SERVICE_NAME, application_name)
+
     # Set any extra attributes passed in.
     for key, value in extra_attrs.items():
         span.set_attribute(key, value)
@@ -80,106 +80,15 @@ def record_common_metrics(metrics, application_name, environment, request_model,
         server_port=server_port,
         response_model=request_model,
     )
-    metrics['genai_client_usage_tokens'].record(input_tokens + output_tokens, attributes)
-    metrics['genai_client_operation_duration'].record(end_time - start_time, attributes)
+    metrics["genai_client_usage_tokens"].record(input_tokens + output_tokens, attributes)
+    metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
     if include_tbt and tbt_value is not None:
-        metrics['genai_server_tbt'].record(tbt_value, attributes)
-    metrics['genai_server_ttft'].record(end_time - start_time, attributes)
-    metrics['genai_requests'].add(1, attributes)
-    metrics['genai_completion_tokens'].add(output_tokens, attributes)
-    metrics['genai_prompt_tokens'].add(input_tokens, attributes)
-    metrics['genai_cost'].record(cost, attributes)
-
-def emit_common_events(event_provider, choices, finish_reason, llmresponse, formatted_messages,
-                       capture_message_content, n):
-    """
-    Emit events common to both chat and chat rag operations.
-    """
-
-    if n > 1:
-        for choice in choices:
-            choice_event_body = {
-                'finish_reason': finish_reason,
-                'index': choice.get('index', 0),
-                'message': {
-                    **({'content': choice.get('message', {}).get('content', '')} if capture_message_content else {}),
-                    'role': choice.get('message', {}).get('role', 'assistant')
-                }
-            }
-            # If tool calls exist, emit an event for each tool call.
-            tool_calls = choice.get('message', {}).get('tool_calls')
-            if tool_calls:
-                for tool_call in tool_calls:
-                    choice_event_body['message'].update({
-                        'tool_calls': {
-                            'function': {
-                                'name': tool_call.get('function', {}).get('name', ''),
-                                'arguments': tool_call.get('function', {}).get('arguments', '')
-                            },
-                            'id': tool_call.get('id', ''),
-                            'type': tool_call.get('type', 'function')
-                        }
-                    })
-                    event = otel_event(
-                        name=SemanticConvention.GEN_AI_CHOICE,
-                        attributes={SemanticConvention.GEN_AI_SYSTEM: SemanticConvention.GEN_AI_SYSTEM_AI21},
-                        body=choice_event_body
-                    )
-                    event_provider.emit(event)
-            else:
-                event = otel_event(
-                    name=SemanticConvention.GEN_AI_CHOICE,
-                    attributes={SemanticConvention.GEN_AI_SYSTEM: SemanticConvention.GEN_AI_SYSTEM_AI21},
-                    body=choice_event_body
-                )
-                event_provider.emit(event)
-    else:
-        # Single choice case.
-        choice_event_body = {
-            'finish_reason': finish_reason,
-            'index': 0,
-            'message': {
-                **({'content': llmresponse} if capture_message_content else {}),
-                'role': 'assistant'
-            }
-        }
-        event = otel_event(
-            name=SemanticConvention.GEN_AI_CHOICE,
-            attributes={SemanticConvention.GEN_AI_SYSTEM: SemanticConvention.GEN_AI_SYSTEM_AI21},
-            body=choice_event_body
-        )
-        event_provider.emit(event)
-
-    # Emit additional role-based events (if formatted messages are available).
-    for role in ['user', 'system', 'assistant', 'tool']:
-        msg = formatted_messages.get(role, {})
-        if msg.get('content', ''):
-            event_body = {
-                **({'content': msg.get('content', '')} if capture_message_content else {}),
-                'role': msg.get('role', [])
-            }
-            # For assistant messages, attach tool call details if they exist.
-            if role == 'assistant' and choices:
-                tool_calls = choices[0].get('message', {}).get('tool_calls', [])
-                if tool_calls:
-                    event_body['tool_calls'] = {
-                        'function': {
-                            'name': tool_calls[0].get('function', {}).get('name', ''),
-                            'arguments': tool_calls[0].get('function', {}).get('arguments', '')
-                        },
-                        'id': tool_calls[0].get('id', ''),
-                        'type': 'function'
-                    }
-            if role == 'tool' and choices:
-                tool_calls = choices[0].get('message', {}).get('tool_calls', [])
-                if tool_calls:
-                    event_body['id'] = tool_calls[0].get('id', '')
-            event = otel_event(
-                name=getattr(SemanticConvention, f'GEN_AI_{role.upper()}_MESSAGE'),
-                attributes={SemanticConvention.GEN_AI_SYSTEM: SemanticConvention.GEN_AI_SYSTEM_AI21},
-                body=event_body
-            )
-            event_provider.emit(event)
+        metrics["genai_server_tbt"].record(tbt_value, attributes)
+    metrics["genai_server_ttft"].record(end_time - start_time, attributes)
+    metrics["genai_requests"].add(1, attributes)
+    metrics["genai_completion_tokens"].add(output_tokens, attributes)
+    metrics["genai_prompt_tokens"].add(input_tokens, attributes)
+    metrics["genai_cost"].record(cost, attributes)
 
 def process_chunk(self, chunk):
     """
@@ -194,21 +103,20 @@ def process_chunk(self, chunk):
         self._ttft = calculate_ttft(self._timestamps, self._start_time)
 
     chunked = response_as_dict(chunk)
-    if (len(chunked.get('choices')) > 0 and
-            'delta' in chunked.get('choices')[0] and
-            'content' in chunked.get('choices')[0].get('delta')):
-        content = chunked.get('choices')[0].get('delta').get('content')
-        if content:
+    if (len(chunked.get("choices")) > 0 and
+            "delta" in chunked.get("choices")[0] and
+            "content" in chunked.get("choices")[0].get("delta")):
+        if content := chunked.get("choices")[0].get("delta").get("content"):
             self._llmresponse += content
-        if chunked.get('usage'):
-            self._input_tokens = chunked.get('usage').get('prompt_tokens')
-            self._output_tokens = chunked.get('usage').get('completion_tokens')
-    self._response_id = chunked.get('id')
-    self._choices += chunked.get('choices')
-    self._finish_reason = chunked.get('choices')[0].get('finish_reason')
+        if chunked.get("usage"):
+            self._input_tokens = chunked.get("usage").get("prompt_tokens")
+            self._output_tokens = chunked.get("usage").get("completion_tokens")
+    self._response_id = chunked.get("id")
+    self._choices += chunked.get("choices")
+    self._finish_reason = chunked.get("choices")[0].get("finish_reason")
 
 def common_chat_logic(scope, pricing_info, environment, application_name, metrics,
-                      event_provider, capture_message_content, disable_metrics, version, is_stream):
+    capture_message_content, disable_metrics, version, is_stream):
     """
     Process chat request and generate Telemetry.
     """
@@ -218,19 +126,19 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
         scope._tbt = calculate_tbt(scope._timestamps)
 
     # Extract and format input messages.
-    formatted_messages = extract_and_format_input(scope._kwargs.get('messages', ''))
+    formatted_messages = extract_and_format_input(scope._kwargs.get("messages", ""))
     prompt = concatenate_all_contents(formatted_messages)
-    request_model = scope._kwargs.get('model', 'jamba-1.5-mini')
+    request_model = scope._kwargs.get("model", "jamba-1.5-mini")
 
     # Calculate cost based on token usage.
     cost = get_chat_model_cost(request_model, pricing_info, scope._input_tokens, scope._output_tokens)
     # Prepare tokens dictionary.
     tokens = {
-        'finish_reason': scope._finish_reason,
-        'response_id': scope._response_id,
-        'input_tokens': scope._input_tokens,
-        'output_tokens': scope._output_tokens,
-        'total_tokens': scope._input_tokens + scope._output_tokens,
+        "finish_reason": scope._finish_reason,
+        "response_id": scope._response_id,
+        "input_tokens": scope._input_tokens,
+        "output_tokens": scope._output_tokens,
+        "total_tokens": scope._input_tokens + scope._output_tokens,
     }
     extra_attrs = {
         SemanticConvention.GEN_AI_REQUEST_IS_STREAM: is_stream,
@@ -239,14 +147,13 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
         SemanticConvention.GEN_AI_SERVER_TBT: scope._tbt,
         SemanticConvention.GEN_AI_SERVER_TTFT: scope._ttft,
         SemanticConvention.GEN_AI_SDK_VERSION: version,
-        SemanticConvention.GEN_AI_OUTPUT_TYPE: 'text' if isinstance(scope._llmresponse, str) else 'json'
+        SemanticConvention.GEN_AI_OUTPUT_TYPE: "text" if isinstance(scope._llmresponse, str) else "json"
     }
     # Set span attributes.
     setup_common_span_attributes(scope._span, request_model, scope._kwargs, tokens,
                                  scope._server_port, scope._server_address, environment,
                                  application_name, extra_attrs)
 
-    # Optionally add events capturing the prompt and completion.
     if capture_message_content:
         scope._span.add_event(
             name=SemanticConvention.GEN_AI_CONTENT_PROMPT_EVENT,
@@ -257,11 +164,6 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
             attributes={SemanticConvention.GEN_AI_CONTENT_COMPLETION: scope._llmresponse},
         )
 
-    # Emit events for each choice and message role.
-    n = scope._kwargs.get('n', 1)
-    emit_common_events(event_provider, scope._choices, scope._finish_reason, scope._llmresponse,
-                       formatted_messages, capture_message_content, n)
-
     scope._span.set_status(Status(StatusCode.OK))
 
     if not disable_metrics:
@@ -272,23 +174,23 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
                               include_tbt=True, tbt_value=scope._tbt)
 
 def process_streaming_chat_response(self, pricing_info, environment, application_name, metrics,
-                                    event_provider, capture_message_content=False, disable_metrics=False, version=''):
+    capture_message_content=False, disable_metrics=False, version=""):
     """
     Process a streaming chat response and generate Telemetry.
     """
 
     common_chat_logic(self, pricing_info, environment, application_name, metrics,
-                      event_provider, capture_message_content, disable_metrics, version, is_stream=True)
+        capture_message_content, disable_metrics, version, is_stream=True)
 
 def process_chat_response(response, request_model, pricing_info, server_port, server_address,
-                          environment, application_name, metrics, event_provider, start_time,
-                          span, capture_message_content=False, disable_metrics=False, version='1.0.0', **kwargs):
+                          environment, application_name, metrics, start_time,
+                          span, capture_message_content=False, disable_metrics=False, version="1.0.0", **kwargs):
     """
     Process a synchronous chat response and generate Telemetry.
     """
 
     # Create a generic scope object to hold telemetry data.
-    self = type('GenericScope', (), {})()
+    self = type("GenericScope", (), {})()
     response_dict = response_as_dict(response)
 
     # pylint: disable = no-member
@@ -297,113 +199,102 @@ def process_chat_response(response, request_model, pricing_info, server_port, se
 
     self._span = span
     # Concatenate content from all choices.
-    self._llmresponse = ''.join(
-        (choice.get('message', {}).get('content') or '')
-        for choice in response_dict.get('choices', [])
+    self._llmresponse = "".join(
+        (choice.get("message", {}).get("content") or "")
+        for choice in response_dict.get("choices", [])
     )
-    self._response_role = response_dict.get('message', {}).get('role', 'assistant')
-    self._input_tokens = response_dict.get('usage', {}).get('prompt_tokens', 0)
-    self._output_tokens = response_dict.get('usage', {}).get('completion_tokens', 0)
-    self._response_id = response_dict.get('id', '')
+    self._response_role = response_dict.get("message", {}).get("role", "assistant")
+    self._input_tokens = response_dict.get("usage", {}).get("prompt_tokens", 0)
+    self._output_tokens = response_dict.get("usage", {}).get("completion_tokens", 0)
+    self._response_id = response_dict.get("id", "")
     self._response_model = request_model
-    self._finish_reason = response_dict.get('choices', [{}])[0].get('finish_reason')
+    self._finish_reason = response_dict.get("choices", [{}])[0].get("finish_reason")
     self._timestamps = []
     self._ttft, self._tbt = self._end_time - self._start_time, 0
     self._server_address, self._server_port = server_address, server_port
     self._kwargs = kwargs
-    self._choices = response_dict.get('choices')
+    self._choices = response_dict.get("choices")
 
     common_chat_logic(self, pricing_info, environment, application_name, metrics,
-                      event_provider, capture_message_content, disable_metrics, version, is_stream=False)
+        capture_message_content, disable_metrics, version, is_stream=False)
 
     return response
 
 def process_chat_rag_response(response, request_model, pricing_info, server_port, server_address,
-                              environment, application_name, metrics, event_provider, start_time,
-                              span, capture_message_content=False, disable_metrics=False, version='1.0.0', **kwargs):
+                              environment, application_name, metrics, start_time,
+                              span, capture_message_content=False, disable_metrics=False, version="1.0.0", **kwargs):
     """
     Process a chat response and generate Telemetry.
     """
     end_time = time.time()
     response_dict = response_as_dict(response)
-    try:
-        # Format input messages into a single prompt string.
-        messages_input = kwargs.get('messages', '')
-        formatted_messages = extract_and_format_input(messages_input)
-        prompt = concatenate_all_contents(formatted_messages)
-        input_tokens = general_tokens(prompt)
-
-        # Create tokens dict and RAG-specific extra attributes.
-        tokens = {'response_id': response_dict.get('id'), 'input_tokens': input_tokens}
-        extra_attrs = {
-            SemanticConvention.GEN_AI_REQUEST_IS_STREAM: False,
-            SemanticConvention.GEN_AI_SERVER_TTFT: end_time - start_time,
-            SemanticConvention.GEN_AI_SDK_VERSION: version,
-            SemanticConvention.GEN_AI_RAG_MAX_SEGMENTS: kwargs.get('max_segments', -1),
-            SemanticConvention.GEN_AI_RAG_STRATEGY: kwargs.get('retrieval_strategy', 'segments'),
-            SemanticConvention.GEN_AI_RAG_SIMILARITY_THRESHOLD: kwargs.get('retrieval_similarity_threshold', -1),
-            SemanticConvention.GEN_AI_RAG_MAX_NEIGHBORS: kwargs.get('max_neighbors', -1),
-            SemanticConvention.GEN_AI_RAG_FILE_IDS: str(kwargs.get('file_ids', '')),
-            SemanticConvention.GEN_AI_RAG_DOCUMENTS_PATH: kwargs.get('path', '')
-        }
-        # Set common span attributes.
-        setup_common_span_attributes(span, request_model, kwargs, tokens,
-                                     server_port, server_address, environment, application_name,
-                                     extra_attrs)
-
-        # Record the prompt event if requested.
-        if capture_message_content:
-            span.add_event(
-                name=SemanticConvention.GEN_AI_CONTENT_PROMPT_EVENT,
-                attributes={SemanticConvention.GEN_AI_CONTENT_PROMPT: prompt},
-            )
-
-        output_tokens = 0
-        choices = response_dict.get('choices', [])
-        # Instead of adding a separate event per choice, we aggregate all completion content.
-        aggregated_completion = []
-        for i in range(kwargs.get('n', 1)):
-            # Get the response content from each choice and count tokens.
-            content = choices[i].get('content', '')
-            aggregated_completion.append(content)
-            output_tokens += general_tokens(content)
-            if kwargs.get('tools'):
-                span.set_attribute(SemanticConvention.GEN_AI_TOOL_CALLS,
-                                   str(choices[i].get('message', {}).get('tool_calls')))
-            # Set output type based on actual content type.
-            if isinstance(content, str):
-                span.set_attribute(SemanticConvention.GEN_AI_OUTPUT_TYPE, 'text')
-            elif content is not None:
-                span.set_attribute(SemanticConvention.GEN_AI_OUTPUT_TYPE, 'json')
-
-        # Concatenate completion responses.
-        llmresponse = ''.join(aggregated_completion)
-        tokens['output_tokens'] = output_tokens
-        tokens['total_tokens'] = input_tokens + output_tokens
-
-        cost = get_chat_model_cost(request_model, pricing_info, input_tokens, output_tokens)
-        span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
-        span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, output_tokens)
-        span.set_attribute(SemanticConvention.GEN_AI_USAGE_TOTAL_TOKENS, input_tokens + output_tokens)
-
-        span.set_status(Status(StatusCode.OK))
-        # Emit a single aggregated completion event.
-        if capture_message_content:
-            span.add_event(
-                name=SemanticConvention.GEN_AI_CONTENT_COMPLETION_EVENT,
-                attributes={SemanticConvention.GEN_AI_CONTENT_COMPLETION: llmresponse},
-            )
-        # Emit the rest of the events (choice and role-based events) as before.
-        n = kwargs.get('n', 1)
-        emit_common_events(event_provider, choices, choices[0].get('finish_reason', ''),
-                           llmresponse, formatted_messages, capture_message_content, n)
-
-        if not disable_metrics:
-            record_common_metrics(metrics, application_name, environment, request_model,
-                                  server_address, server_port, start_time, end_time,
-                                  input_tokens, output_tokens, cost, include_tbt=False)
-        return response
-
-    except Exception as e:
-        handle_exception(span, e)
-        return response
+    # Format input messages into a single prompt string.
+    messages_input = kwargs.get("messages", "")
+    formatted_messages = extract_and_format_input(messages_input)
+    prompt = concatenate_all_contents(formatted_messages)
+    input_tokens = general_tokens(prompt)
+
+    # Create tokens dict and RAG-specific extra attributes.
+    tokens = {"response_id": response_dict.get("id"), "input_tokens": input_tokens}
+    extra_attrs = {
+        SemanticConvention.GEN_AI_REQUEST_IS_STREAM: False,
+        SemanticConvention.GEN_AI_SERVER_TTFT: end_time - start_time,
+        SemanticConvention.GEN_AI_SDK_VERSION: version,
+        SemanticConvention.GEN_AI_RAG_MAX_SEGMENTS: kwargs.get("max_segments", -1),
+        SemanticConvention.GEN_AI_RAG_STRATEGY: kwargs.get("retrieval_strategy", "segments"),
+        SemanticConvention.GEN_AI_RAG_SIMILARITY_THRESHOLD: kwargs.get("retrieval_similarity_threshold", -1),
+        SemanticConvention.GEN_AI_RAG_MAX_NEIGHBORS: kwargs.get("max_neighbors", -1),
+        SemanticConvention.GEN_AI_RAG_FILE_IDS: str(kwargs.get("file_ids", "")),
+        SemanticConvention.GEN_AI_RAG_DOCUMENTS_PATH: kwargs.get("path", "")
+    }
+    # Set common span attributes.
+    setup_common_span_attributes(span, request_model, kwargs, tokens,
+                                    server_port, server_address, environment, application_name,
+                                    extra_attrs)
+
+    if capture_message_content:
+        span.add_event(
+            name=SemanticConvention.GEN_AI_CONTENT_PROMPT_EVENT,
+            attributes={SemanticConvention.GEN_AI_CONTENT_PROMPT: prompt},
+        )
+
+    output_tokens = 0
+    choices = response_dict.get("choices", [])
+    aggregated_completion = []
+    for i in range(kwargs.get("n", 1)):
+        # Get the response content from each choice and count tokens.
+        content = choices[i].get("content", "")
+        aggregated_completion.append(content)
+        output_tokens += general_tokens(content)
+        if kwargs.get("tools"):
+            span.set_attribute(SemanticConvention.GEN_AI_TOOL_CALLS,
+                                str(choices[i].get("message", {}).get("tool_calls")))
+        # Set output type based on actual content type.
+        if isinstance(content, str):
+            span.set_attribute(SemanticConvention.GEN_AI_OUTPUT_TYPE, "text")
+        elif content is not None:
+            span.set_attribute(SemanticConvention.GEN_AI_OUTPUT_TYPE, "json")
+
+    # Concatenate completion responses.
+    llmresponse = "".join(aggregated_completion)
+    tokens["output_tokens"] = output_tokens
+    tokens["total_tokens"] = input_tokens + output_tokens
+
+    cost = get_chat_model_cost(request_model, pricing_info, input_tokens, output_tokens)
+    span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
+    span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, output_tokens)
+    span.set_attribute(SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE, input_tokens + output_tokens)
+
+    span.set_status(Status(StatusCode.OK))
+
+    if capture_message_content:
+        span.add_event(
+            name=SemanticConvention.GEN_AI_CONTENT_COMPLETION_EVENT,
+            attributes={SemanticConvention.GEN_AI_CONTENT_COMPLETION: llmresponse},
+        )
+
+    if not disable_metrics:
+        record_common_metrics(metrics, application_name, environment, request_model,
+                                server_address, server_port, start_time, end_time,
+                                input_tokens, output_tokens, cost, include_tbt=False)
+    return response

{openlit-1.34.3.dist-info → openlit-1.34.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: openlit
-Version: 1.34.3
+Version: 1.34.4
 Summary: OpenTelemetry-native Auto instrumentation library for monitoring LLM Applications and GPUs, facilitating the integration of observability into your GenAI-driven projects
 License: Apache-2.0
 Keywords: OpenTelemetry,otel,otlp,llm,tracing,openai,anthropic,claude,cohere,llm monitoring,observability,monitoring,gpt,Generative AI,chatGPT,gpu

{openlit-1.34.3.dist-info → openlit-1.34.4.dist-info}/RECORD

@@ -14,10 +14,10 @@ openlit/guard/sensitive_topic.py,sha256=RgVw_laFERv0nNdzBsAd2_3yLomMOK-gVq-P7oj1
 openlit/guard/utils.py,sha256=6hE3rCRjFXYjKRQYUo8YsqUSlvod48nOWp8MwoQEYdw,7670
 openlit/instrumentation/ag2/__init__.py,sha256=KgyLJBmwAxRWu7Z0S8FDDK4TZ13EFoAAIalvG5Oq4wc,1839
 openlit/instrumentation/ag2/ag2.py,sha256=eNQziyeZl4396GsIp5qI1Dne2KcnQMmhftW7joKQvNU,6934
-openlit/instrumentation/ai21/__init__.py,sha256=QXMByKCUhFITUIwUR01m0Fjpr20txV_GWcRJ66dTu_Q,2703
-openlit/instrumentation/ai21/ai21.py,sha256=J1QlBg56LWhkiD9IDmu0tJMOGnqKz0AiHtrMnACjU0Y,6814
-openlit/instrumentation/ai21/async_ai21.py,sha256=fIIGrAjF2xfGlAJiW6itqL88PSaA9vxy7F8nFpyVjZg,6918
-openlit/instrumentation/ai21/utils.py,sha256=7PcxzAwrLNTF5df5SR0Mp6UAkLGTATWxYrKS98c9iig,19770
+openlit/instrumentation/ai21/__init__.py,sha256=U24XlK1aHX0zubyUyBY6PBCa59fwp5sU5f-VD1EkCjc,2583
+openlit/instrumentation/ai21/ai21.py,sha256=1fJ1MvVIRQG-gh5YXkDycuTriT7_VB77vjXDKb7GZY8,6965
+openlit/instrumentation/ai21/async_ai21.py,sha256=uUJUXCKJcokYi6fPfcgBABSMVLj9CQsWJySakiZcSiU,7003
+openlit/instrumentation/ai21/utils.py,sha256=TiJtzG6kcrTf4FMJkAeHYUwZxkjp6JS3xoM2qn4gw54,14215
 openlit/instrumentation/anthropic/__init__.py,sha256=QEsiwdxcQDzzlVYR4_x7KTdf0-UJDJt8FjwNQMspnxM,1929
 openlit/instrumentation/anthropic/anthropic.py,sha256=NxJJjhsu9sSFIlBp322olGkPlLt9Bn5sndaugYA68dE,5149
 openlit/instrumentation/anthropic/async_anthropic.py,sha256=ivJGygKWVTS2hWWX12_g1tiq-5mpeHXETZsWoFZL3UE,5235
@@ -131,7 +131,7 @@ openlit/otel/events.py,sha256=VrMjTpvnLtYRBHCiFwJojTQqqNpRCxoD4yJYeQrtPsk,3560
 openlit/otel/metrics.py,sha256=GM2PDloBGRhBTkHHkYaqmOwIAQkY124ZhW4sEqW1Fgk,7086
 openlit/otel/tracing.py,sha256=tjV2bEbEDPUB1Z46gE-UsJsb04sRdFrfbhIDkxViZc0,3103
 openlit/semcov/__init__.py,sha256=ptyo37PY-FHDx_PShEvbdns71cD4YvvXw15bCRXKCKM,13461
-openlit-1.34.3.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-openlit-1.34.3.dist-info/METADATA,sha256=8_jDnUBC1cxAr2DNwkg5IXbNQX2qru-_nC7OpwC6Jh8,23469
-openlit-1.34.3.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
-openlit-1.34.3.dist-info/RECORD,,
+openlit-1.34.4.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+openlit-1.34.4.dist-info/METADATA,sha256=0EfABzey6V_Rxze1-maNqQN1Y9Vftu54Ix0-5t7vaNo,23469
+openlit-1.34.4.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+openlit-1.34.4.dist-info/RECORD,,