Skip to content

Same agent Smaller bill. The tools and infrastructure to get more work from your AI. Fewer tokens. Less compute. Lower cost.

[Feat]: Update AssemblyAI, Bedrock, Gemini, ElevenLabs and Azure AI Inference Instrumentation for compatibility with OTel Semconv (#620)

Patcher2f5de0a279ba
diff --git a/sdk/python/.pylintrc b/sdk/python/.pylintrc
index 9542de74..fe2409c9 100644
--- a/sdk/python/.pylintrc
+++ b/sdk/python/.pylintrc
@@ -24,4 +24,6 @@ disable=
unused-argument,
too-many-branches,
too-many-locals,
- too-many-lines
\ No newline at end of file
+ too-many-lines,
+ no-name-in-module,
+ protected-access,
\ No newline at end of file
diff --git a/sdk/python/pyproject.toml b/sdk/python/pyproject.toml
index 9e20da8b..dde0383f 100644
--- a/sdk/python/pyproject.toml
+++ b/sdk/python/pyproject.toml
@@ -1,6 +1,6 @@
[tool.poetry]
name = "openlit"
-version = "1.33.8"
+version = "1.33.9"
description = "OpenTelemetry-native Auto instrumentation library for monitoring LLM Applications and GPUs, facilitating the integration of observability into your GenAI-driven projects"
authors = ["OpenLIT"]
license = "Apache-2.0"
diff --git a/sdk/python/src/openlit/__helpers.py b/sdk/python/src/openlit/__helpers.py
index 7ae739b5..01a12c6c 100644
--- a/sdk/python/src/openlit/__helpers.py
+++ b/sdk/python/src/openlit/__helpers.py
@@ -257,9 +257,15 @@ def set_server_address_and_port(client_instance: Any,
using defaults if none found or values are None.
"""
+ # Try getting base_url from multiple potential attributes
base_client = getattr(client_instance, "_client", None)
base_url = getattr(base_client, "base_url", None)
+ if not base_url:
+ # Attempt to get endpoint from instance._config.endpoint if base_url is not set
+ config = getattr(client_instance, "_config", None)
+ base_url = getattr(config, "endpoint", None)
+
if base_url:
if isinstance(base_url, str):
url = urlparse(base_url)
@@ -269,7 +275,7 @@ def set_server_address_and_port(client_instance: Any,
server_address = getattr(base_url, "host", None) or default_server_address
port_attr = getattr(base_url, "port", None)
server_port = port_attr if port_attr is not None else default_server_port
- else: # no base_url provided; use defaults.
+ else: # no base_url or endpoint provided; use defaults.
server_address = default_server_address
server_port = default_server_port
diff --git a/sdk/python/src/openlit/__init__.py b/sdk/python/src/openlit/__init__.py
index f14fdb6c..8bba065f 100644
--- a/sdk/python/src/openlit/__init__.py
+++ b/sdk/python/src/openlit/__init__.py
@@ -251,7 +251,7 @@ def init(
"gpt4all": "gpt4all",
"elevenlabs": "elevenlabs",
"vllm": "vllm",
- "google-ai-studio": "google.generativeai",
+ "google-ai-studio": "google.genai",
"azure-ai-inference": "azure.ai.inference",
"langchain": "langchain",
"llama_index": "llama_index",
diff --git a/sdk/python/src/openlit/instrumentation/assemblyai/__init__.py b/sdk/python/src/openlit/instrumentation/assemblyai/__init__.py
index 897c3e60..2280414f 100644
--- a/sdk/python/src/openlit/instrumentation/assemblyai/__init__.py
+++ b/sdk/python/src/openlit/instrumentation/assemblyai/__init__.py
@@ -34,7 +34,7 @@ class AssemblyAIInstrumentor(BaseInstrumentor):
wrap_function_wrapper(
"assemblyai.transcriber",
"Transcriber.transcribe",
- transcribe("assemblyai.transcribe", version, environment, application_name,
+ transcribe(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
diff --git a/sdk/python/src/openlit/instrumentation/assemblyai/assemblyai.py b/sdk/python/src/openlit/instrumentation/assemblyai/assemblyai.py
index 2bdfdf96..62db15c0 100644
--- a/sdk/python/src/openlit/instrumentation/assemblyai/assemblyai.py
+++ b/sdk/python/src/openlit/instrumentation/assemblyai/assemblyai.py
@@ -1,19 +1,23 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment
"""
Module for monitoring Assembly AI API calls.
"""
import logging
+import time
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
-from openlit.__helpers import get_audio_model_cost
-from openlit.__helpers import handle_exception
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
+from openlit.__helpers import (
+ get_audio_model_cost,
+ handle_exception,
+ create_metrics_attributes,
+ set_server_address_and_port
+)
from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
logger = logging.getLogger(__name__)
-def transcribe(gen_ai_endpoint, version, environment, application_name,
+def transcribe(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for creating speech audio to collect metrics.
@@ -48,43 +52,56 @@ def transcribe(gen_ai_endpoint, version, environment, application_name,
The response from the original 'transcribe' method.
"""
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
+ server_address, server_port = set_server_address_and_port(instance, 'api.assemblyai.com', 443)
+ request_model = kwargs.get('speech_model', 'best')
+
+ span_name = f'{SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO} {request_model}'
+
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
response = wrapped(*args, **kwargs)
+ end_time = time.time()
try:
- llm_model = response.speech_model if response.speech_model else "best"
-
# Calculate cost of the operation
- cost = get_audio_model_cost(llm_model,
+ cost = get_audio_model_cost(request_model,
pricing_info, None, response.audio_duration)
- # Set Span attributes
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_ASSEMBLYAI)
+ # Set Span attributes (OTel Semconv)
+ span.set_attribute(TELEMETRY_SDK_NAME, 'openlit')
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_ASSEMBLYAI)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ 'text')
+
+ # Set Span attributes (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
+ span.set_attribute(SERVICE_NAME,
application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
- response.id)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- llm_model)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_DURATION,
response.audio_duration)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
if trace_content:
span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: response.audio_url,
- },
+ name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: response.audio_url,
+ },
)
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
@@ -96,30 +113,29 @@ def transcribe(gen_ai_endpoint, version, environment, application_name,
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_ASSEMBLYAI,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- llm_model,
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO,
+ system=SemanticConvetion.GEN_AI_SYSTEM_ASSEMBLYAI,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=request_model,
+ )
+
+ metrics['genai_client_operation_duration'].record(
+ end_time - start_time, attributes
+ )
+ metrics['genai_requests'].add(1, attributes)
+ metrics['genai_cost'].record(cost, attributes)
# Return original response
return response
except Exception as e:
handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
+ logger.error('Error in trace creation: %s', e)
# Return original response
return response
diff --git a/sdk/python/src/openlit/instrumentation/azure_ai_inference/__init__.py b/sdk/python/src/openlit/instrumentation/azure_ai_inference/__init__.py
index 81fcf234..cb499bed 100644
--- a/sdk/python/src/openlit/instrumentation/azure_ai_inference/__init__.py
+++ b/sdk/python/src/openlit/instrumentation/azure_ai_inference/__init__.py
@@ -38,7 +38,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
wrap_function_wrapper(
"azure.ai.inference",
"ChatCompletionsClient.complete",
- complete("azure_ai.complete", version, environment, application_name,
+ complete(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
@@ -46,7 +46,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
wrap_function_wrapper(
"azure.ai.inference",
"EmbeddingsClient.embed",
- embedding("azure_ai.embed", version, environment, application_name,
+ embedding(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
@@ -54,7 +54,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
wrap_function_wrapper(
"azure.ai.inference.aio",
"ChatCompletionsClient.complete",
- async_complete("azure_ai.complete", version, environment, application_name,
+ async_complete(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
@@ -62,7 +62,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
wrap_function_wrapper(
"azure.ai.inference.aio",
"EmbeddingsClient.embed",
- async_embedding("azure_ai.embed", version, environment, application_name,
+ async_embedding(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
diff --git a/sdk/python/src/openlit/instrumentation/azure_ai_inference/async_azure_ai_inference.py b/sdk/python/src/openlit/instrumentation/azure_ai_inference/async_azure_ai_inference.py
index e22b638d..b19578da 100644
--- a/sdk/python/src/openlit/instrumentation/azure_ai_inference/async_azure_ai_inference.py
+++ b/sdk/python/src/openlit/instrumentation/azure_ai_inference/async_azure_ai_inference.py
@@ -1,15 +1,20 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment, protected-access
"""
Module for monitoring Azure AI Inference API calls.
"""
import logging
+import time
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
from openlit.__helpers import (
- handle_exception,
get_chat_model_cost,
get_embed_model_cost,
+ handle_exception,
+ response_as_dict,
+ calculate_ttft,
+ calculate_tbt,
+ create_metrics_attributes,
+ set_server_address_and_port,
general_tokens
)
from openlit.semcov import SemanticConvetion
@@ -17,13 +22,12 @@ from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
logger = logging.getLogger(__name__)
-def async_complete(gen_ai_endpoint, version, environment, application_name,
+def async_complete(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for chat to collect metrics.
Args:
- gen_ai_endpoint: Endpoint identifier for logging and tracing.
version: Version of the monitoring package.
environment: Deployment environment (e.g., production, staging).
application_name: Name of the application using the Azure AI Inference API.
@@ -35,165 +39,275 @@ def async_complete(gen_ai_endpoint, version, environment, application_name,
A function that wraps the chat method to add telemetry.
"""
+ class TracedAsyncStream:
+ """
+ Wrapper for streaming responses to collect metrics and trace data.
+ Wraps the 'az.ai.inference.AsyncStream' response to collect message IDs and aggregated response.
+
+ This class implements the '__aiter__' and '__anext__' methods that
+ handle asynchronous streaming responses.
+
+ This class also implements '__aenter__' and '__aexit__' methods that
+ handle asynchronous context management protocol.
+ """
+ def __init__(
+ self,
+ wrapped,
+ span,
+ kwargs,
+ server_address,
+ server_port,
+ **args,
+ ):
+ self.__wrapped__ = wrapped
+ self._span = span
+ # Placeholder for aggregating streaming response
+ self._llmresponse = ""
+ self._response_id = ""
+ self._response_model = ""
+ self._finish_reason = ""
+ self._system_fingerprint = ""
+
+ self._args = args
+ self._kwargs = kwargs
+ self._start_time = time.time()
+ self._end_time = None
+ self._timestamps = []
+ self._ttft = 0
+ self._tbt = 0
+ self._server_address = server_address
+ self._server_port = server_port
+
+ async def __aenter__(self):
+ await self.__wrapped__.__aenter__()
+ return self
+
+ async def __aexit__(self, exc_type, exc_value, traceback):
+ await self.__wrapped__.__aexit__(exc_type, exc_value, traceback)
+
+ def __aiter__(self):
+ return self
+
+ async def __getattr__(self, name):
+ """Delegate attribute access to the wrapped object."""
+ return getattr(await self.__wrapped__, name)
+
+ async def __anext__(self):
+ try:
+ chunk = await self.__wrapped__.__anext__()
+ end_time = time.time()
+ # Record the timestamp for the current chunk
+ self._timestamps.append(end_time)
+
+ if len(self._timestamps) == 1:
+ # Calculate time to first chunk
+ self._ttft = calculate_ttft(self._timestamps, self._start_time)
+
+ chunked = response_as_dict(chunk)
+ # Collect message IDs and aggregated response from events
+ 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:
+ self._llmresponse += content
+ self._response_id = chunked.get('id')
+ self._response_model = chunked.get('model')
+ self._finish_reason = chunked.get('choices')[0].get('finish_reason')
+ self._system_fingerprint = chunked.get('system_fingerprint')
+ return chunk
+ except StopAsyncIteration:
+ # Handling exception ensure observability without disrupting operation
+ try:
+ self._end_time = time.time()
+ if len(self._timestamps) > 1:
+ self._tbt = calculate_tbt(self._timestamps)
+
+ # Format 'messages' into a single string
+ message_prompt = self._kwargs.get("messages", "")
+ formatted_messages = []
+ for message in message_prompt:
+ role = message["role"]
+ content = message["content"]
+
+ if isinstance(content, list):
+ content_str_list = []
+ for item in content:
+ if item["type"] == "text":
+ content_str_list.append(f'text: {item["text"]}')
+ elif (item["type"] == "image_url" and
+ not item["image_url"]["url"].startswith("data:")):
+ content_str_list.append(f'image_url: {item["image_url"]["url"]}')
+ content_str = ", ".join(content_str_list)
+ formatted_messages.append(f"{role}: {content_str}")
+ else:
+ formatted_messages.append(f"{role}: {content}")
+ prompt = "\n".join(formatted_messages)
+
+ request_model = self._kwargs.get("model", "gpt-4o")
+
+ # Calculate tokens using input prompt and aggregated response
+ input_tokens = general_tokens(prompt)
+ output_tokens = general_tokens(self._llmresponse)
+
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model,
+ pricing_info, input_tokens,
+ output_tokens)
+
+ # Set Span attributes (OTel Semconv)
+ self._span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
+ SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
+ self._kwargs.get("seed", ""))
+ self._span.set_attribute(SemanticConvetion.SERVER_PORT,
+ self._server_port)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
+ self._kwargs.get("frequency_penalty", 0.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
+ self._kwargs.get("max_tokens", -1))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
+ self._kwargs.get("presence_penalty", 0.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES,
+ self._kwargs.get("stop", []))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
+ self._kwargs.get("temperature", 1.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
+ self._kwargs.get("top_p", 1.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
+ [self._finish_reason])
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
+ self._response_id)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ self._response_model)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
+ input_tokens)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+ output_tokens)
+ self._span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ self._server_address)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OPENAI_RESPONSE_SYSTEM_FINGERPRINT,
+ self._system_fingerprint)
+ if isinstance(self._llmresponse, str):
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "text")
+ else:
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "json")
+
+ # Set Span attributes (Extra)
+ self._span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ self._span.set_attribute(SERVICE_NAME,
+ application_name)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ True)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
+ input_tokens + output_tokens)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TBT,
+ self._tbt)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ self._ttft)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
+ if trace_content:
+ self._span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
+ },
+ )
+ self._span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: self._llmresponse,
+ },
+ )
+ self._span.set_status(Status(StatusCode.OK))
+
+ if disable_metrics is False:
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ request_model=request_model,
+ server_address=self._server_address,
+ server_port=self._server_port,
+ response_model=self._response_model,
+ )
+
+ metrics["genai_client_usage_tokens"].record(
+ input_tokens + output_tokens, attributes
+ )
+ metrics["genai_client_operation_duration"].record(
+ self._end_time - self._start_time, attributes
+ )
+ metrics["genai_server_tbt"].record(
+ self._tbt, attributes
+ )
+ metrics["genai_server_ttft"].record(
+ self._ttft, 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)
+
+ except Exception as e:
+ handle_exception(self._span, e)
+ logger.error("Error in trace creation: %s", e)
+ finally:
+ self._span.end()
+ raise
+
async def wrapper(wrapped, instance, args, kwargs):
"""
- Wraps the 'chat' API call to add telemetry.
-
+ Wraps the 'chat.completions' API call to add telemetry.
+
This collects metrics such as execution time, cost, and token usage, and handles errors
gracefully, adding details to the trace for observability.
Args:
- wrapped: The original 'chat' method to be wrapped.
+ wrapped: The original 'chat.completions' method to be wrapped.
instance: The instance of the class where the original method is defined.
- args: Positional arguments for the 'chat' method.
- kwargs: Keyword arguments for the 'chat' method.
+ args: Positional arguments for the 'chat.completions' method.
+ kwargs: Keyword arguments for the 'chat.completions' method.
Returns:
- The response from the original 'chat' method.
+ The response from the original 'chat.completions' method.
"""
- # pylint: disable=no-else-return
- if kwargs.get("stream", False) is True:
- # Special handling for streaming response to accommodate the nature of data flow
- async def stream_generator():
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- # Placeholder for aggregating streaming response
- llmresponse = ""
-
- # Loop through streaming events capturing relevant details
- async for chunk in await wrapped(*args, **kwargs):
- if chunk.choices:
- # Collect message IDs and aggregated response from events
- content = chunk.choices[0].delta.content
- if content:
- llmresponse += content
-
- yield chunk
- response_id = chunk.id
-
- # Handling exception ensure observability without disrupting operation
- try:
- # Format 'messages' into a single string
- message_prompt = kwargs.get("messages", "")
- formatted_messages = []
- for message in message_prompt:
- role = message["role"]
- content = message["content"]
-
- if isinstance(content, list):
- content_str = ", ".join(
- # pylint: disable=line-too-long
- f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
- if "type" in item else f'text: {item["text"]}'
- for item in content
- )
- formatted_messages.append(f"{role}: {content_str}")
- else:
- formatted_messages.append(f"{role}: {content}")
- prompt = "\n".join(formatted_messages)
-
- model = kwargs.get("model", "phi3-mini-4k")
-
- # Calculate tokens using input prompt and aggregated response
- input_tokens = general_tokens(prompt)
- output_tokens = general_tokens(llmresponse)
-
- total_tokens = input_tokens + output_tokens
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens,
- output_tokens)
-
- # Set base span attribues
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
- span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- True)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
- kwargs.get("user", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
- kwargs.get("top_p", 1.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
- kwargs.get("max_tokens", -1))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
- kwargs.get("temperature", 1.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
- kwargs.get("presence_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
- kwargs.get("frequency_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
- kwargs.get("seed", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
- response_id)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- input_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
- output_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ # Check if streaming is enabled for the API call
+ streaming = kwargs.get("stream", False)
+ server_address, server_port = set_server_address_and_port(instance, "models.github.ai", 443)
+ request_model = kwargs.get("model", "gpt-4o")
- if trace_content:
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
- },
- )
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: llmresponse,
- },
- )
+ span_name = f"{SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
- span.set_status(Status(StatusCode.OK))
-
- if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(
- total_tokens, attributes
- )
- metrics["genai_completion_tokens"].add(output_tokens, attributes)
- metrics["genai_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ # pylint: disable=no-else-return
+ if streaming:
+ # Special handling for streaming response to accommodate the nature of data flow
+ awaited_wrapped = await wrapped(*args, **kwargs)
+ span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
- except Exception as e:
- handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
+ return TracedAsyncStream(awaited_wrapped, span, kwargs, server_address, server_port)
- return stream_generator()
+ # Handling for non-streaming responses
else:
- # pylint: disable=line-too-long
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
response = await wrapped(*args, **kwargs)
+ end_time = time.time()
+
+ response_dict = response_as_dict(response)
try:
# Format 'messages' into a single string
@@ -205,7 +319,6 @@ def async_complete(gen_ai_endpoint, version, environment, application_name,
if isinstance(content, list):
content_str = ", ".join(
- # pylint: disable=line-too-long
f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
if "type" in item else f'text: {item["text"]}'
for item in content
@@ -215,43 +328,66 @@ def async_complete(gen_ai_endpoint, version, environment, application_name,
formatted_messages.append(f"{role}: {content}")
prompt = "\n".join(formatted_messages)
- model = kwargs.get("model", "phi3-mini-4k")
+ input_tokens = response_dict.get('usage').get('prompt_tokens')
+ output_tokens = response_dict.get('usage').get('completion_tokens')
+
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model,
+ pricing_info, input_tokens,
+ output_tokens)
- # Set base span attribues
+ # Set base span attribues (OTel Semconv)
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- False)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
- kwargs.get("user", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
- kwargs.get("top_p", 1.0))
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
+ kwargs.get("seed", ""))
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
+ kwargs.get("frequency_penalty", 0.0))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
kwargs.get("max_tokens", -1))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
- kwargs.get("temperature", 1.0))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
kwargs.get("presence_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
- kwargs.get("frequency_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
- kwargs.get("seed", ""))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES,
+ kwargs.get("stop", []))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
+ kwargs.get("temperature", 1.0))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
+ kwargs.get("top_p", 1.0))
span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
- response.id)
- span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
- [response.choices[0]["finish_reason"]])
+ response_dict.get("id"))
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ response_dict.get('model'))
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
+ input_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+ output_tokens)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.GEN_AI_OPENAI_RESPONSE_SYSTEM_FINGERPRINT,
+ response_dict.get('system_fingerprint'))
+ # Set base span attribues (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ False)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
+ input_tokens + output_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ end_time - start_time)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
if trace_content:
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
@@ -259,49 +395,53 @@ def async_complete(gen_ai_endpoint, version, environment, application_name,
SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
},
)
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.choices[0].message.content,
- },
- )
- input_tokens = response.usage.prompt_tokens
- output_tokens = response.usage.completion_tokens
- total_tokens = response.usage.total_tokens
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens, output_tokens)
+ for i in range(kwargs.get('n',1)):
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
+ [response_dict.get('choices')[i].get('finish_reason')])
+ if trace_content:
+ span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+ attributes={
+ # pylint: disable=line-too-long
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: str(response_dict.get('choices')[i].get('message').get('content')),
+ },
+ )
+ if kwargs.get('tools'):
+ span.set_attribute(SemanticConvetion.GEN_AI_TOOL_CALLS,
+ str(response_dict.get('choices')[i].get('message').get('tool_calls')))
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- input_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
- output_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ if isinstance(response_dict.get('choices')[i].get('message').get('content'), str):
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "text")
+ elif response_dict.get('choices')[i].get('message').get('content') is not None:
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "json")
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=response_dict.get('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_server_ttft"].record(
+ end_time - start_time, attributes
+ )
metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(total_tokens, attributes)
metrics["genai_completion_tokens"].add(output_tokens, attributes)
metrics["genai_prompt_tokens"].add(input_tokens, attributes)
metrics["genai_cost"].record(cost, attributes)
@@ -318,18 +458,17 @@ def async_complete(gen_ai_endpoint, version, environment, application_name,
return wrapper
-def async_embedding(gen_ai_endpoint, version, environment, application_name,
+def async_embedding(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for embeddings to collect metrics.
Args:
- gen_ai_endpoint: Endpoint identifier for logging and tracing.
version: Version of the monitoring package.
environment: Deployment environment (e.g., production, staging).
- application_name: Name of the application using the Azure AI Inference API.
+ application_name: Name of the application using the Azure Inference API.
tracer: OpenTelemetry tracer for creating spans.
- pricing_info: Information used for calculating the cost of Azure AI Inference usage.
+ pricing_info: Information used for calculating the cost of Azure Inference usage.
trace_content: Flag indicating whether to trace the actual content.
Returns:
@@ -353,69 +492,84 @@ def async_embedding(gen_ai_endpoint, version, environment, application_name,
The response from the original 'embeddings' method.
"""
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
+ server_address, server_port = set_server_address_and_port(instance, "models.github.ai", 443)
+ request_model = kwargs.get("model", "text-embedding-ada-002")
+
+ span_name = f"{SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING} {request_model}"
+
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
response = await wrapped(*args, **kwargs)
+ end_time = time.time()
+ response_dict = response_as_dict(response)
try:
+ input_tokens = response_dict.get('usage').get('prompt_tokens')
+
# Calculate cost of the operation
- cost = get_embed_model_cost(kwargs.get("model", "text-embedding-ada-002"),
- pricing_info, response.usage.prompt_tokens)
+ cost = get_embed_model_cost(request_model,
+ pricing_info, input_tokens)
- # Set Span attributes
+ # Set Span attributes (OTel Semconv)
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- kwargs.get("model", "text-embedding-ada-002"))
+ request_model)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_ENCODING_FORMATS,
- kwargs.get("encoding_format", "float"))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_EMBEDDING_DIMENSION,
- kwargs.get("dimensions", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
- kwargs.get("user", ""))
+ [kwargs.get('encoding_format', 'float')])
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- response.usage.prompt_tokens)
+ input_tokens)
+
+ # Set Span attributes (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- response.usage.total_tokens)
+ input_tokens)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
+
if trace_content:
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: kwargs.get("input", ""),
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: str(kwargs.get("input", "")),
},
)
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- kwargs.get("model", "text-embedding-ada-002")
- }
-
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=request_model,
+ )
+ metrics["genai_client_usage_tokens"].record(
+ input_tokens, attributes
+ )
+ metrics["genai_client_operation_duration"].record(
+ end_time - start_time, attributes
+ )
metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(response.usage.total_tokens, attributes)
- metrics["genai_prompt_tokens"].add(response.usage.prompt_tokens, attributes)
+ metrics["genai_prompt_tokens"].add(input_tokens, attributes)
metrics["genai_cost"].record(cost, attributes)
# Return original response
diff --git a/sdk/python/src/openlit/instrumentation/azure_ai_inference/azure_ai_inference.py b/sdk/python/src/openlit/instrumentation/azure_ai_inference/azure_ai_inference.py
index 2612afc8..f5c56d1f 100644
--- a/sdk/python/src/openlit/instrumentation/azure_ai_inference/azure_ai_inference.py
+++ b/sdk/python/src/openlit/instrumentation/azure_ai_inference/azure_ai_inference.py
@@ -1,15 +1,20 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment, protected-access
"""
Module for monitoring Azure AI Inference API calls.
"""
import logging
+import time
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
from openlit.__helpers import (
- handle_exception,
get_chat_model_cost,
get_embed_model_cost,
+ handle_exception,
+ response_as_dict,
+ calculate_ttft,
+ calculate_tbt,
+ create_metrics_attributes,
+ set_server_address_and_port,
general_tokens
)
from openlit.semcov import SemanticConvetion
@@ -17,13 +22,12 @@ from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
logger = logging.getLogger(__name__)
-def complete(gen_ai_endpoint, version, environment, application_name,
+def complete(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for chat to collect metrics.
Args:
- gen_ai_endpoint: Endpoint identifier for logging and tracing.
version: Version of the monitoring package.
environment: Deployment environment (e.g., production, staging).
application_name: Name of the application using the Azure AI Inference API.
@@ -35,165 +39,275 @@ def complete(gen_ai_endpoint, version, environment, application_name,
A function that wraps the chat method to add telemetry.
"""
+ class TracedSyncStream:
+ """
+ Wrapper for streaming responses to collect metrics and trace data.
+ Wraps the 'az.ai.inference.AsyncStream' response to collect message IDs and aggregated response.
+
+ This class implements the '__aiter__' and '__anext__' methods that
+ handle asynchronous streaming responses.
+
+ This class also implements '__aenter__' and '__aexit__' methods that
+ handle asynchronous context management protocol.
+ """
+ def __init__(
+ self,
+ wrapped,
+ span,
+ kwargs,
+ server_address,
+ server_port,
+ **args,
+ ):
+ self.__wrapped__ = wrapped
+ self._span = span
+ # Placeholder for aggregating streaming response
+ self._llmresponse = ""
+ self._response_id = ""
+ self._response_model = ""
+ self._finish_reason = ""
+ self._system_fingerprint = ""
+
+ self._args = args
+ self._kwargs = kwargs
+ self._start_time = time.time()
+ self._end_time = None
+ self._timestamps = []
+ self._ttft = 0
+ self._tbt = 0
+ self._server_address = server_address
+ self._server_port = server_port
+
+ def __enter__(self):
+ self.__wrapped__.__enter__()
+ return self
+
+ def __exit__(self, exc_type, exc_value, traceback):
+ self.__wrapped__.__exit__(exc_type, exc_value, traceback)
+
+ def __iter__(self):
+ return self
+
+ def __getattr__(self, name):
+ """Delegate attribute access to the wrapped object."""
+ return getattr(self.__wrapped__, name)
+
+ def __next__(self):
+ try:
+ chunk = self.__wrapped__.__next__()
+ end_time = time.time()
+ # Record the timestamp for the current chunk
+ self._timestamps.append(end_time)
+
+ if len(self._timestamps) == 1:
+ # Calculate time to first chunk
+ self._ttft = calculate_ttft(self._timestamps, self._start_time)
+
+ chunked = response_as_dict(chunk)
+ # Collect message IDs and aggregated response from events
+ 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:
+ self._llmresponse += content
+ self._response_id = chunked.get('id')
+ self._response_model = chunked.get('model')
+ self._finish_reason = chunked.get('choices')[0].get('finish_reason')
+ self._system_fingerprint = chunked.get('system_fingerprint')
+ return chunk
+ except StopIteration:
+ # Handling exception ensure observability without disrupting operation
+ try:
+ self._end_time = time.time()
+ if len(self._timestamps) > 1:
+ self._tbt = calculate_tbt(self._timestamps)
+
+ # Format 'messages' into a single string
+ message_prompt = self._kwargs.get("messages", "")
+ formatted_messages = []
+ for message in message_prompt:
+ role = message["role"]
+ content = message["content"]
+
+ if isinstance(content, list):
+ content_str_list = []
+ for item in content:
+ if item["type"] == "text":
+ content_str_list.append(f'text: {item["text"]}')
+ elif (item["type"] == "image_url" and
+ not item["image_url"]["url"].startswith("data:")):
+ content_str_list.append(f'image_url: {item["image_url"]["url"]}')
+ content_str = ", ".join(content_str_list)
+ formatted_messages.append(f"{role}: {content_str}")
+ else:
+ formatted_messages.append(f"{role}: {content}")
+ prompt = "\n".join(formatted_messages)
+
+ request_model = self._kwargs.get("model", "gpt-4o")
+
+ # Calculate tokens using input prompt and aggregated response
+ input_tokens = general_tokens(prompt)
+ output_tokens = general_tokens(self._llmresponse)
+
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model,
+ pricing_info, input_tokens,
+ output_tokens)
+
+ # Set Span attributes (OTel Semconv)
+ self._span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
+ SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
+ self._kwargs.get("seed", ""))
+ self._span.set_attribute(SemanticConvetion.SERVER_PORT,
+ self._server_port)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
+ self._kwargs.get("frequency_penalty", 0.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
+ self._kwargs.get("max_tokens", -1))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
+ self._kwargs.get("presence_penalty", 0.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES,
+ self._kwargs.get("stop", []))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
+ self._kwargs.get("temperature", 1.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
+ self._kwargs.get("top_p", 1.0))
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
+ [self._finish_reason])
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
+ self._response_id)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ self._response_model)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
+ input_tokens)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+ output_tokens)
+ self._span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ self._server_address)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OPENAI_RESPONSE_SYSTEM_FINGERPRINT,
+ self._system_fingerprint)
+ if isinstance(self._llmresponse, str):
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "text")
+ else:
+ self._span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "json")
+
+ # Set Span attributes (Extra)
+ self._span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ self._span.set_attribute(SERVICE_NAME,
+ application_name)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ True)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
+ input_tokens + output_tokens)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TBT,
+ self._tbt)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ self._ttft)
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
+ if trace_content:
+ self._span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
+ },
+ )
+ self._span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: self._llmresponse,
+ },
+ )
+ self._span.set_status(Status(StatusCode.OK))
+
+ if disable_metrics is False:
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ request_model=request_model,
+ server_address=self._server_address,
+ server_port=self._server_port,
+ response_model=self._response_model,
+ )
+
+ metrics["genai_client_usage_tokens"].record(
+ input_tokens + output_tokens, attributes
+ )
+ metrics["genai_client_operation_duration"].record(
+ self._end_time - self._start_time, attributes
+ )
+ metrics["genai_server_tbt"].record(
+ self._tbt, attributes
+ )
+ metrics["genai_server_ttft"].record(
+ self._ttft, 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)
+
+ except Exception as e:
+ handle_exception(self._span, e)
+ logger.error("Error in trace creation: %s", e)
+ finally:
+ self._span.end()
+ raise
+
def wrapper(wrapped, instance, args, kwargs):
"""
- Wraps the 'chat' API call to add telemetry.
-
+ Wraps the 'chat.completions' API call to add telemetry.
+
This collects metrics such as execution time, cost, and token usage, and handles errors
gracefully, adding details to the trace for observability.
Args:
- wrapped: The original 'chat' method to be wrapped.
+ wrapped: The original 'chat.completions' method to be wrapped.
instance: The instance of the class where the original method is defined.
- args: Positional arguments for the 'chat' method.
- kwargs: Keyword arguments for the 'chat' method.
+ args: Positional arguments for the 'chat.completions' method.
+ kwargs: Keyword arguments for the 'chat.completions' method.
Returns:
- The response from the original 'chat' method.
+ The response from the original 'chat.completions' method.
"""
- # pylint: disable=no-else-return
- if kwargs.get("stream", False) is True:
- # Special handling for streaming response to accommodate the nature of data flow
- def stream_generator():
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- # Placeholder for aggregating streaming response
- llmresponse = ""
-
- # Loop through streaming events capturing relevant details
- for chunk in wrapped(*args, **kwargs):
- if chunk.choices:
- # Collect message IDs and aggregated response from events
- content = chunk.choices[0].delta.content
- if content:
- llmresponse += content
-
- yield chunk
- response_id = chunk.id
-
- # Handling exception ensure observability without disrupting operation
- try:
- # Format 'messages' into a single string
- message_prompt = kwargs.get("messages", "")
- formatted_messages = []
- for message in message_prompt:
- role = message["role"]
- content = message["content"]
-
- if isinstance(content, list):
- content_str = ", ".join(
- # pylint: disable=line-too-long
- f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
- if "type" in item else f'text: {item["text"]}'
- for item in content
- )
- formatted_messages.append(f"{role}: {content_str}")
- else:
- formatted_messages.append(f"{role}: {content}")
- prompt = "\n".join(formatted_messages)
-
- model = kwargs.get("model", "phi3-mini-4k")
-
- # Calculate tokens using input prompt and aggregated response
- input_tokens = general_tokens(prompt)
- output_tokens = general_tokens(llmresponse)
-
- total_tokens = input_tokens + output_tokens
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens,
- output_tokens)
-
- # Set base span attribues
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
- span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- True)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
- kwargs.get("user", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
- kwargs.get("top_p", 1.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
- kwargs.get("max_tokens", -1))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
- kwargs.get("temperature", 1.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
- kwargs.get("presence_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
- kwargs.get("frequency_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
- kwargs.get("seed", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
- response_id)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- input_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
- output_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ # Check if streaming is enabled for the API call
+ streaming = kwargs.get("stream", False)
+ server_address, server_port = set_server_address_and_port(instance, "models.github.ai", 443)
+ request_model = kwargs.get("model", "gpt-4o")
- if trace_content:
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
- },
- )
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: llmresponse,
- },
- )
+ span_name = f"{SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
- span.set_status(Status(StatusCode.OK))
-
- if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(
- total_tokens, attributes
- )
- metrics["genai_completion_tokens"].add(output_tokens, attributes)
- metrics["genai_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ # pylint: disable=no-else-return
+ if streaming:
+ # Special handling for streaming response to accommodate the nature of data flow
+ awaited_wrapped = wrapped(*args, **kwargs)
+ span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
- except Exception as e:
- handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
+ return TracedSyncStream(awaited_wrapped, span, kwargs, server_address, server_port)
- return stream_generator()
+ # Handling for non-streaming responses
else:
- # pylint: disable=line-too-long
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
response = wrapped(*args, **kwargs)
+ end_time = time.time()
+
+ response_dict = response_as_dict(response)
try:
# Format 'messages' into a single string
@@ -205,7 +319,6 @@ def complete(gen_ai_endpoint, version, environment, application_name,
if isinstance(content, list):
content_str = ", ".join(
- # pylint: disable=line-too-long
f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
if "type" in item else f'text: {item["text"]}'
for item in content
@@ -215,43 +328,66 @@ def complete(gen_ai_endpoint, version, environment, application_name,
formatted_messages.append(f"{role}: {content}")
prompt = "\n".join(formatted_messages)
- model = kwargs.get("model", "phi3-mini-4k")
+ input_tokens = response_dict.get('usage').get('prompt_tokens')
+ output_tokens = response_dict.get('usage').get('completion_tokens')
+
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model,
+ pricing_info, input_tokens,
+ output_tokens)
- # Set base span attribues
+ # Set base span attribues (OTel Semconv)
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- False)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
- kwargs.get("user", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
- kwargs.get("top_p", 1.0))
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
+ kwargs.get("seed", ""))
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
+ kwargs.get("frequency_penalty", 0.0))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
kwargs.get("max_tokens", -1))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
- kwargs.get("temperature", 1.0))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
kwargs.get("presence_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
- kwargs.get("frequency_penalty", 0.0))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
- kwargs.get("seed", ""))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES,
+ kwargs.get("stop", []))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
+ kwargs.get("temperature", 1.0))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
+ kwargs.get("top_p", 1.0))
span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
- response.id)
- span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
- [response.choices[0]["finish_reason"]])
+ response_dict.get("id"))
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ response_dict.get('model'))
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
+ input_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+ output_tokens)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.GEN_AI_OPENAI_RESPONSE_SYSTEM_FINGERPRINT,
+ response_dict.get('system_fingerprint'))
+ # Set base span attribues (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ False)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
+ input_tokens + output_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ end_time - start_time)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
if trace_content:
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
@@ -259,49 +395,53 @@ def complete(gen_ai_endpoint, version, environment, application_name,
SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
},
)
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.choices[0].message.content,
- },
- )
- input_tokens = response.usage.prompt_tokens
- output_tokens = response.usage.completion_tokens
- total_tokens = response.usage.total_tokens
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens, output_tokens)
+ for i in range(kwargs.get('n',1)):
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
+ [response_dict.get('choices')[i].get('finish_reason')])
+ if trace_content:
+ span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+ attributes={
+ # pylint: disable=line-too-long
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: str(response_dict.get('choices')[i].get('message').get('content')),
+ },
+ )
+ if kwargs.get('tools'):
+ span.set_attribute(SemanticConvetion.GEN_AI_TOOL_CALLS,
+ str(response_dict.get('choices')[i].get('message').get('tool_calls')))
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- input_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
- output_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ if isinstance(response_dict.get('choices')[i].get('message').get('content'), str):
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "text")
+ elif response_dict.get('choices')[i].get('message').get('content') is not None:
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "json")
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=response_dict.get('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_server_ttft"].record(
+ end_time - start_time, attributes
+ )
metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(total_tokens, attributes)
metrics["genai_completion_tokens"].add(output_tokens, attributes)
metrics["genai_prompt_tokens"].add(input_tokens, attributes)
metrics["genai_cost"].record(cost, attributes)
@@ -318,18 +458,17 @@ def complete(gen_ai_endpoint, version, environment, application_name,
return wrapper
-def embedding(gen_ai_endpoint, version, environment, application_name,
+def embedding(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for embeddings to collect metrics.
Args:
- gen_ai_endpoint: Endpoint identifier for logging and tracing.
version: Version of the monitoring package.
environment: Deployment environment (e.g., production, staging).
- application_name: Name of the application using the Azure AI Inference API.
+ application_name: Name of the application using the Azure Inference API.
tracer: OpenTelemetry tracer for creating spans.
- pricing_info: Information used for calculating the cost of Azure AI Inference usage.
+ pricing_info: Information used for calculating the cost of Azure Inference usage.
trace_content: Flag indicating whether to trace the actual content.
Returns:
@@ -353,69 +492,84 @@ def embedding(gen_ai_endpoint, version, environment, application_name,
The response from the original 'embeddings' method.
"""
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
+ server_address, server_port = set_server_address_and_port(instance, "models.github.ai", 443)
+ request_model = kwargs.get("model", "text-embedding-ada-002")
+
+ span_name = f"{SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING} {request_model}"
+
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
response = wrapped(*args, **kwargs)
+ end_time = time.time()
+ response_dict = response_as_dict(response)
try:
+ input_tokens = response_dict.get('usage').get('prompt_tokens')
+
# Calculate cost of the operation
- cost = get_embed_model_cost(kwargs.get("model", "text-embedding-ada-002"),
- pricing_info, response.usage.prompt_tokens)
+ cost = get_embed_model_cost(request_model,
+ pricing_info, input_tokens)
- # Set Span attributes
+ # Set Span attributes (OTel Semconv)
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- kwargs.get("model", "text-embedding-ada-002"))
+ request_model)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_ENCODING_FORMATS,
- kwargs.get("encoding_format", "float"))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_EMBEDDING_DIMENSION,
- kwargs.get("dimensions", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
- kwargs.get("user", ""))
+ [kwargs.get('encoding_format', 'float')])
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- response.usage.prompt_tokens)
+ input_tokens)
+
+ # Set Span attributes (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- response.usage.total_tokens)
+ input_tokens)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
+
if trace_content:
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: kwargs.get("input", ""),
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: str(kwargs.get("input", "")),
},
)
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- kwargs.get("model", "text-embedding-ada-002")
- }
-
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=request_model,
+ )
+ metrics["genai_client_usage_tokens"].record(
+ input_tokens, attributes
+ )
+ metrics["genai_client_operation_duration"].record(
+ end_time - start_time, attributes
+ )
metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(response.usage.total_tokens, attributes)
- metrics["genai_prompt_tokens"].add(response.usage.prompt_tokens, attributes)
+ metrics["genai_prompt_tokens"].add(input_tokens, attributes)
metrics["genai_cost"].record(cost, attributes)
# Return original response
diff --git a/sdk/python/src/openlit/instrumentation/bedrock/__init__.py b/sdk/python/src/openlit/instrumentation/bedrock/__init__.py
index f26c4e86..71a1f1b8 100644
--- a/sdk/python/src/openlit/instrumentation/bedrock/__init__.py
+++ b/sdk/python/src/openlit/instrumentation/bedrock/__init__.py
@@ -32,7 +32,7 @@ class BedrockInstrumentor(BaseInstrumentor):
wrap_function_wrapper(
"botocore.client",
"ClientCreator.create_client",
- converse("bedrock.converse", version, environment, application_name,
+ converse(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
diff --git a/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py b/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py
index 281dd991..8cda881d 100644
--- a/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py
+++ b/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py
@@ -1,17 +1,22 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, protected-access, too-many-branches
"""
Module for monitoring Amazon Bedrock API calls.
"""
import logging
+import time
from botocore.response import StreamingBody
from botocore.exceptions import ReadTimeoutError, ResponseStreamingError
from urllib3.exceptions import ProtocolError as URLLib3ProtocolError
from urllib3.exceptions import ReadTimeoutError as URLLib3ReadTimeoutError
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
-from openlit.__helpers import get_chat_model_cost
-from openlit.__helpers import handle_exception
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
+from openlit.__helpers import (
+ get_chat_model_cost,
+ handle_exception,
+ response_as_dict,
+ create_metrics_attributes,
+ set_server_address_and_port
+)
from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
@@ -47,8 +52,7 @@ class CustomStreamWrapper(StreamingBody):
return data_chunk
-
-def converse(gen_ai_endpoint, version, environment, application_name, tracer,
+def converse(version, environment, application_name, tracer,
pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for messages to collect metrics.
@@ -91,52 +95,104 @@ def converse(gen_ai_endpoint, version, environment, application_name, tracer,
Returns:
The modified response with telemetry.
"""
- with tracer.start_as_current_span(gen_ai_endpoint, kind=SpanKind.CLIENT) as span:
+
+ server_address, server_port = set_server_address_and_port(instance, 'aws.amazon.com', 443)
+ request_model = method_kwargs.get('modelId', 'amazon.titan-text-express-v1')
+
+ span_name = f'{SemanticConvetion.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 = original_method(*method_args, **method_kwargs)
+ end_time = time.time()
+
+ response_dict = response_as_dict(response)
try:
- message_prompt = method_kwargs.get("messages", "")
+ message_prompt = method_kwargs.get('messages', '')
formatted_messages = []
for message in message_prompt:
- role = message["role"]
- content = message["content"]
+ role = message['role']
+ content = message['content']
if isinstance(content, list):
- # pylint: disable=line-too-long
content_str = ", ".join(f'text: {item["text"]}' for item in content if "text" in item)
- formatted_messages.append(f"{role}: {content_str}")
+ formatted_messages.append(f'{role}: {content_str}')
else:
- formatted_messages.append(f"{role}: {content}")
- prompt = "\n".join(formatted_messages)
+ formatted_messages.append(f'{role}: {content}')
+ prompt = '\n'.join(formatted_messages)
+
+ input_tokens = response_dict.get('usage').get('inputTokens')
+ output_tokens = response_dict.get('usage').get('outputTokens')
- model = method_kwargs.get("modelId", "amazon.titan-text-express-v1")
- input_tokens = response["usage"]["inputTokens"]
- output_tokens = response["usage"]["outputTokens"]
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model, pricing_info,
+ input_tokens, output_tokens)
+
+ llm_response = response_dict.get('output').get('message').get('content')[0].get('text')
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
+ # Set base span attribues (OTel Semconv)
+ span.set_attribute(TELEMETRY_SDK_NAME, 'openlit')
+ span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
+ SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
SemanticConvetion.GEN_AI_SYSTEM_AWS_BEDROCK)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+
+ inference_config = method_kwargs.get('inferenceConfig', {})
+
+ # List of attributes and their config keys
+ attributes = [
+ (SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY, 'frequencyPenalty'),
+ (SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS, 'maxTokens'),
+ (SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY, 'presencePenalty'),
+ (SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES, 'stopSequences'),
+ (SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE, 'temperature'),
+ (SemanticConvetion.GEN_AI_REQUEST_TOP_P, 'topP'),
+ (SemanticConvetion.GEN_AI_REQUEST_TOP_K, 'topK'),
+ ]
+
+ # Set each attribute if the corresponding value exists and is not None
+ for attribute, key in attributes:
+ value = inference_config.get(key)
+ if value is not None:
+ span.set_attribute(attribute, value)
+
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
+ response_dict.get('ResponseMetadata').get('RequestId'))
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ request_model)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
input_tokens)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
output_tokens)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ if isinstance(llm_response, str):
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ 'text')
+ else:
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ 'json')
+
+ # Set base span attribues (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ False)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
input_tokens + output_tokens)
-
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens,
- output_tokens)
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ end_time - start_time)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
if trace_content:
span.add_event(
@@ -148,42 +204,43 @@ def converse(gen_ai_endpoint, version, environment, application_name, tracer,
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
attributes={
- # pylint: disable=line-too-long
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response["output"]["message"]["content"][0]["text"],
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: llm_response,
},
)
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_AWS_BEDROCK,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_AWS_BEDROCK,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=request_model,
+ )
+
+ metrics['genai_client_usage_tokens'].record(
input_tokens + output_tokens, attributes
)
- metrics["genai_completion_tokens"].add(output_tokens, attributes)
- metrics["genai_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ metrics['genai_client_operation_duration'].record(
+ end_time - start_time, 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)
return response
except Exception as e:
handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
+ logger.error('Error in trace creation: %s', e)
# Return original response
return response
@@ -192,7 +249,7 @@ def converse(gen_ai_endpoint, version, environment, application_name, tracer,
client = wrapped(*args, **kwargs)
# Replace the original method with the instrumented one
- if kwargs.get("service_name") == "bedrock-runtime":
+ if kwargs.get('service_name') == 'bedrock-runtime':
original_invoke_model = client.converse
client.converse = lambda *args, **kwargs: converse_wrapper(original_invoke_model,
*args, **kwargs)
diff --git a/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py b/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py
index ac9b78d7..2aa83a09 100644
--- a/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py
+++ b/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py
@@ -244,7 +244,7 @@ def async_chat(version, environment, application_name, tracer,
kwargs.get("stop_sequences", []))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
kwargs.get("temperature", 0.3))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_K,
kwargs.get("k", 1.0))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
kwargs.get("p", 1.0))
diff --git a/sdk/python/src/openlit/instrumentation/cohere/cohere.py b/sdk/python/src/openlit/instrumentation/cohere/cohere.py
index 7158c604..4bd4e132 100644
--- a/sdk/python/src/openlit/instrumentation/cohere/cohere.py
+++ b/sdk/python/src/openlit/instrumentation/cohere/cohere.py
@@ -244,7 +244,7 @@ def chat(version, environment, application_name, tracer,
kwargs.get("stop_sequences", []))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
kwargs.get("temperature", 0.3))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_K,
kwargs.get("k", 1.0))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
kwargs.get("p", 1.0))
diff --git a/sdk/python/src/openlit/instrumentation/elevenlabs/async_elevenlabs.py b/sdk/python/src/openlit/instrumentation/elevenlabs/async_elevenlabs.py
index 529a6817..54440aec 100644
--- a/sdk/python/src/openlit/instrumentation/elevenlabs/async_elevenlabs.py
+++ b/sdk/python/src/openlit/instrumentation/elevenlabs/async_elevenlabs.py
@@ -1,13 +1,17 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment
"""
Module for monitoring Ollama API calls.
"""
import logging
+import time
+from urllib.parse import urlparse
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
-from openlit.__helpers import get_audio_model_cost
-from openlit.__helpers import handle_exception
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
+from openlit.__helpers import (
+ get_audio_model_cost,
+ handle_exception,
+ create_metrics_attributes,
+)
from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
@@ -19,7 +23,6 @@ def async_generate(gen_ai_endpoint, version, environment, application_name,
Generates a telemetry wrapper for creating speech audio to collect metrics.
Args:
- gen_ai_endpoint: Endpoint identifier for logging and tracing.
version: Version of the monitoring package.
environment: Deployment environment (e.g., production, staging).
application_name: Name of the application using the ElevenLabs API.
@@ -47,73 +50,96 @@ def async_generate(gen_ai_endpoint, version, environment, application_name,
Returns:
The response from the original 'generate' method.
"""
+ url = urlparse(instance._client_wrapper._base_url)
+ server_address, server_port = url.hostname, url.port or 443
+ request_model = kwargs.get('model', kwargs.get('model_id', 'eleven_multilingual_v2'))
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- response = await wrapped(*args, **kwargs)
+ span_name = f'{SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO} {request_model}'
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
+ response = await wrapped(*args, **kwargs)
+ end_time = time.time()
+ for key, value in instance.__dict__.items():
+ print(f'{key}: {value}')
try:
# Calculate cost of the operation
- cost = get_audio_model_cost(kwargs.get("model", "eleven_multilingual_v2"),
- pricing_info, kwargs.get("text", ""))
+ cost = get_audio_model_cost(request_model,
+ pricing_info, kwargs.get('text', ''))
# Set Span attributes
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_ELEVENLABS)
+ span.set_attribute(TELEMETRY_SDK_NAME, 'openlit')
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_ASSEMBLYAI)
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- kwargs.get("model", "eleven_multilingual_v2"))
- if isinstance(kwargs.get("voice", "Rachel"), str):
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ 'audio')
+
+ # Set Span attributes (Extras)
+ if gen_ai_endpoint == 'elevenlabs.generate':
+ if isinstance(kwargs.get('voice', 'Rachel'), str):
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_VOICE,
+ kwargs.get('voice', 'Rachel'))
+ else:
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_VOICE,
- kwargs.get("voice", "Rachel"))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_RESPONSE_FORMAT,
- kwargs.get("output_format", "mp3"))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_SETTINGS,
- str(kwargs.get("voice_settings", "")))
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ kwargs.get('voice_id', ''))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_RESPONSE_FORMAT,
+ kwargs.get('output_format', 'mp3'))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_SETTINGS,
+ str(kwargs.get('voice_settings', '')))
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
if trace_content:
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: str(kwargs.get("text", "")),
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: str(kwargs.get('text', '')),
},
)
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_ELEVENLABS,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- kwargs.get("model", "eleven_multilingual_v2")
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO,
+ system=SemanticConvetion.GEN_AI_SYSTEM_ELEVENLABS,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=request_model,
+ )
+
+ metrics['genai_client_operation_duration'].record(
+ end_time - start_time, attributes
+ )
+ metrics['genai_requests'].add(1, attributes)
+ metrics['genai_cost'].record(cost, attributes)
# Return original response
return response
except Exception as e:
handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
+ logger.error('Error in trace creation: %s', e)
# Return original response
return response
diff --git a/sdk/python/src/openlit/instrumentation/elevenlabs/elevenlabs.py b/sdk/python/src/openlit/instrumentation/elevenlabs/elevenlabs.py
index b4df23c0..7b3d9cec 100644
--- a/sdk/python/src/openlit/instrumentation/elevenlabs/elevenlabs.py
+++ b/sdk/python/src/openlit/instrumentation/elevenlabs/elevenlabs.py
@@ -1,13 +1,17 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment
"""
Module for monitoring Ollama API calls.
"""
import logging
+import time
+from urllib.parse import urlparse
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
-from openlit.__helpers import get_audio_model_cost
-from openlit.__helpers import handle_exception
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
+from openlit.__helpers import (
+ get_audio_model_cost,
+ handle_exception,
+ create_metrics_attributes,
+)
from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
@@ -19,7 +23,6 @@ def generate(gen_ai_endpoint, version, environment, application_name,
Generates a telemetry wrapper for creating speech audio to collect metrics.
Args:
- gen_ai_endpoint: Endpoint identifier for logging and tracing.
version: Version of the monitoring package.
environment: Deployment environment (e.g., production, staging).
application_name: Name of the application using the ElevenLabs API.
@@ -48,78 +51,96 @@ def generate(gen_ai_endpoint, version, environment, application_name,
The response from the original 'generate' method.
"""
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- response = wrapped(*args, **kwargs)
+ url = urlparse(instance._client_wrapper._base_url)
+ server_address, server_port = url.hostname, url.port or 443
+ request_model = kwargs.get('model', kwargs.get('model_id', 'eleven_multilingual_v2'))
+
+ span_name = f'{SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO} {request_model}'
+ with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+ start_time = time.time()
+ response = wrapped(*args, **kwargs)
+ end_time = time.time()
+ for key, value in instance.__dict__.items():
+ print(f'{key}: {value}')
try:
# Calculate cost of the operation
- cost = get_audio_model_cost(kwargs.get("model", "eleven_multilingual_v2"),
- pricing_info, kwargs.get("text", ""))
+ cost = get_audio_model_cost(request_model,
+ pricing_info, kwargs.get('text', ''))
# Set Span attributes
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_ELEVENLABS)
+ span.set_attribute(TELEMETRY_SDK_NAME, 'openlit')
span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- if gen_ai_endpoint == "elevenlabs.generate":
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- kwargs.get("model", "eleven_multilingual_v2"))
- if isinstance(kwargs.get("voice", "Rachel"), str):
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_ASSEMBLYAI)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ 'audio')
+
+ # Set Span attributes (Extras)
+ if gen_ai_endpoint == 'elevenlabs.generate':
+ if isinstance(kwargs.get('voice', 'Rachel'), str):
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_VOICE,
- kwargs.get("voice", "Rachel"))
+ kwargs.get('voice', 'Rachel'))
else:
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- kwargs.get("model_id", "eleven_multilingual_v2"))
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_VOICE,
- kwargs.get("voice_id", ""))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_RESPONSE_FORMAT,
- kwargs.get("output_format", "mp3"))
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_SETTINGS,
- str(kwargs.get("voice_settings", "")))
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
+ kwargs.get('voice_id', ''))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_RESPONSE_FORMAT,
+ kwargs.get('output_format', 'mp3'))
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_AUDIO_SETTINGS,
+ str(kwargs.get('voice_settings', '')))
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
if trace_content:
span.add_event(
name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: str(kwargs.get("text", "")),
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: str(kwargs.get('text', '')),
},
)
span.set_status(Status(StatusCode.OK))
if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_ELEVENLABS,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- kwargs.get("model", "eleven_multilingual_v2")
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_AUDIO,
+ system=SemanticConvetion.GEN_AI_SYSTEM_ELEVENLABS,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=request_model,
+ )
+
+ metrics['genai_client_operation_duration'].record(
+ end_time - start_time, attributes
+ )
+ metrics['genai_requests'].add(1, attributes)
+ metrics['genai_cost'].record(cost, attributes)
# Return original response
return response
except Exception as e:
handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
+ logger.error('Error in trace creation: %s', e)
# Return original response
return response
diff --git a/sdk/python/src/openlit/instrumentation/google_ai_studio/__init__.py b/sdk/python/src/openlit/instrumentation/google_ai_studio/__init__.py
index 897f8e9a..2219b009 100644
--- a/sdk/python/src/openlit/instrumentation/google_ai_studio/__init__.py
+++ b/sdk/python/src/openlit/instrumentation/google_ai_studio/__init__.py
@@ -14,11 +14,11 @@ from openlit.instrumentation.google_ai_studio.async_google_ai_studio import (
async_generate
)
-_instruments = ("google-generativeai >= 0.2.0",)
+_instruments = ("google-genai >= 1.3.0",)
class GoogleAIStudioInstrumentor(BaseInstrumentor):
"""
- An instrumentor for google-generativeai's client library.
+ An instrumentor for google-genai's client library.
"""
def instrumentation_dependencies(self) -> Collection[str]:
@@ -32,21 +32,21 @@ class GoogleAIStudioInstrumentor(BaseInstrumentor):
pricing_info = kwargs.get("pricing_info", {})
trace_content = kwargs.get("trace_content", False)
disable_metrics = kwargs.get("disable_metrics")
- version = importlib.metadata.version("google-generativeai")
+ version = importlib.metadata.version("google-genai")
# sync generate
wrap_function_wrapper(
- "google.generativeai.generative_models",
- "GenerativeModel.generate_content",
- generate("google_ai_studio.generate_content", version, environment, application_name,
+ "google.genai.models",
+ "Models.generate_content",
+ generate(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics),
)
# async generate
wrap_function_wrapper(
- "google.generativeai.generative_models",
- "GenerativeModel.generate_content_async",
- async_generate("google_ai_studio.generate_content", version, environment,
+ "google.genai.models",
+ "AsyncModels.generate_content",
+ async_generate(version, environment,
application_name, tracer, pricing_info, trace_content, metrics,
disable_metrics),
)
diff --git a/sdk/python/src/openlit/instrumentation/google_ai_studio/async_google_ai_studio.py b/sdk/python/src/openlit/instrumentation/google_ai_studio/async_google_ai_studio.py
index 943453ad..bb4721d4 100644
--- a/sdk/python/src/openlit/instrumentation/google_ai_studio/async_google_ai_studio.py
+++ b/sdk/python/src/openlit/instrumentation/google_ai_studio/async_google_ai_studio.py
@@ -1,21 +1,24 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment, protected-access
"""
Module for monitoring Google AI Studio API calls.
"""
import logging
+import time
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
from openlit.__helpers import (
- handle_exception,
get_chat_model_cost,
+ handle_exception,
+ response_as_dict,
+ create_metrics_attributes,
+ set_server_address_and_port
)
from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
logger = logging.getLogger(__name__)
-def async_generate(gen_ai_endpoint, version, environment, application_name,
+def async_generate(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for chat to collect metrics.
@@ -35,229 +38,190 @@ def async_generate(gen_ai_endpoint, version, environment, application_name,
async def wrapper(wrapped, instance, args, kwargs):
"""
- Wraps the 'chat' API call to add telemetry.
-
+ Wraps the 'chat.completions' API call to add telemetry.
+
This collects metrics such as execution time, cost, and token usage, and handles errors
gracefully, adding details to the trace for observability.
Args:
- wrapped: The original 'chat' method to be wrapped.
+ wrapped: The original 'chat.completions' method to be wrapped.
instance: The instance of the class where the original method is defined.
- args: Positional arguments for the 'chat' method.
- kwargs: Keyword arguments for the 'chat' method.
+ args: Positional arguments for the 'chat.completions' method.
+ kwargs: Keyword arguments for the 'chat.completions' method.
Returns:
- The response from the original 'chat' method.
+ The response from the original 'chat.completions' method.
"""
- # pylint: disable=no-else-return
- if kwargs.get('stream', False) is True:
- # Special handling for streaming response to accommodate the nature of data flow
- async def stream_generator():
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- # Placeholder for aggregating streaming response
- llmresponse = ""
-
- # Loop through streaming events capturing relevant details
- async for chunk in await wrapped(*args, **kwargs):
- # Collect message IDs and aggregated response from events
- content = chunk.text
- if content:
- llmresponse += content
-
- input_tokens = chunk.usage_metadata.prompt_token_count
- output_tokens = chunk.usage_metadata.candidates_token_count
- yield chunk
-
- # Handling exception ensure observability without disrupting operation
- try:
- prompt = ""
- for arg in args:
- if isinstance(arg, str):
- prompt = f"{prompt}{arg}\n"
- elif isinstance(arg, list):
- for subarg in arg:
- prompt = f"{prompt}{subarg}\n"
- if hasattr(instance, "_model_id"):
- model = instance._model_id
- if hasattr(instance, "_model_name"):
- model = instance._model_name.replace("publishers/google/models/", "")
- if model.startswith("models/"):
- model = model[len("models/"):]
-
- total_tokens = input_tokens + output_tokens
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens,
- output_tokens)
-
- # Set Span attributes
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- True)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- input_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+
+ server_address, server_port = set_server_address_and_port(instance, "generativelanguage.googleapis.com", 443)
+ request_model = kwargs.get("model", "gemini-2.0-flash")
+
+ span_name = f"{SemanticConvetion.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)
+ end_time = time.time()
+
+ response_dict = response_as_dict(response)
+
+ try:
+ # Format 'messages' into a single string
+ message_prompt = kwargs.get("contents", [])
+ formatted_messages = []
+
+ for content in message_prompt:
+ role = content.role
+ parts = content.parts
+ content_str = []
+
+ for part in parts:
+ # Collect relevant fields and handle each type of data that Part could contain
+ if part.text:
+ content_str.append(f"text: {part.text}")
+ if part.video_metadata:
+ content_str.append(f"video_metadata: {part.video_metadata}")
+ if part.thought:
+ content_str.append(f"thought: {part.thought}")
+ if part.code_execution_result:
+ content_str.append(f"code_execution_result: {part.code_execution_result}")
+ if part.executable_code:
+ content_str.append(f"executable_code: {part.executable_code}")
+ if part.file_data:
+ content_str.append(f"file_data: {part.file_data}")
+ if part.function_call:
+ content_str.append(f"function_call: {part.function_call}")
+ if part.function_response:
+ content_str.append(f"function_response: {part.function_response}")
+ if part.inline_data:
+ content_str.append(f"inline_data: {part.inline_data}")
+
+ formatted_messages.append(f"{role}: {', '.join(content_str)}")
+
+ prompt = "\n".join(formatted_messages)
+
+ input_tokens = response_dict.get('usage_metadata').get('prompt_token_count')
+ output_tokens = response_dict.get('usage_metadata').get('candidates_token_count')
+
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model,
+ pricing_info, input_tokens,
output_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
- if trace_content:
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
- },
- )
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: llmresponse,
- },
- )
-
- span.set_status(Status(StatusCode.OK))
-
- if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(
- total_tokens, attributes
- )
- metrics["genai_completion_tokens"].add(output_tokens, attributes)
- metrics["genai_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
-
- except Exception as e:
- handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
-
- return stream_generator()
- else:
- # pylint: disable=line-too-long
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- response = await wrapped(*args, **kwargs)
-
- try:
- prompt = ""
- for arg in args:
- if isinstance(arg, str):
- prompt = f"{prompt}{arg}\n"
- elif isinstance(arg, list):
- for subarg in arg:
- prompt = f"{prompt}{subarg}\n"
- if hasattr(instance, "_model_id"):
- model = instance._model_id
- if hasattr(instance, "_model_name"):
- model = instance._model_name.replace("publishers/google/models/", "")
- if model.startswith("models/"):
- model = model[len("models/"):]
-
- # Set base span attribues
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- False)
-
- if trace_content:
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
- },
- )
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.text,
- },
- )
-
- prompt_tokens = response.usage_metadata.prompt_token_count
- completion_tokens = response.usage_metadata.candidates_token_count
- total_tokens = response.usage_metadata.total_token_count
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, prompt_tokens, completion_tokens)
-
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- prompt_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
- completion_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
-
- span.set_status(Status(StatusCode.OK))
-
- if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(total_tokens, attributes)
- metrics["genai_completion_tokens"].add(completion_tokens, attributes)
- metrics["genai_prompt_tokens"].add(prompt_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
-
- # Return original response
- return response
-
- except Exception as e:
- handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
-
- # Return original response
- return response
+
+ # Set base span attribues (OTel Semconv)
+ span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
+ span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
+ SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_GEMINI)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+
+ inference_config = kwargs.get('config', {})
+
+ # List of attributes and their config keys
+ attributes = [
+ (SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY, 'frequency_penalty'),
+ (SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS, 'max_tokens'),
+ (SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY, 'presence_penalty'),
+ (SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES, 'stop_sequences'),
+ (SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE, 'temperature'),
+ (SemanticConvetion.GEN_AI_REQUEST_TOP_P, 'top_p'),
+ (SemanticConvetion.GEN_AI_REQUEST_TOP_K, 'top_k'),
+ ]
+
+ # Set each attribute if the corresponding value exists and is not None
+ for attribute, key in attributes:
+ # Use getattr to get the attribute value from the object
+ value = getattr(inference_config, key, None)
+ if value is not None:
+ span.set_attribute(attribute, value)
+
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ response_dict.get('model_version'))
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
+ input_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+ output_tokens)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
+ [str(response_dict.get('candidates')[0].get('finish_reason'))])
+
+ # Set base span attribues (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ False)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
+ input_tokens + output_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ end_time - start_time)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
+ if trace_content:
+ span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
+ },
+ )
+ span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.text,
+ },
+ )
+
+ if isinstance(response_dict.get('text'), str):
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "text")
+ elif response_dict.get('text') is not None:
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "json")
+
+ span.set_status(Status(StatusCode.OK))
+
+ if disable_metrics is False:
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_GEMINI,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=response_dict.get('model_version'),
+ )
+
+ metrics["genai_client_usage_tokens"].record(
+ input_tokens + output_tokens, attributes
+ )
+ metrics["genai_client_operation_duration"].record(
+ end_time - start_time, 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)
+
+ # Return original response
+ return response
+
+ except Exception as e:
+ handle_exception(span, e)
+ logger.error("Error in trace creation: %s", e)
+
+ # Return original response
+ return response
return wrapper
diff --git a/sdk/python/src/openlit/instrumentation/google_ai_studio/google_ai_studio.py b/sdk/python/src/openlit/instrumentation/google_ai_studio/google_ai_studio.py
index 3c2eb8ec..7a54b4cf 100644
--- a/sdk/python/src/openlit/instrumentation/google_ai_studio/google_ai_studio.py
+++ b/sdk/python/src/openlit/instrumentation/google_ai_studio/google_ai_studio.py
@@ -1,21 +1,24 @@
-# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument, possibly-used-before-assignment, protected-access
"""
Module for monitoring Google AI Studio API calls.
"""
import logging
+import time
from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
+from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
from openlit.__helpers import (
- handle_exception,
get_chat_model_cost,
+ handle_exception,
+ response_as_dict,
+ create_metrics_attributes,
+ set_server_address_and_port
)
from openlit.semcov import SemanticConvetion
# Initialize logger for logging potential issues and operations
logger = logging.getLogger(__name__)
-def generate(gen_ai_endpoint, version, environment, application_name,
+def generate(version, environment, application_name,
tracer, pricing_info, trace_content, metrics, disable_metrics):
"""
Generates a telemetry wrapper for chat to collect metrics.
@@ -35,230 +38,190 @@ def generate(gen_ai_endpoint, version, environment, application_name,
def wrapper(wrapped, instance, args, kwargs):
"""
- Wraps the 'chat' API call to add telemetry.
-
+ Wraps the 'chat.completions' API call to add telemetry.
+
This collects metrics such as execution time, cost, and token usage, and handles errors
gracefully, adding details to the trace for observability.
Args:
- wrapped: The original 'chat' method to be wrapped.
+ wrapped: The original 'chat.completions' method to be wrapped.
instance: The instance of the class where the original method is defined.
- args: Positional arguments for the 'chat' method.
- kwargs: Keyword arguments for the 'chat' method.
+ args: Positional arguments for the 'chat.completions' method.
+ kwargs: Keyword arguments for the 'chat.completions' method.
Returns:
- The response from the original 'chat' method.
+ The response from the original 'chat.completions' method.
"""
- # pylint: disable=no-else-return
- if kwargs.get("stream", False) is True:
- # Special handling for streaming response to accommodate the nature of data flow
- def stream_generator():
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- # Placeholder for aggregating streaming response
- llmresponse = ""
-
- # Loop through streaming events capturing relevant details
- for chunk in wrapped(*args, **kwargs):
- # Collect message IDs and aggregated response from events
- content = chunk.text
- if content:
- llmresponse += content
-
- input_tokens = chunk.usage_metadata.prompt_token_count
- output_tokens = chunk.usage_metadata.candidates_token_count
- yield chunk
-
- # Handling exception ensure observability without disrupting operation
- try:
- prompt = ""
- for arg in args:
- if isinstance(arg, str):
- prompt = f"{prompt}{arg}\n"
- elif isinstance(arg, list):
- for subarg in arg:
- prompt = f"{prompt}{subarg}\n"
- if hasattr(instance, "_model_id"):
- model = instance._model_id
- if hasattr(instance, "_model_name"):
- model = instance._model_name.replace("publishers/google/models/", "")
- if model.startswith("models/"):
- model = model[len("models/"):]
-
- total_tokens = input_tokens + output_tokens
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, input_tokens,
- output_tokens)
-
- # Set Span attributes
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- True)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- input_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+
+ server_address, server_port = set_server_address_and_port(instance, "generativelanguage.googleapis.com", 443)
+ request_model = kwargs.get("model", "gemini-2.0-flash")
+
+ span_name = f"{SemanticConvetion.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)
+ end_time = time.time()
+
+ response_dict = response_as_dict(response)
+
+ try:
+ # Format 'messages' into a single string
+ message_prompt = kwargs.get("contents", [])
+ formatted_messages = []
+
+ for content in message_prompt:
+ role = content.role
+ parts = content.parts
+ content_str = []
+
+ for part in parts:
+ # Collect relevant fields and handle each type of data that Part could contain
+ if part.text:
+ content_str.append(f"text: {part.text}")
+ if part.video_metadata:
+ content_str.append(f"video_metadata: {part.video_metadata}")
+ if part.thought:
+ content_str.append(f"thought: {part.thought}")
+ if part.code_execution_result:
+ content_str.append(f"code_execution_result: {part.code_execution_result}")
+ if part.executable_code:
+ content_str.append(f"executable_code: {part.executable_code}")
+ if part.file_data:
+ content_str.append(f"file_data: {part.file_data}")
+ if part.function_call:
+ content_str.append(f"function_call: {part.function_call}")
+ if part.function_response:
+ content_str.append(f"function_response: {part.function_response}")
+ if part.inline_data:
+ content_str.append(f"inline_data: {part.inline_data}")
+
+ formatted_messages.append(f"{role}: {', '.join(content_str)}")
+
+ prompt = "\n".join(formatted_messages)
+
+ input_tokens = response_dict.get('usage_metadata').get('prompt_token_count')
+ output_tokens = response_dict.get('usage_metadata').get('candidates_token_count')
+
+ # Calculate cost of the operation
+ cost = get_chat_model_cost(request_model,
+ pricing_info, input_tokens,
output_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
- if trace_content:
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
- },
- )
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: llmresponse,
- },
- )
-
- span.set_status(Status(StatusCode.OK))
-
- if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(
- total_tokens, attributes
- )
- metrics["genai_completion_tokens"].add(output_tokens, attributes)
- metrics["genai_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
-
- except Exception as e:
- handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
-
- return stream_generator()
- else:
- # pylint: disable=line-too-long
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
- response = wrapped(*args, **kwargs)
-
- try:
- prompt = ""
- for arg in args:
- if isinstance(arg, str):
- prompt = f"{prompt}{arg}\n"
- elif isinstance(arg, list):
- for subarg in arg:
- prompt = f"{prompt}{subarg}\n"
-
- if hasattr(instance, "_model_id"):
- model = instance._model_id
- if hasattr(instance, "_model_name"):
- model = instance._model_name.replace("publishers/google/models/", "")
- if model.startswith("models/"):
- model = model[len("models/"):]
-
- # Set base span attribues
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO)
- span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
- gen_ai_endpoint)
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
- environment)
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
- application_name)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
- model)
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
- False)
-
- if trace_content:
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
- },
- )
- span.add_event(
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
- attributes={
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.text,
- },
- )
-
- prompt_tokens = response.usage_metadata.prompt_token_count
- completion_tokens = response.usage_metadata.candidates_token_count
- total_tokens = response.usage_metadata.total_token_count
- # Calculate cost of the operation
- cost = get_chat_model_cost(model,
- pricing_info, prompt_tokens, completion_tokens)
-
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
- prompt_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
- completion_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
- total_tokens)
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
- cost)
-
- span.set_status(Status(StatusCode.OK))
-
- if disable_metrics is False:
- attributes = {
- TELEMETRY_SDK_NAME:
- "openlit",
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
- application_name,
- SemanticConvetion.GEN_AI_SYSTEM:
- SemanticConvetion.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO,
- SemanticConvetion.GEN_AI_ENVIRONMENT:
- environment,
- SemanticConvetion.GEN_AI_OPERATION:
- SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
- model
- }
-
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_total_tokens"].add(total_tokens, attributes)
- metrics["genai_completion_tokens"].add(completion_tokens, attributes)
- metrics["genai_prompt_tokens"].add(prompt_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
-
- # Return original response
- return response
-
- except Exception as e:
- handle_exception(span, e)
- logger.error("Error in trace creation: %s", e)
-
- # Return original response
- return response
+
+ # Set base span attribues (OTel Semconv)
+ span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
+ span.set_attribute(SemanticConvetion.GEN_AI_OPERATION,
+ SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
+ span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
+ SemanticConvetion.GEN_AI_SYSTEM_GEMINI)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
+ request_model)
+ span.set_attribute(SemanticConvetion.SERVER_PORT,
+ server_port)
+
+ inference_config = kwargs.get('config', {})
+
+ # List of attributes and their config keys
+ attributes = [
+ (SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY, 'frequency_penalty'),
+ (SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS, 'max_tokens'),
+ (SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY, 'presence_penalty'),
+ (SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES, 'stop_sequences'),
+ (SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE, 'temperature'),
+ (SemanticConvetion.GEN_AI_REQUEST_TOP_P, 'top_p'),
+ (SemanticConvetion.GEN_AI_REQUEST_TOP_K, 'top_k'),
+ ]
+
+ # Set each attribute if the corresponding value exists and is not None
+ for attribute, key in attributes:
+ # Use getattr to get the attribute value from the object
+ value = getattr(inference_config, key, None)
+ if value is not None:
+ span.set_attribute(attribute, value)
+
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL,
+ response_dict.get('model_version'))
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS,
+ input_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS,
+ output_tokens)
+ span.set_attribute(SemanticConvetion.SERVER_ADDRESS,
+ server_address)
+ span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
+ [str(response_dict.get('candidates')[0].get('finish_reason'))])
+
+ # Set base span attribues (Extras)
+ span.set_attribute(DEPLOYMENT_ENVIRONMENT,
+ environment)
+ span.set_attribute(SERVICE_NAME,
+ application_name)
+ span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
+ False)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
+ input_tokens + output_tokens)
+ span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
+ cost)
+ span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT,
+ end_time - start_time)
+ span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION,
+ version)
+ if trace_content:
+ span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
+ },
+ )
+ span.add_event(
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+ attributes={
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.text,
+ },
+ )
+
+ if isinstance(response_dict.get('text'), str):
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "text")
+ elif response_dict.get('text') is not None:
+ span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+ "json")
+
+ span.set_status(Status(StatusCode.OK))
+
+ if disable_metrics is False:
+ attributes = create_metrics_attributes(
+ service_name=application_name,
+ deployment_environment=environment,
+ operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+ system=SemanticConvetion.GEN_AI_SYSTEM_GEMINI,
+ request_model=request_model,
+ server_address=server_address,
+ server_port=server_port,
+ response_model=response_dict.get('model_version'),
+ )
+
+ metrics["genai_client_usage_tokens"].record(
+ input_tokens + output_tokens, attributes
+ )
+ metrics["genai_client_operation_duration"].record(
+ end_time - start_time, 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)
+
+ # Return original response
+ return response
+
+ except Exception as e:
+ handle_exception(span, e)
+ logger.error("Error in trace creation: %s", e)
+
+ # Return original response
+ return response
return wrapper
diff --git a/sdk/python/src/openlit/instrumentation/openai/async_openai.py b/sdk/python/src/openlit/instrumentation/openai/async_openai.py
index 19739159..5d1acde7 100644
--- a/sdk/python/src/openlit/instrumentation/openai/async_openai.py
+++ b/sdk/python/src/openlit/instrumentation/openai/async_openai.py
@@ -580,7 +580,7 @@ def async_embedding(version, environment, application_name,
request_model=request_model,
server_address=server_address,
server_port=server_port,
- response_model=response_dict.get('model'),
+ response_model=request_model,
)
metrics["genai_client_usage_tokens"].record(
input_tokens, attributes
diff --git a/sdk/python/src/openlit/instrumentation/openai/openai.py b/sdk/python/src/openlit/instrumentation/openai/openai.py
index 1ec61b2f..7c39adb2 100644
--- a/sdk/python/src/openlit/instrumentation/openai/openai.py
+++ b/sdk/python/src/openlit/instrumentation/openai/openai.py
@@ -580,7 +580,7 @@ def embedding(version, environment, application_name,
request_model=request_model,
server_address=server_address,
server_port=server_port,
- response_model=response_dict.get('model'),
+ response_model=request_model,
)
metrics["genai_client_usage_tokens"].record(
input_tokens, attributes
diff --git a/sdk/python/src/openlit/semcov/__init__.py b/sdk/python/src/openlit/semcov/__init__.py
index 4e09da3c..616722a5 100644
--- a/sdk/python/src/openlit/semcov/__init__.py
+++ b/sdk/python/src/openlit/semcov/__init__.py
@@ -98,7 +98,7 @@ class SemanticConvetion:
GEN_AI_SYSTEM_GPT4ALL = "gpt4all"
GEN_AI_SYSTEM_ELEVENLABS = "elevenlabs"
GEN_AI_SYSTEM_VLLM = "vLLM"
- GEN_AI_SYSTEM_GOOGLE_AI_STUDIO = "google-ai-studio"
+ GEN_AI_SYSTEM_GOOGLE_AI_STUDIO = "google.ai.studio"
GEN_AI_SYSTEM_REKAAI = "rekaai"
GEN_AI_SYSTEM_PREMAI = "premai"
GEN_AI_SYSTEM_LANGCHAIN = "langchain"