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: Align with OpenTelemetry GenAI Semconv (#1018)

Ishan Jain06d2a8718439
diff --git a/CODEOWNERS b/CODEOWNERS
index 8b0fd624..e1947054 100644
--- a/CODEOWNERS
+++ b/CODEOWNERS
@@ -1,2 +1,5 @@
# Default owners for everything in the repo, unless a later match takes precedence.
* @openlit/litters
+
+# Python SDK Codeowners
+/sdk/python @openlit/litters @ishanjainn
diff --git a/sdk/python/pyproject.toml b/sdk/python/pyproject.toml
index ddf0f6ff..4b105200 100644
--- a/sdk/python/pyproject.toml
+++ b/sdk/python/pyproject.toml
@@ -1,6 +1,6 @@
[tool.poetry]
name = "openlit"
-version = "1.37.0"
+version = "1.37.1"
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 25f7d3c5..2f19f553 100644
--- a/sdk/python/src/openlit/__helpers.py
+++ b/sdk/python/src/openlit/__helpers.py
@@ -214,12 +214,19 @@ def create_metrics_attributes(
server_address: str,
server_port: int,
response_model: str,
+ token_type: str = None,
+ error_type: str = None,
) -> Dict[Any, Any]:
"""
- Returns OTel metrics attributes
+ Returns OTel metrics attributes.
+
+ Args:
+ token_type: For gen_ai.client.token.usage metric only "input" and "output"
+ are allowed per OTel GenAI semconv; do not use "reasoning" or "total".
+ error_type: Optional error type for failed operations
"""
- return {
+ attributes = {
TELEMETRY_SDK_NAME: "openlit",
SERVICE_NAME: service_name,
DEPLOYMENT_ENVIRONMENT: deployment_environment,
@@ -231,6 +238,14 @@ def create_metrics_attributes(
SemanticConvention.GEN_AI_RESPONSE_MODEL: response_model,
}
+ # Add optional attributes for OTel compliance
+ if token_type:
+ attributes[SemanticConvention.GEN_AI_TOKEN_TYPE] = token_type
+ if error_type:
+ attributes[SemanticConvention.ERROR_TYPE] = error_type
+
+ return attributes
+
def create_db_metrics_attributes(
service_name: str,
@@ -485,12 +500,28 @@ def record_completion_metrics(
cost,
tbt,
ttft,
+ error_type=None,
+ is_stream=False,
+ time_per_chunk_observations=None,
):
"""
- Record completion metrics for the operation.
+ Record completion metrics for the operation with proper OTel token type attributes.
+
+ For gen_ai.client.token.usage, only gen_ai.token.type "input" and "output" are
+ used per OTel GenAI semantic conventions; reasoning/total must not be recorded
+ on this metric.
+
+ When the operation ended in an error, pass error_type so that
+ gen_ai.client.operation.duration is recorded with error.type (OTel semconv
+ conditionally required).
+
+ When is_stream is True and ttft is set, records gen_ai.client.operation.time_to_first_chunk.
+ When time_per_chunk_observations is provided (list of inter-chunk durations in seconds),
+ records each on gen_ai.client.operation.time_per_output_chunk (streaming only).
"""
- attributes = create_metrics_attributes(
+ # Base attributes without token type
+ base_attributes = create_metrics_attributes(
operation=gen_ai_operation,
system=GEN_AI_PROVIDER_NAME,
server_address=server_address,
@@ -499,17 +530,62 @@ def record_completion_metrics(
response_model=response_model,
service_name=application_name,
deployment_environment=environment,
+ error_type=error_type,
)
- metrics["genai_client_usage_tokens"].record(
- input_tokens + output_tokens, attributes
+
+ # Record token usage with proper token type (OTel compliant)
+ if input_tokens:
+ input_attributes = {
+ **base_attributes,
+ SemanticConvention.GEN_AI_TOKEN_TYPE: SemanticConvention.GEN_AI_TOKEN_TYPE_INPUT,
+ }
+ metrics["genai_client_usage_tokens"].record(input_tokens, input_attributes)
+
+ if output_tokens:
+ output_attributes = {
+ **base_attributes,
+ SemanticConvention.GEN_AI_TOKEN_TYPE: SemanticConvention.GEN_AI_TOKEN_TYPE_OUTPUT,
+ }
+ metrics["genai_client_usage_tokens"].record(output_tokens, output_attributes)
+
+ # Operation duration
+ metrics["genai_client_operation_duration"].record(
+ end_time - start_time, base_attributes
)
- metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
- metrics["genai_server_tbt"].record(tbt, attributes)
- metrics["genai_server_ttft"].record(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)
+
+ # Client streaming metrics (OTel: only for streaming)
+ if is_stream and ttft and "genai_client_time_to_first_chunk" in metrics:
+ metrics["genai_client_time_to_first_chunk"].record(ttft, base_attributes)
+ if time_per_chunk_observations and "genai_client_time_per_output_chunk" in metrics:
+ for duration_sec in time_per_chunk_observations:
+ metrics["genai_client_time_per_output_chunk"].record(
+ duration_sec, base_attributes
+ )
+
+ # Server metrics
+ if tbt:
+ metrics["genai_server_tbt"].record(tbt, base_attributes)
+ if ttft:
+ metrics["genai_server_ttft"].record(ttft, base_attributes)
+
+ # gen_ai.server.request.duration: server-side processing time
+ if "genai_server_request_duration" in metrics:
+ if (
+ ttft is not None
+ and tbt
+ and output_tokens is not None
+ and output_tokens >= 1
+ ):
+ server_duration = ttft + tbt * (output_tokens - 1)
+ else:
+ server_duration = end_time - start_time
+ metrics["genai_server_request_duration"].record(
+ server_duration, base_attributes
+ )
+
+ # Cost (OpenLIT vendor extension; not in OTel GenAI semconv)
+ if "genai_cost" in metrics:
+ metrics["genai_cost"].record(cost, base_attributes)
def record_embedding_metrics(
@@ -526,12 +602,18 @@ def record_embedding_metrics(
end_time,
input_tokens,
cost,
+ error_type=None,
):
"""
- Record embedding-specific metrics for the operation.
+ Record embedding-specific metrics for the operation with proper OTel token type.
+
+ For gen_ai.client.token.usage, only gen_ai.token.type "input" and "output" are
+ used per OTel GenAI semantic conventions.
+
+ When the operation ended in an error, pass error_type for gen_ai.client.operation.duration.
"""
- attributes = create_metrics_attributes(
+ base_attributes = create_metrics_attributes(
operation=gen_ai_operation,
system=GEN_AI_PROVIDER_NAME,
server_address=server_address,
@@ -540,12 +622,24 @@ def record_embedding_metrics(
response_model=response_model,
service_name=application_name,
deployment_environment=environment,
+ error_type=error_type,
)
- 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_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
+
+ # Record token usage with proper token type (OTel compliant)
+ if input_tokens:
+ input_attributes = {
+ **base_attributes,
+ SemanticConvention.GEN_AI_TOKEN_TYPE: SemanticConvention.GEN_AI_TOKEN_TYPE_INPUT,
+ }
+ metrics["genai_client_usage_tokens"].record(input_tokens, input_attributes)
+
+ metrics["genai_client_operation_duration"].record(
+ end_time - start_time, base_attributes
+ )
+
+ # Cost (OpenLIT vendor extension; not in OTel GenAI semconv)
+ if "genai_cost" in metrics:
+ metrics["genai_cost"].record(cost, base_attributes)
def record_audio_metrics(
@@ -577,8 +671,8 @@ def record_audio_metrics(
deployment_environment=environment,
)
metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ if "genai_cost" in metrics:
+ metrics["genai_cost"].record(cost, attributes)
def record_image_metrics(
@@ -610,8 +704,8 @@ def record_image_metrics(
deployment_environment=environment,
)
metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_cost"].record(cost, attributes)
+ if "genai_cost" in metrics:
+ metrics["genai_cost"].record(cost, attributes)
def common_db_span_attributes(
@@ -843,7 +937,6 @@ def record_framework_metrics(
service_name=application_name,
deployment_environment=environment,
)
- metrics["genai_requests"].add(1, attributes)
metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
diff --git a/sdk/python/src/openlit/instrumentation/ai21/ai21.py b/sdk/python/src/openlit/instrumentation/ai21/ai21.py
index d0340005..79b25630 100644
--- a/sdk/python/src/openlit/instrumentation/ai21/ai21.py
+++ b/sdk/python/src/openlit/instrumentation/ai21/ai21.py
@@ -7,6 +7,7 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
)
from openlit.instrumentation.ai21.utils import (
process_chunk,
@@ -156,6 +157,26 @@ def chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AI21,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -211,6 +232,26 @@ def chat_rag(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AI21,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/ai21/async_ai21.py b/sdk/python/src/openlit/instrumentation/ai21/async_ai21.py
index 26401f42..2ef8ac65 100644
--- a/sdk/python/src/openlit/instrumentation/ai21/async_ai21.py
+++ b/sdk/python/src/openlit/instrumentation/ai21/async_ai21.py
@@ -7,6 +7,7 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
)
from openlit.instrumentation.ai21.utils import (
process_chunk,
@@ -156,6 +157,26 @@ def async_chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AI21,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -211,6 +232,26 @@ def async_chat_rag(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AI21,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/ai21/utils.py b/sdk/python/src/openlit/instrumentation/ai21/utils.py
index 32f5042b..cef34601 100644
--- a/sdk/python/src/openlit/instrumentation/ai21/utils.py
+++ b/sdk/python/src/openlit/instrumentation/ai21/utils.py
@@ -486,6 +486,12 @@ def common_chat_logic(
# Metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -503,6 +509,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/anthropic/anthropic.py b/sdk/python/src/openlit/instrumentation/anthropic/anthropic.py
index c89704d0..e813e697 100644
--- a/sdk/python/src/openlit/instrumentation/anthropic/anthropic.py
+++ b/sdk/python/src/openlit/instrumentation/anthropic/anthropic.py
@@ -4,7 +4,11 @@ Module for monitoring Anthropic API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.anthropic.utils import (
process_chunk,
process_chat_response,
@@ -160,6 +164,26 @@ def messages(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_ANTHROPIC,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/anthropic/async_anthropic.py b/sdk/python/src/openlit/instrumentation/anthropic/async_anthropic.py
index f987145c..302a6329 100644
--- a/sdk/python/src/openlit/instrumentation/anthropic/async_anthropic.py
+++ b/sdk/python/src/openlit/instrumentation/anthropic/async_anthropic.py
@@ -4,7 +4,11 @@ Module for monitoring Anthropic API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.anthropic.utils import (
process_chunk,
process_chat_response,
@@ -160,6 +164,26 @@ def async_messages(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_ANTHROPIC,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/anthropic/utils.py b/sdk/python/src/openlit/instrumentation/anthropic/utils.py
index 0aaaf04a..753ed14b 100644
--- a/sdk/python/src/openlit/instrumentation/anthropic/utils.py
+++ b/sdk/python/src/openlit/instrumentation/anthropic/utils.py
@@ -665,6 +665,12 @@ def common_chat_logic(
# Record metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -682,6 +688,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
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 93f79007..23f68ccf 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
@@ -8,6 +8,8 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
)
from openlit.instrumentation.azure_ai_inference.utils import (
process_chunk,
@@ -138,23 +140,47 @@ def async_complete(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
response = await wrapped(*args, **kwargs)
- response = process_chat_response(
- response=response,
- request_model=request_model,
- pricing_info=pricing_info,
- server_port=server_port,
- server_address=server_address,
- environment=environment,
- application_name=application_name,
- metrics=metrics,
- start_time=start_time,
- span=span,
- capture_message_content=capture_message_content,
- disable_metrics=disable_metrics,
- version=version,
- event_provider=event_provider,
- **kwargs,
- )
+
+ try:
+ response = process_chat_response(
+ response=response,
+ request_model=request_model,
+ pricing_info=pricing_info,
+ server_port=server_port,
+ server_address=server_address,
+ environment=environment,
+ application_name=application_name,
+ metrics=metrics,
+ start_time=start_time,
+ span=span,
+ capture_message_content=capture_message_content,
+ disable_metrics=disable_metrics,
+ version=version,
+ event_provider=event_provider,
+ **kwargs,
+ )
+ except Exception as e:
+ handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -215,6 +241,23 @@ def async_embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return 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 704d35d3..3e048382 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
@@ -8,6 +8,8 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
)
from openlit.instrumentation.azure_ai_inference.utils import (
process_chunk,
@@ -138,23 +140,47 @@ def complete(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
response = wrapped(*args, **kwargs)
- response = process_chat_response(
- response=response,
- request_model=request_model,
- pricing_info=pricing_info,
- server_port=server_port,
- server_address=server_address,
- environment=environment,
- application_name=application_name,
- metrics=metrics,
- start_time=start_time,
- span=span,
- capture_message_content=capture_message_content,
- disable_metrics=disable_metrics,
- version=version,
- event_provider=event_provider,
- **kwargs,
- )
+
+ try:
+ response = process_chat_response(
+ response=response,
+ request_model=request_model,
+ pricing_info=pricing_info,
+ server_port=server_port,
+ server_address=server_address,
+ environment=environment,
+ application_name=application_name,
+ metrics=metrics,
+ start_time=start_time,
+ span=span,
+ capture_message_content=capture_message_content,
+ disable_metrics=disable_metrics,
+ version=version,
+ event_provider=event_provider,
+ **kwargs,
+ )
+ except Exception as e:
+ handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -215,6 +241,23 @@ def embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/azure_ai_inference/utils.py b/sdk/python/src/openlit/instrumentation/azure_ai_inference/utils.py
index 230fd479..9b515c90 100644
--- a/sdk/python/src/openlit/instrumentation/azure_ai_inference/utils.py
+++ b/sdk/python/src/openlit/instrumentation/azure_ai_inference/utils.py
@@ -672,6 +672,12 @@ def common_chat_logic(
# Metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -689,6 +695,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py b/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py
index 5dbf6b58..83034dee 100644
--- a/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py
+++ b/sdk/python/src/openlit/instrumentation/bedrock/bedrock.py
@@ -4,7 +4,11 @@ Module for monitoring Amazon Bedrock API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.bedrock.utils import (
process_chunk,
process_chat_response,
@@ -74,6 +78,26 @@ def converse(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_AWS_BEDROCK,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/bedrock/utils.py b/sdk/python/src/openlit/instrumentation/bedrock/utils.py
index ef23cd65..3a2a9a9a 100644
--- a/sdk/python/src/openlit/instrumentation/bedrock/utils.py
+++ b/sdk/python/src/openlit/instrumentation/bedrock/utils.py
@@ -572,6 +572,12 @@ def common_chat_logic(
# Record metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -589,6 +595,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py b/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py
index 1abd76c3..53ce83dc 100644
--- a/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py
+++ b/sdk/python/src/openlit/instrumentation/cohere/async_cohere.py
@@ -7,6 +7,8 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
)
from openlit.instrumentation.cohere.utils import (
process_chunk,
@@ -47,23 +49,47 @@ def async_chat(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
response = await wrapped(*args, **kwargs)
- response = process_chat_response(
- response=response,
- request_model=request_model,
- pricing_info=pricing_info,
- server_port=server_port,
- server_address=server_address,
- environment=environment,
- application_name=application_name,
- metrics=metrics,
- start_time=start_time,
- span=span,
- capture_message_content=capture_message_content,
- disable_metrics=disable_metrics,
- version=version,
- event_provider=event_provider,
- **kwargs,
- )
+
+ try:
+ response = process_chat_response(
+ response=response,
+ request_model=request_model,
+ pricing_info=pricing_info,
+ server_port=server_port,
+ server_address=server_address,
+ environment=environment,
+ application_name=application_name,
+ metrics=metrics,
+ start_time=start_time,
+ span=span,
+ capture_message_content=capture_message_content,
+ disable_metrics=disable_metrics,
+ version=version,
+ event_provider=event_provider,
+ **kwargs,
+ )
+ except Exception as e:
+ handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_COHERE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -248,6 +274,23 @@ def async_embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_COHERE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/cohere/cohere.py b/sdk/python/src/openlit/instrumentation/cohere/cohere.py
index 2b102987..78d97e39 100644
--- a/sdk/python/src/openlit/instrumentation/cohere/cohere.py
+++ b/sdk/python/src/openlit/instrumentation/cohere/cohere.py
@@ -7,6 +7,8 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
)
from openlit.instrumentation.cohere.utils import (
process_chunk,
@@ -47,23 +49,47 @@ def chat(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
response = wrapped(*args, **kwargs)
- response = process_chat_response(
- response=response,
- request_model=request_model,
- pricing_info=pricing_info,
- server_port=server_port,
- server_address=server_address,
- environment=environment,
- application_name=application_name,
- metrics=metrics,
- start_time=start_time,
- span=span,
- capture_message_content=capture_message_content,
- disable_metrics=disable_metrics,
- version=version,
- event_provider=event_provider,
- **kwargs,
- )
+
+ try:
+ response = process_chat_response(
+ response=response,
+ request_model=request_model,
+ pricing_info=pricing_info,
+ server_port=server_port,
+ server_address=server_address,
+ environment=environment,
+ application_name=application_name,
+ metrics=metrics,
+ start_time=start_time,
+ span=span,
+ capture_message_content=capture_message_content,
+ disable_metrics=disable_metrics,
+ version=version,
+ event_provider=event_provider,
+ **kwargs,
+ )
+ except Exception as e:
+ handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_COHERE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -248,6 +274,23 @@ def embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_COHERE,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/cohere/utils.py b/sdk/python/src/openlit/instrumentation/cohere/utils.py
index be9e0595..efb5bade 100644
--- a/sdk/python/src/openlit/instrumentation/cohere/utils.py
+++ b/sdk/python/src/openlit/instrumentation/cohere/utils.py
@@ -510,6 +510,12 @@ def common_chat_logic(
# Span status and metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -527,6 +533,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/crewai/utils.py b/sdk/python/src/openlit/instrumentation/crewai/utils.py
index b4baa1ce..89c27679 100644
--- a/sdk/python/src/openlit/instrumentation/crewai/utils.py
+++ b/sdk/python/src/openlit/instrumentation/crewai/utils.py
@@ -588,10 +588,6 @@ def _record_crewai_metrics(
duration_ms / 1000, attributes
)
- # Record operation count
- if "genai_requests" in metrics:
- metrics["genai_requests"].add(1, attributes)
-
except Exception:
# Graceful degradation
pass
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 aa16ea2d..faa84068 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
@@ -4,7 +4,11 @@ Module for monitoring Google AI Studio API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.google_ai_studio.utils import (
process_chat_response,
process_chunk,
@@ -67,6 +71,26 @@ def async_generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
# Return original response
return response
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 4046cd5e..aaf4a822 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
@@ -4,7 +4,11 @@ Module for monitoring Google AI Studio API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.google_ai_studio.utils import (
process_chat_response,
process_chunk,
@@ -67,6 +71,26 @@ def generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_GOOGLE_AI_STUDIO,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
# Return original response
return response
diff --git a/sdk/python/src/openlit/instrumentation/google_ai_studio/utils.py b/sdk/python/src/openlit/instrumentation/google_ai_studio/utils.py
index 14da6648..6f1babee 100644
--- a/sdk/python/src/openlit/instrumentation/google_ai_studio/utils.py
+++ b/sdk/python/src/openlit/instrumentation/google_ai_studio/utils.py
@@ -699,6 +699,12 @@ def common_chat_logic(
# Record metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -716,6 +722,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/gpt4all/gpt4all.py b/sdk/python/src/openlit/instrumentation/gpt4all/gpt4all.py
index 956e5ab6..325acf0e 100644
--- a/sdk/python/src/openlit/instrumentation/gpt4all/gpt4all.py
+++ b/sdk/python/src/openlit/instrumentation/gpt4all/gpt4all.py
@@ -4,7 +4,12 @@ Module for monitoring GPT4All API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.gpt4all.utils import (
process_generate_response,
process_chunk,
@@ -168,6 +173,26 @@ def generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_GPT4ALL,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -229,6 +254,23 @@ def embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_GPT4ALL,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/gpt4all/utils.py b/sdk/python/src/openlit/instrumentation/gpt4all/utils.py
index f57e13f4..70df031d 100644
--- a/sdk/python/src/openlit/instrumentation/gpt4all/utils.py
+++ b/sdk/python/src/openlit/instrumentation/gpt4all/utils.py
@@ -370,6 +370,12 @@ def common_generate_logic(
# Span status and metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -387,6 +393,8 @@ def common_generate_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/groq/async_groq.py b/sdk/python/src/openlit/instrumentation/groq/async_groq.py
index ab836700..d963f521 100644
--- a/sdk/python/src/openlit/instrumentation/groq/async_groq.py
+++ b/sdk/python/src/openlit/instrumentation/groq/async_groq.py
@@ -4,7 +4,11 @@ Module for monitoring Groq API calls (async version).
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.groq.utils import (
process_chunk,
process_streaming_chat_response,
@@ -154,6 +158,26 @@ def async_chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_GROQ,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/groq/groq.py b/sdk/python/src/openlit/instrumentation/groq/groq.py
index 43159664..87c0cde4 100644
--- a/sdk/python/src/openlit/instrumentation/groq/groq.py
+++ b/sdk/python/src/openlit/instrumentation/groq/groq.py
@@ -4,7 +4,11 @@ Module for monitoring Groq API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.groq.utils import (
process_chunk,
process_streaming_chat_response,
@@ -154,6 +158,26 @@ def chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_GROQ,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/groq/utils.py b/sdk/python/src/openlit/instrumentation/groq/utils.py
index a7c87bb1..5623a525 100644
--- a/sdk/python/src/openlit/instrumentation/groq/utils.py
+++ b/sdk/python/src/openlit/instrumentation/groq/utils.py
@@ -488,6 +488,12 @@ def common_chat_logic(
# Metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -505,6 +511,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/langgraph/async_langgraph.py b/sdk/python/src/openlit/instrumentation/langgraph/async_langgraph.py
index 9e2bf895..d59a1816 100644
--- a/sdk/python/src/openlit/instrumentation/langgraph/async_langgraph.py
+++ b/sdk/python/src/openlit/instrumentation/langgraph/async_langgraph.py
@@ -485,8 +485,5 @@ def _record_checkpoint_metrics(
if "genai_client_operation_duration" in metrics:
metrics["genai_client_operation_duration"].record(duration, attributes)
- if "genai_requests" in metrics:
- metrics["genai_requests"].add(1, attributes)
-
except Exception:
pass
diff --git a/sdk/python/src/openlit/instrumentation/langgraph/utils.py b/sdk/python/src/openlit/instrumentation/langgraph/utils.py
index 68d2e39a..5e06de19 100644
--- a/sdk/python/src/openlit/instrumentation/langgraph/utils.py
+++ b/sdk/python/src/openlit/instrumentation/langgraph/utils.py
@@ -659,9 +659,5 @@ def _record_langgraph_metrics(
if "genai_client_operation_duration" in metrics:
metrics["genai_client_operation_duration"].record(duration, attributes)
- # Record operation count
- if "genai_requests" in metrics:
- metrics["genai_requests"].add(1, attributes)
-
except Exception:
pass
diff --git a/sdk/python/src/openlit/instrumentation/litellm/async_litellm.py b/sdk/python/src/openlit/instrumentation/litellm/async_litellm.py
index 0ed38e2b..7265e6b9 100644
--- a/sdk/python/src/openlit/instrumentation/litellm/async_litellm.py
+++ b/sdk/python/src/openlit/instrumentation/litellm/async_litellm.py
@@ -4,7 +4,12 @@ Module for monitoring LiteLLM API calls (async version).
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.litellm.utils import (
process_chunk,
process_streaming_chat_response,
@@ -164,6 +169,26 @@ def acompletion(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_LITELLM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -223,6 +248,23 @@ def aembedding(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_LITELLM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/litellm/litellm.py b/sdk/python/src/openlit/instrumentation/litellm/litellm.py
index 05019cec..2513fc0d 100644
--- a/sdk/python/src/openlit/instrumentation/litellm/litellm.py
+++ b/sdk/python/src/openlit/instrumentation/litellm/litellm.py
@@ -4,7 +4,12 @@ Module for monitoring LiteLLM API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.litellm.utils import (
process_chunk,
process_streaming_chat_response,
@@ -164,6 +169,26 @@ def completion(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_LITELLM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -223,6 +248,23 @@ def embedding(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_LITELLM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/litellm/utils.py b/sdk/python/src/openlit/instrumentation/litellm/utils.py
index 74835662..b6db8b19 100644
--- a/sdk/python/src/openlit/instrumentation/litellm/utils.py
+++ b/sdk/python/src/openlit/instrumentation/litellm/utils.py
@@ -549,6 +549,12 @@ def common_chat_logic(
# Span status and metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -566,6 +572,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/mistral/async_mistral.py b/sdk/python/src/openlit/instrumentation/mistral/async_mistral.py
index 7ce46ae5..84bb21c0 100644
--- a/sdk/python/src/openlit/instrumentation/mistral/async_mistral.py
+++ b/sdk/python/src/openlit/instrumentation/mistral/async_mistral.py
@@ -7,6 +7,8 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
)
from openlit.instrumentation.mistral.utils import (
process_chunk,
@@ -47,23 +49,47 @@ def async_complete(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
response = await wrapped(*args, **kwargs)
- response = process_chat_response(
- response=response,
- request_model=request_model,
- pricing_info=pricing_info,
- server_port=server_port,
- server_address=server_address,
- environment=environment,
- application_name=application_name,
- metrics=metrics,
- start_time=start_time,
- span=span,
- capture_message_content=capture_message_content,
- disable_metrics=disable_metrics,
- version=version,
- event_provider=event_provider,
- **kwargs,
- )
+
+ try:
+ response = process_chat_response(
+ response=response,
+ request_model=request_model,
+ pricing_info=pricing_info,
+ server_port=server_port,
+ server_address=server_address,
+ environment=environment,
+ application_name=application_name,
+ metrics=metrics,
+ start_time=start_time,
+ span=span,
+ capture_message_content=capture_message_content,
+ disable_metrics=disable_metrics,
+ version=version,
+ event_provider=event_provider,
+ **kwargs,
+ )
+ except Exception as e:
+ handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_MISTRAL,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -239,6 +265,23 @@ def async_embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_MISTRAL,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/mistral/mistral.py b/sdk/python/src/openlit/instrumentation/mistral/mistral.py
index 3808f0b8..cff73f0e 100644
--- a/sdk/python/src/openlit/instrumentation/mistral/mistral.py
+++ b/sdk/python/src/openlit/instrumentation/mistral/mistral.py
@@ -7,6 +7,8 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
)
from openlit.instrumentation.mistral.utils import (
process_chunk,
@@ -47,23 +49,47 @@ def complete(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
response = wrapped(*args, **kwargs)
- response = process_chat_response(
- response=response,
- request_model=request_model,
- pricing_info=pricing_info,
- server_port=server_port,
- server_address=server_address,
- environment=environment,
- application_name=application_name,
- metrics=metrics,
- start_time=start_time,
- span=span,
- capture_message_content=capture_message_content,
- disable_metrics=disable_metrics,
- version=version,
- event_provider=event_provider,
- **kwargs,
- )
+
+ try:
+ response = process_chat_response(
+ response=response,
+ request_model=request_model,
+ pricing_info=pricing_info,
+ server_port=server_port,
+ server_address=server_address,
+ environment=environment,
+ application_name=application_name,
+ metrics=metrics,
+ start_time=start_time,
+ span=span,
+ capture_message_content=capture_message_content,
+ disable_metrics=disable_metrics,
+ version=version,
+ event_provider=event_provider,
+ **kwargs,
+ )
+ except Exception as e:
+ handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_MISTRAL,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -239,6 +265,23 @@ def embed(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_MISTRAL,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/mistral/utils.py b/sdk/python/src/openlit/instrumentation/mistral/utils.py
index 80fa2cd0..0dd3ce30 100644
--- a/sdk/python/src/openlit/instrumentation/mistral/utils.py
+++ b/sdk/python/src/openlit/instrumentation/mistral/utils.py
@@ -539,6 +539,12 @@ def common_chat_logic(
# Span status and metrics
scope._span.set_status(Status(StatusCode.OK))
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -556,6 +562,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/ollama/async_ollama.py b/sdk/python/src/openlit/instrumentation/ollama/async_ollama.py
index 331199d1..f919b49b 100644
--- a/sdk/python/src/openlit/instrumentation/ollama/async_ollama.py
+++ b/sdk/python/src/openlit/instrumentation/ollama/async_ollama.py
@@ -4,7 +4,12 @@ Module for monitoring Ollama API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.ollama.utils import (
process_chunk,
process_chat_response,
@@ -165,6 +170,26 @@ def async_chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OLLAMA,
+ server_address,
+ server_port,
+ request_model or "unknown",
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.monotonic(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -320,6 +345,26 @@ def async_generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TEXT_COMPLETION,
+ SemanticConvention.GEN_AI_SYSTEM_OLLAMA,
+ server_address,
+ server_port,
+ request_model or "unknown",
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.monotonic(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -381,6 +426,23 @@ def async_embeddings(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_OLLAMA,
+ server_address,
+ server_port,
+ request_model or "unknown",
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.monotonic(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/ollama/ollama.py b/sdk/python/src/openlit/instrumentation/ollama/ollama.py
index 21efc97f..a25d51a9 100644
--- a/sdk/python/src/openlit/instrumentation/ollama/ollama.py
+++ b/sdk/python/src/openlit/instrumentation/ollama/ollama.py
@@ -4,7 +4,12 @@ Module for monitoring Ollama API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.ollama.utils import (
process_chunk,
process_chat_response,
@@ -165,6 +170,26 @@ def chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OLLAMA,
+ server_address,
+ server_port,
+ request_model or "unknown",
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.monotonic(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -320,6 +345,26 @@ def generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TEXT_COMPLETION,
+ SemanticConvention.GEN_AI_SYSTEM_OLLAMA,
+ server_address,
+ server_port,
+ request_model or "unknown",
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.monotonic(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -381,6 +426,23 @@ def embeddings(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_OLLAMA,
+ server_address,
+ server_port,
+ request_model or "unknown",
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.monotonic(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/ollama/utils.py b/sdk/python/src/openlit/instrumentation/ollama/utils.py
index 3b1fa87a..79ce563a 100644
--- a/sdk/python/src/openlit/instrumentation/ollama/utils.py
+++ b/sdk/python/src/openlit/instrumentation/ollama/utils.py
@@ -15,9 +15,9 @@ from openlit.__helpers import (
general_tokens,
get_chat_model_cost,
get_embed_model_cost,
- create_metrics_attributes,
common_span_attributes,
record_completion_metrics,
+ record_embedding_metrics,
otel_event,
)
from openlit.semcov import SemanticConvention
@@ -283,42 +283,6 @@ def process_chunk(self, chunk):
self._finish_reason = chunked.get("done_reason", "")
-def record_embedding_metrics(
- metrics,
- gen_ai_operation,
- GEN_AI_PROVIDER_NAME,
- server_address,
- server_port,
- request_model,
- response_model,
- environment,
- application_name,
- start_time,
- end_time,
- cost,
- input_tokens,
-):
- """
- Record embedding metrics for the operation.
- """
-
- attributes = create_metrics_attributes(
- operation=gen_ai_operation,
- system=GEN_AI_PROVIDER_NAME,
- server_address=server_address,
- server_port=server_port,
- request_model=request_model,
- response_model=response_model,
- service_name=application_name,
- deployment_environment=environment,
- )
- 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_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_cost"].record(cost, attributes)
-
-
def common_chat_logic(
scope,
pricing_info,
@@ -485,6 +449,12 @@ def common_chat_logic(
# Span status and metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -502,6 +472,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
@@ -672,6 +644,12 @@ def common_generate_logic(
# Span status and metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_TEXT_COMPLETION,
@@ -689,6 +667,8 @@ def common_generate_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
@@ -792,8 +772,8 @@ def common_embedding_logic(
application_name,
scope._start_time,
scope._end_time,
- cost,
input_tokens,
+ cost,
)
diff --git a/sdk/python/src/openlit/instrumentation/openai/async_openai.py b/sdk/python/src/openlit/instrumentation/openai/async_openai.py
index e8bf22f8..d16ec586 100644
--- a/sdk/python/src/openlit/instrumentation/openai/async_openai.py
+++ b/sdk/python/src/openlit/instrumentation/openai/async_openai.py
@@ -4,7 +4,12 @@ Module for monitoring OpenAI API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ record_completion_metrics,
+ record_embedding_metrics,
+ set_server_address_and_port,
+)
from openlit.instrumentation.openai.utils import (
process_chat_chunk,
process_response_chunk,
@@ -159,6 +164,26 @@ def async_chat_completions(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ kwargs.get("model", "unknown"),
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -309,6 +334,26 @@ def async_responses(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -367,6 +412,26 @@ def async_chat_completions_parse(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -426,6 +491,23 @@ def async_embedding(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -485,6 +567,26 @@ def async_image_generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ end_time,
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -544,6 +646,26 @@ def async_image_variations(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ end_time,
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -603,6 +725,26 @@ def async_audio_create(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ end_time,
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/openai/openai.py b/sdk/python/src/openlit/instrumentation/openai/openai.py
index 83f03a81..c09275c5 100644
--- a/sdk/python/src/openlit/instrumentation/openai/openai.py
+++ b/sdk/python/src/openlit/instrumentation/openai/openai.py
@@ -4,7 +4,12 @@ Module for monitoring OpenAI API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.openai.utils import (
process_chat_chunk,
process_response_chunk,
@@ -159,6 +164,27 @@ def chat_completions(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ err_type = type(e).__name__ or "_OTHER"
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ kwargs.get("model", "unknown"),
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=err_type,
+ )
return response
@@ -310,6 +336,27 @@ def responses(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ err_type = type(e).__name__ or "_OTHER"
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ kwargs.get("model", "unknown"),
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=err_type,
+ )
return response
@@ -368,6 +415,26 @@ def chat_completions_parse(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -429,6 +496,23 @@ def embedding(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -490,6 +574,26 @@ def image_generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ end_time,
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -550,6 +654,26 @@ def image_variatons(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ end_time,
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -610,6 +734,26 @@ def audio_create(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ SemanticConvention.GEN_AI_SYSTEM_OPENAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ end_time,
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/openai/utils.py b/sdk/python/src/openlit/instrumentation/openai/utils.py
index 39da9303..a916493f 100644
--- a/sdk/python/src/openlit/instrumentation/openai/utils.py
+++ b/sdk/python/src/openlit/instrumentation/openai/utils.py
@@ -974,6 +974,12 @@ def common_response_logic(
# Record metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -991,6 +997,8 @@ def common_response_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
@@ -1389,6 +1397,12 @@ def common_chat_logic(
# Record metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -1406,6 +1420,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/premai/premai.py b/sdk/python/src/openlit/instrumentation/premai/premai.py
index 6743fc63..fbb4824f 100644
--- a/sdk/python/src/openlit/instrumentation/premai/premai.py
+++ b/sdk/python/src/openlit/instrumentation/premai/premai.py
@@ -4,7 +4,12 @@ Module for monitoring PremAI API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+ record_embedding_metrics,
+)
from openlit.instrumentation.premai.utils import (
process_chat_response,
process_chunk,
@@ -152,6 +157,26 @@ def chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_PREMAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -210,6 +235,23 @@ def embedding(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_embedding_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ SemanticConvention.GEN_AI_SYSTEM_PREMAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/premai/utils.py b/sdk/python/src/openlit/instrumentation/premai/utils.py
index 5cb18788..9e618a31 100644
--- a/sdk/python/src/openlit/instrumentation/premai/utils.py
+++ b/sdk/python/src/openlit/instrumentation/premai/utils.py
@@ -515,6 +515,12 @@ def common_chat_logic(
# Span status and metrics
scope._span.set_status(Status(StatusCode.OK))
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -532,6 +538,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/reka/async_reka.py b/sdk/python/src/openlit/instrumentation/reka/async_reka.py
index b6e07994..d965bbc0 100644
--- a/sdk/python/src/openlit/instrumentation/reka/async_reka.py
+++ b/sdk/python/src/openlit/instrumentation/reka/async_reka.py
@@ -4,7 +4,11 @@ Module for monitoring Reka API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.reka.utils import process_chat_response
from openlit.semcov import SemanticConvention
@@ -61,6 +65,26 @@ def async_chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_REKAAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/reka/reka.py b/sdk/python/src/openlit/instrumentation/reka/reka.py
index 1670a181..dbb70b16 100644
--- a/sdk/python/src/openlit/instrumentation/reka/reka.py
+++ b/sdk/python/src/openlit/instrumentation/reka/reka.py
@@ -4,7 +4,11 @@ Module for monitoring Reka API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.reka.utils import process_chat_response
from openlit.semcov import SemanticConvention
@@ -61,6 +65,26 @@ def chat(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_REKAAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/reka/utils.py b/sdk/python/src/openlit/instrumentation/reka/utils.py
index dc4f2e8b..60d31590 100644
--- a/sdk/python/src/openlit/instrumentation/reka/utils.py
+++ b/sdk/python/src/openlit/instrumentation/reka/utils.py
@@ -436,6 +436,12 @@ def common_chat_logic(
# Metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -453,6 +459,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/sarvam/async_sarvam.py b/sdk/python/src/openlit/instrumentation/sarvam/async_sarvam.py
index 7dedc9da..dd9895d3 100644
--- a/sdk/python/src/openlit/instrumentation/sarvam/async_sarvam.py
+++ b/sdk/python/src/openlit/instrumentation/sarvam/async_sarvam.py
@@ -5,7 +5,11 @@ Module for monitoring async Sarvam AI API calls.
import time
from opentelemetry import context as context_api
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.sarvam.utils import (
process_chunk,
process_chat_response,
@@ -167,6 +171,26 @@ def async_chat_completions(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -229,6 +253,26 @@ def async_text_translate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TRANSLATE,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -291,6 +335,26 @@ def async_speech_to_text_transcribe(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -353,6 +417,26 @@ def async_text_to_speech_convert(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TEXT_TO_SPEECH,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -415,6 +499,26 @@ def async_text_transliterate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TRANSLITERATE,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -477,6 +581,26 @@ def async_text_identify_language(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_LANGUAGE_IDENTIFICATION,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -537,6 +661,26 @@ def async_speech_to_text_translate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT_TRANSLATE,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/sarvam/sarvam.py b/sdk/python/src/openlit/instrumentation/sarvam/sarvam.py
index ca5eb550..62dffd30 100644
--- a/sdk/python/src/openlit/instrumentation/sarvam/sarvam.py
+++ b/sdk/python/src/openlit/instrumentation/sarvam/sarvam.py
@@ -7,6 +7,7 @@ from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
set_server_address_and_port,
+ record_completion_metrics,
)
from openlit.semcov import SemanticConvention
from openlit.instrumentation.sarvam.utils import (
@@ -169,6 +170,26 @@ def chat_completions(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -231,6 +252,26 @@ def text_translate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TRANSLATE,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -293,6 +334,26 @@ def speech_to_text_transcribe(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -355,6 +416,26 @@ def text_to_speech_convert(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TEXT_TO_SPEECH,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -417,6 +498,26 @@ def text_transliterate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_TRANSLITERATE,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -479,6 +580,26 @@ def text_identify_language(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_LANGUAGE_IDENTIFICATION,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -539,6 +660,26 @@ def speech_to_text_translate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT_TRANSLATE,
+ SemanticConvention.GEN_AI_SYSTEM_SARVAM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/sarvam/utils.py b/sdk/python/src/openlit/instrumentation/sarvam/utils.py
index 869797f5..73638ad7 100644
--- a/sdk/python/src/openlit/instrumentation/sarvam/utils.py
+++ b/sdk/python/src/openlit/instrumentation/sarvam/utils.py
@@ -455,6 +455,12 @@ def common_chat_logic(
# Span status and metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -472,6 +478,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/together/async_together.py b/sdk/python/src/openlit/instrumentation/together/async_together.py
index d901c346..029ea28e 100644
--- a/sdk/python/src/openlit/instrumentation/together/async_together.py
+++ b/sdk/python/src/openlit/instrumentation/together/async_together.py
@@ -4,7 +4,11 @@ Module for monitoring Together API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.together.utils import (
process_chat_response,
process_chunk,
@@ -152,6 +156,26 @@ def async_completion(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_TOGETHER,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -212,6 +236,26 @@ def async_image_generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ SemanticConvention.GEN_AI_SYSTEM_TOGETHER,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
return wrapper
diff --git a/sdk/python/src/openlit/instrumentation/together/together.py b/sdk/python/src/openlit/instrumentation/together/together.py
index bc241327..2e6c24e2 100644
--- a/sdk/python/src/openlit/instrumentation/together/together.py
+++ b/sdk/python/src/openlit/instrumentation/together/together.py
@@ -4,7 +4,11 @@ Module for monitoring Together API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.together.utils import (
process_chat_response,
process_chunk,
@@ -152,6 +156,26 @@ def completion(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_TOGETHER,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
@@ -212,6 +236,26 @@ def image_generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ SemanticConvention.GEN_AI_SYSTEM_TOGETHER,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
return wrapper
diff --git a/sdk/python/src/openlit/instrumentation/together/utils.py b/sdk/python/src/openlit/instrumentation/together/utils.py
index 4e929c95..901a721f 100644
--- a/sdk/python/src/openlit/instrumentation/together/utils.py
+++ b/sdk/python/src/openlit/instrumentation/together/utils.py
@@ -17,7 +17,7 @@ from openlit.__helpers import (
calculate_tbt,
get_chat_model_cost,
get_image_model_cost,
- create_metrics_attributes,
+ record_completion_metrics,
)
from openlit.semcov import SemanticConvention
@@ -131,33 +131,33 @@ def record_common_metrics(
cost,
tbt=None,
ttft=None,
+ is_stream=False,
+ time_per_chunk_observations=None,
):
"""
Record common metrics for the operation.
+ Delegates to the OTel-compliant helper function.
"""
-
- attributes = create_metrics_attributes(
- operation=gen_ai_operation,
- system=GEN_AI_PROVIDER_NAME,
+ record_completion_metrics(
+ metrics=metrics,
+ gen_ai_operation=gen_ai_operation,
+ GEN_AI_PROVIDER_NAME=GEN_AI_PROVIDER_NAME,
server_address=server_address,
server_port=server_port,
request_model=request_model,
response_model=response_model,
- service_name=application_name,
- deployment_environment=environment,
- )
- metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
- metrics["genai_requests"].add(1, attributes)
- metrics["genai_prompt_tokens"].add(input_tokens, attributes)
- metrics["genai_completion_tokens"].add(output_tokens, attributes)
- metrics["genai_client_usage_tokens"].record(
- input_tokens + output_tokens, attributes
+ environment=environment,
+ application_name=application_name,
+ start_time=start_time,
+ end_time=end_time,
+ input_tokens=input_tokens,
+ output_tokens=output_tokens,
+ cost=cost,
+ tbt=tbt,
+ ttft=ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=time_per_chunk_observations,
)
- metrics["genai_cost"].record(cost, attributes)
- if tbt is not None:
- metrics["genai_server_tbt"].record(tbt, attributes)
- if ttft is not None:
- metrics["genai_server_ttft"].record(ttft, attributes)
def common_chat_logic(
@@ -288,6 +288,12 @@ def common_chat_logic(
# Metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_common_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -305,6 +311,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/vertexai/async_vertexai.py b/sdk/python/src/openlit/instrumentation/vertexai/async_vertexai.py
index a90fb5bf..ccb2a7c4 100644
--- a/sdk/python/src/openlit/instrumentation/vertexai/async_vertexai.py
+++ b/sdk/python/src/openlit/instrumentation/vertexai/async_vertexai.py
@@ -7,6 +7,7 @@ import time
from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
+ record_completion_metrics,
)
from openlit.instrumentation.vertexai.utils import (
process_chunk,
@@ -158,6 +159,26 @@ def async_send_message(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_VERTEXAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/vertexai/utils.py b/sdk/python/src/openlit/instrumentation/vertexai/utils.py
index ba0deb44..15463446 100644
--- a/sdk/python/src/openlit/instrumentation/vertexai/utils.py
+++ b/sdk/python/src/openlit/instrumentation/vertexai/utils.py
@@ -479,6 +479,12 @@ def common_chat_logic(
# Record metrics
if not disable_metrics:
+ inter_chunk_durations = None
+ if getattr(scope, "_timestamps", None) and len(scope._timestamps) > 1:
+ inter_chunk_durations = [
+ scope._timestamps[i] - scope._timestamps[i - 1]
+ for i in range(1, len(scope._timestamps))
+ ]
record_completion_metrics(
metrics,
SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
@@ -496,6 +502,8 @@ def common_chat_logic(
cost,
scope._tbt,
scope._ttft,
+ is_stream=is_stream,
+ time_per_chunk_observations=inter_chunk_durations,
)
diff --git a/sdk/python/src/openlit/instrumentation/vertexai/vertexai.py b/sdk/python/src/openlit/instrumentation/vertexai/vertexai.py
index c77d37f3..9ec97345 100644
--- a/sdk/python/src/openlit/instrumentation/vertexai/vertexai.py
+++ b/sdk/python/src/openlit/instrumentation/vertexai/vertexai.py
@@ -7,6 +7,7 @@ import time
from opentelemetry.trace import SpanKind
from openlit.__helpers import (
handle_exception,
+ record_completion_metrics,
)
from openlit.instrumentation.vertexai.utils import (
process_chunk,
@@ -158,6 +159,26 @@ def send_message(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_VERTEXAI,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/instrumentation/vllm/vllm.py b/sdk/python/src/openlit/instrumentation/vllm/vllm.py
index ddba1ec0..675a84ac 100644
--- a/sdk/python/src/openlit/instrumentation/vllm/vllm.py
+++ b/sdk/python/src/openlit/instrumentation/vllm/vllm.py
@@ -4,7 +4,11 @@ Module for monitoring vLLM API calls.
import time
from opentelemetry.trace import SpanKind
-from openlit.__helpers import handle_exception, set_server_address_and_port
+from openlit.__helpers import (
+ handle_exception,
+ set_server_address_and_port,
+ record_completion_metrics,
+)
from openlit.instrumentation.vllm.utils import process_chat_response
from openlit.semcov import SemanticConvention
@@ -62,6 +66,26 @@ def generate(
except Exception as e:
handle_exception(span, e)
+ if not disable_metrics and metrics:
+ record_completion_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ SemanticConvention.GEN_AI_SYSTEM_VLLM,
+ server_address,
+ server_port,
+ request_model,
+ "unknown",
+ environment,
+ application_name,
+ start_time,
+ time.time(),
+ 0,
+ 0,
+ 0,
+ None,
+ None,
+ error_type=type(e).__name__ or "_OTHER",
+ )
return response
diff --git a/sdk/python/src/openlit/otel/metrics.py b/sdk/python/src/openlit/otel/metrics.py
index fd2c6445..db066ea8 100644
--- a/sdk/python/src/openlit/otel/metrics.py
+++ b/sdk/python/src/openlit/otel/metrics.py
@@ -133,6 +133,13 @@ _MCP_PAYLOAD_SIZE_BUCKETS = [
5242880, # 5MB
]
+# Client-side streaming metrics buckets (OTel semconv: same as operation.duration)
+_GEN_AI_CLIENT_OPERATION_TIME_TO_FIRST_CHUNK = _GEN_AI_CLIENT_OPERATION_DURATION_BUCKETS
+
+_GEN_AI_CLIENT_OPERATION_TIME_PER_OUTPUT_CHUNK = (
+ _GEN_AI_CLIENT_OPERATION_DURATION_BUCKETS
+)
+
def setup_meter(application_name, environment, meter, otlp_endpoint, otlp_headers):
"""
@@ -249,36 +256,33 @@ def setup_meter(application_name, environment, meter, otlp_endpoint, otlp_header
unit="s",
explicit_bucket_boundaries_advisory=_GEN_AI_SERVER_TFTT,
),
+ "genai_client_time_to_first_chunk": meter.create_histogram(
+ name=SemanticConvention.GEN_AI_CLIENT_OPERATION_TIME_TO_FIRST_CHUNK,
+ description="Time from client request to first response chunk (OTel compliant)",
+ unit="s",
+ explicit_bucket_boundaries_advisory=_GEN_AI_CLIENT_OPERATION_TIME_TO_FIRST_CHUNK,
+ ),
+ "genai_client_time_per_output_chunk": meter.create_histogram(
+ name=SemanticConvention.GEN_AI_CLIENT_OPERATION_TIME_PER_OUTPUT_CHUNK,
+ description="Time between consecutive response chunks from client perspective (OTel compliant)",
+ unit="s",
+ explicit_bucket_boundaries_advisory=_GEN_AI_CLIENT_OPERATION_TIME_PER_OUTPUT_CHUNK,
+ ),
+ "genai_server_request_duration": meter.create_histogram(
+ name=SemanticConvention.GEN_AI_SERVER_REQUEST_DURATION,
+ description="Total server-side processing time from request receipt to response transmission",
+ unit="s",
+ explicit_bucket_boundaries_advisory=_GEN_AI_CLIENT_OPERATION_DURATION_BUCKETS,
+ ),
"db_client_operation_duration": meter.create_histogram(
name=SemanticConvention.DB_CLIENT_OPERATION_DURATION,
description="DB operation duration",
unit="s",
explicit_bucket_boundaries_advisory=_DB_CLIENT_OPERATION_DURATION_BUCKETS,
),
- # Extra
- "genai_requests": meter.create_counter(
- name=SemanticConvention.GEN_AI_REQUESTS,
- description="Number of requests to GenAI",
- unit="1",
- ),
- "genai_prompt_tokens": meter.create_counter(
- name=SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS,
- description="Number of prompt tokens processed.",
- unit="1",
- ),
- "genai_completion_tokens": meter.create_counter(
- name=SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS,
- description="Number of completion tokens processed.",
- unit="1",
- ),
- "genai_reasoning_tokens": meter.create_counter(
- name=SemanticConvention.GEN_AI_USAGE_REASONING_TOKENS,
- description="Number of reasoning thought tokens processed.",
- unit="1",
- ),
"genai_cost": meter.create_histogram(
name=SemanticConvention.GEN_AI_USAGE_COST,
- description="The distribution of GenAI request costs.",
+ description="The distribution of GenAI request costs (OpenLIT vendor extension; not part of OTel GenAI semconv).",
unit="USD",
),
"db_requests": meter.create_counter(
diff --git a/sdk/python/src/openlit/semcov/__init__.py b/sdk/python/src/openlit/semcov/__init__.py
index 2664ffe3..1ece3e94 100644
--- a/sdk/python/src/openlit/semcov/__init__.py
+++ b/sdk/python/src/openlit/semcov/__init__.py
@@ -84,10 +84,35 @@ class SemanticConvention:
GEN_AI_CLIENT_TOKEN_USAGE = "gen_ai.client.token.usage"
DB_CLIENT_TOKEN_USAGE = "db.client.token.usage"
GEN_AI_CLIENT_OPERATION_DURATION = "gen_ai.client.operation.duration"
+ GEN_AI_CLIENT_OPERATION_TIME_TO_FIRST_CHUNK = (
+ "gen_ai.client.operation.time_to_first_chunk"
+ )
+ GEN_AI_CLIENT_OPERATION_TIME_PER_OUTPUT_CHUNK = (
+ "gen_ai.client.operation.time_per_output_chunk"
+ )
GEN_AI_SERVER_REQUEST_DURATION = "gen_ai.server.request.duration"
GEN_AI_SERVER_TBT = "gen_ai.server.time_per_output_token"
GEN_AI_SERVER_TTFT = "gen_ai.server.time_to_first_token"
+ # GenAI Token Type Attribute (OTel Semconv - required for token usage metric)
+ GEN_AI_TOKEN_TYPE = "gen_ai.token.type"
+ GEN_AI_TOKEN_TYPE_INPUT = "input"
+ GEN_AI_TOKEN_TYPE_OUTPUT = "output"
+ # For spans/attributes only; do NOT use for gen_ai.client.token.usage metric
+ # (semconv allows only "input" and "output" on that metric).
+ GEN_AI_TOKEN_TYPE_REASONING = "reasoning"
+
+ # Error Type Values (OTel Semconv - conditional for operation failures)
+ ERROR_TYPE_UNAVAILABLE = "unavailable"
+ ERROR_TYPE_AUTHENTICATION = "authentication"
+ ERROR_TYPE_TIMEOUT = "timeout"
+ ERROR_TYPE_RATE_LIMITED = "rate_limited"
+ ERROR_TYPE_PERMISSION = "permission"
+ ERROR_TYPE_NOT_FOUND = "not_found"
+ ERROR_TYPE_INVALID_REQUEST = "invalid_request"
+ ERROR_TYPE_SERVER_ERROR = "server_error"
+ ERROR_TYPE_UNAUTHORIZED = "unauthorized"
+
# GenAI Event Names (OTel Semconv)
GEN_AI_USER_MESSAGE = "gen_ai.user.message"
GEN_AI_SYSTEM_MESSAGE = "gen_ai.system.message"
@@ -174,12 +199,6 @@ class SemanticConvention:
GEN_AI_RESPONSE_MODEL = _get_otel_attr(
OTelGenAIAttributes, "GEN_AI_RESPONSE_MODEL", "gen_ai.response.model"
)
- GEN_AI_USAGE_INPUT_TOKENS = _get_otel_attr(
- OTelGenAIAttributes, "GEN_AI_USAGE_INPUT_TOKENS", "gen_ai.usage.input_tokens"
- )
- GEN_AI_USAGE_OUTPUT_TOKENS = _get_otel_attr(
- OTelGenAIAttributes, "GEN_AI_USAGE_OUTPUT_TOKENS", "gen_ai.usage.output_tokens"
- )
GEN_AI_USAGE_CACHE_CREATION_INPUT_TOKENS = _get_otel_attr(
OTelGenAIAttributes,
"GEN_AI_USAGE_CACHE_CREATION_INPUT_TOKENS",
@@ -439,7 +458,6 @@ class SemanticConvention:
GEN_AI_HUB_OWNER = "gen_ai.hub.owner"
GEN_AI_HUB_REPO = "gen_ai.hub.repo"
GEN_AI_RETRIEVAL_SOURCE = "gen_ai.retrieval.source"
- GEN_AI_REQUESTS = "gen_ai.total.requests"
GEN_AI_DATA_SOURCES = "gen_ai.data_source_count"
GEN_AI_ENVIRONMENT = "gen_ai.environment"
GEN_AI_APPLICATION_NAME = "gen_ai.application_name"
@@ -454,8 +472,11 @@ class SemanticConvention:
# GenAI Response Attributes (Extra/Custom)
GEN_AI_USAGE_TOTAL_TOKENS = "gen_ai.usage.total_tokens"
- GEN_AI_USAGE_COST = "gen_ai.usage.cost"
- GEN_AI_USAGE_REASONING_TOKENS = "gen_ai.usage.reasoning_tokens"
+ GEN_AI_USAGE_INPUT_TOKENS = "gen_ai.usage.input_tokens"
+ GEN_AI_USAGE_OUTPUT_TOKENS = "gen_ai.usage.output_tokens"
+ GEN_AI_USAGE_COST = (
+ "gen_ai.usage.cost" # OpenLIT vendor extension; not OTel GenAI semconv
+ )
GEN_AI_RESPONSE_IMAGE = "gen_ai.response.image"
GEN_AI_TOOL_CALLS = "gen_ai.response.tool_calls"