Skip to content

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

fix: OpenAI and TS Langchain instrumentations (#1141)

Aman Agarwal2fd83fb970f4
diff --git a/sdk/python/src/openlit/instrumentation/openai/async_openai.py b/sdk/python/src/openlit/instrumentation/openai/async_openai.py
index 43cd4840..b80c5c4a 100644
--- a/sdk/python/src/openlit/instrumentation/openai/async_openai.py
+++ b/sdk/python/src/openlit/instrumentation/openai/async_openai.py
@@ -361,6 +361,17 @@ def async_responses(
if streaming:
span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ True,
+ version,
+ )
ctx = trace_api.set_span_in_context(span)
token = context_api.attach(ctx)
try:
@@ -385,7 +396,42 @@ def async_responses(
else:
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ err_type = set_span_error_type(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=err_type,
+ )
+ raise
try:
response = process_response_response(
@@ -463,7 +509,42 @@ def async_chat_completions_parse(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ err_type = set_span_error_type(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=err_type,
+ )
+ raise
try:
response = process_chat_response(
@@ -543,7 +624,39 @@ def async_embedding(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ err_type = set_span_error_type(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=err_type,
+ )
+ raise
try:
response = process_embedding_response(
@@ -618,7 +731,22 @@ def async_image_generate(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -698,7 +826,22 @@ def async_image_variations(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -778,7 +921,22 @@ def async_audio_create(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -856,7 +1014,22 @@ def async_audio_transcription(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -914,7 +1087,22 @@ def async_audio_translation(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -970,7 +1158,22 @@ def async_image_edit(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -1028,7 +1231,22 @@ def async_moderation(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_MODERATION,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -1083,7 +1301,22 @@ def _make_async_lightweight_wrapper(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = await wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ operation_type,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
diff --git a/sdk/python/src/openlit/instrumentation/openai/openai.py b/sdk/python/src/openlit/instrumentation/openai/openai.py
index 62525a1d..e88b37b3 100644
--- a/sdk/python/src/openlit/instrumentation/openai/openai.py
+++ b/sdk/python/src/openlit/instrumentation/openai/openai.py
@@ -363,6 +363,17 @@ def responses(
if streaming:
span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ True,
+ version,
+ )
ctx = trace_api.set_span_in_context(span)
token = context_api.attach(ctx)
try:
@@ -387,7 +398,42 @@ def responses(
else:
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ err_type = set_span_error_type(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=err_type,
+ )
+ raise
try:
response = process_response_response(
@@ -466,7 +512,42 @@ def chat_completions_parse(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ err_type = set_span_error_type(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=err_type,
+ )
+ raise
try:
response = process_chat_response(
@@ -547,7 +628,39 @@ def embedding(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ err_type = set_span_error_type(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=err_type,
+ )
+ raise
try:
response = process_embedding_response(
@@ -623,7 +736,22 @@ def image_generate(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -703,7 +831,22 @@ def image_variatons(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -783,7 +926,22 @@ def audio_create(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -861,7 +1019,22 @@ def audio_transcription(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -919,7 +1092,22 @@ def audio_translation(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -975,7 +1163,22 @@ def image_edit(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_IMAGE,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -1033,7 +1236,22 @@ def moderation(
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_MODERATION,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
@@ -1086,7 +1304,22 @@ def _make_lightweight_wrapper(span_prefix, operation_type, default_model="unknow
with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
start_time = time.time()
- response = wrapped(*args, **kwargs)
+ set_openai_request_span_attributes(
+ span,
+ operation_type,
+ server_address,
+ server_port,
+ request_model,
+ environment,
+ application_name,
+ False,
+ version,
+ )
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ set_span_error_type(span, e)
+ raise
end_time = time.time()
try:
diff --git a/sdk/python/src/openlit/instrumentation/openai/utils.py b/sdk/python/src/openlit/instrumentation/openai/utils.py
index 89d6cd99..f76337aa 100644
--- a/sdk/python/src/openlit/instrumentation/openai/utils.py
+++ b/sdk/python/src/openlit/instrumentation/openai/utils.py
@@ -14,6 +14,7 @@ from opentelemetry.sdk.resources import (
from opentelemetry.trace import Status, StatusCode
from openlit.__helpers import (
+ _apply_custom_span_attributes,
calculate_ttft,
response_as_dict,
calculate_tbt,
@@ -69,10 +70,13 @@ def set_openai_request_span_attributes(
if value is not None:
span.set_attribute(key, value)
+ _apply_custom_span_attributes(span)
+
def set_span_error_type(span, error):
- """Set error status and OTel error.type without recording duplicate events."""
+ """Record exception, set error status and OTel error.type on the span."""
+ span.record_exception(error)
error_type = type(error).__name__ or "_OTHER"
span.set_status(Status(StatusCode.ERROR))
span.set_attribute(SemanticConvention.ERROR_TYPE, error_type)
@@ -1012,7 +1016,8 @@ def common_response_logic(
# Reasoning tokens
if hasattr(scope, "_reasoning_tokens") and scope._reasoning_tokens > 0:
scope._span.set_attribute(
- "gen_ai.usage.reasoning_tokens", scope._reasoning_tokens
+ SemanticConvention.GEN_AI_USAGE_REASONING_TOKENS,
+ scope._reasoning_tokens,
)
# OTel cached token attributes (set even when 0)