Skip to content

Feature: Instrument OpenAI's new Responses API #2782 #2787

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
tongyu0924 wants to merge 3 commits into from
+88 −13
Closed

Feature: Instrument OpenAI's new Responses API #2782 #2787

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ef0d1a3
Update assistant_wrappers.py
tongyu0924 Mar 18, 2025
c8858fa
Update __init__.py
tongyu0924 Mar 18, 2025
d730cbe
Update assistant_wrappers.py
tongyu0924 Mar 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions .../opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/v1/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,7 @@
runs_retrieve_wrapper,
runs_create_and_stream_wrapper,
messages_list_wrapper,
responses_retrieve_wrapper,
)

from opentelemetry.instrumentation.openai.utils import is_metrics_enabled
Expand Down Expand Up @@ -243,6 +244,11 @@ def _instrument(self, **kwargs):
"Messages.list",
messages_list_wrapper(tracer),
)
wrap_function_wrapper(
"openai.resources.responses",
"Responses.retrieve",
responses_retrieve_wrapper(tracer),
)
except (AttributeError, ModuleNotFoundError):
pass

Expand Down
95 changes: 82 additions & 13 deletions ...etry-instrumentation-openai/opentelemetry/instrumentation/openai/v1/assistant_wrappers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,13 +15,46 @@

from openai._legacy_response import LegacyAPIResponse
from openai.types.beta.threads.run import Run
from openai.types.responses import Response

logger = logging.getLogger(__name__)

assistants = {}
runs = {}


@_with_tracer_wrapper
def responses_retrieve_wrapper(tracer, wrapped, instance, args, kwargs):
@dont_throw
def process_response(response):
if isinstance(response, LegacyAPIResponse):
parsed_response = response.parse()
else:
parsed_response = response
assert isinstance(parsed_response, Response)

response_id = parsed_response.id
response_text = parsed_response.content

span = tracer.start_span(
"openai.response.retrieve",
kind=SpanKind.CLIENT,
attributes={SpanAttributes.LLM_REQUEST_TYPE: LLMRequestTypeValues.CHAT.value},
)

_set_span_attribute(span, "gen_ai.response.id", response_id)
_set_span_attribute(span, "gen_ai.response.content", response_text)
span.end()

if context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY):
return wrapped(*args, **kwargs)

response = wrapped(*args, **kwargs)
process_response(response)

return response


@_with_tracer_wrapper
def assistants_create_wrapper(tracer, wrapped, instance, args, kwargs):
if context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY):
Expand All @@ -46,11 +79,13 @@ def runs_create_wrapper(tracer, wrapped, instance, args, kwargs):
instructions = kwargs.get("instructions")

response = wrapped(*args, **kwargs)
response_dict = model_as_dict(response)

runs[thread_id] = {
"start_time": time.time_ns(),
"assistant_id": kwargs.get("assistant_id"),
"instructions": instructions,
"run_id": response_dict.get("id"),
}

return response
Expand All @@ -66,13 +101,15 @@ def process_response(response):
parsed_response = response
assert type(parsed_response) is Run

if parsed_response.id in runs:
if parsed_response.thread_id in runs:
thread_id = parsed_response.thread_id
runs[thread_id]["end_time"] = time.time_ns()
if parsed_response.usage:
runs[thread_id]["usage"] = parsed_response.usage

if context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY):
return wrapped(*args, **kwargs)

thread_id = kwargs.get("thread_id")
response = wrapped(*args, **kwargs)
process_response(response)

Expand Down Expand Up @@ -102,7 +139,7 @@ def messages_list_wrapper(tracer, wrapped, instance, args, kwargs):
start_time=run.get("start_time"),
)

i = 0
prompt_index = 0
if assistants.get(run["assistant_id"]) is not None or Config.enrich_assistant:
if Config.enrich_assistant:
assistant = model_as_dict(
Expand All @@ -112,6 +149,11 @@ def messages_list_wrapper(tracer, wrapped, instance, args, kwargs):
else:
assistant = assistants[run["assistant_id"]]

_set_span_attribute(
span,
SpanAttributes.LLM_SYSTEM,
"openai",
)
_set_span_attribute(
span,
SpanAttributes.LLM_REQUEST_MODEL,
Expand All @@ -122,25 +164,47 @@ def messages_list_wrapper(tracer, wrapped, instance, args, kwargs):
SpanAttributes.LLM_RESPONSE_MODEL,
assistant["model"],
)
_set_span_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{i}.role", "system")
_set_span_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{prompt_index}.role", "system")
_set_span_attribute(
span,
f"{SpanAttributes.LLM_PROMPTS}.{i}.content",
f"{SpanAttributes.LLM_PROMPTS}.{prompt_index}.content",
assistant["instructions"],
)
i += 1
_set_span_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{i}.role", "system")
prompt_index += 1
_set_span_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{prompt_index}.role", "system")
_set_span_attribute(
span, f"{SpanAttributes.LLM_PROMPTS}.{i}.content", run["instructions"]
span, f"{SpanAttributes.LLM_PROMPTS}.{prompt_index}.content", run["instructions"]
)
prompt_index += 1

for i, msg in enumerate(messages):
prefix = f"{SpanAttributes.LLM_COMPLETIONS}.{i}"
completion_index = 0
for msg in messages:
prefix = f"{SpanAttributes.LLM_COMPLETIONS}.{completion_index}"
content = msg.get("content")

_set_span_attribute(span, f"{prefix}.role", msg.get("role"))
message_content = content[0].get("text").get("value")
message_role = msg.get("role")
if message_role in ["user", "system"]:
_set_span_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{prompt_index}.role", message_role)
_set_span_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{prompt_index}.content", message_content)
prompt_index += 1
else:
_set_span_attribute(span, f"{prefix}.role", msg.get("role"))
_set_span_attribute(span, f"{prefix}.content", message_content)
_set_span_attribute(span, f"gen_ai.response.{completion_index}.id", msg.get("id"))
completion_index += 1

if run.get("usage"):
usage_dict = model_as_dict(run.get("usage"))
_set_span_attribute(
span, f"{prefix}.content", content[0].get("text").get("value")
span,
SpanAttributes.LLM_USAGE_COMPLETION_TOKENS,
usage_dict.get("completion_tokens"),
)
_set_span_attribute(
span,
SpanAttributes.LLM_USAGE_PROMPT_TOKENS,
usage_dict.get("prompt_tokens"),
)

span.end(run.get("end_time"))
Expand Down Expand Up @@ -175,6 +239,11 @@ def runs_create_and_stream_wrapper(tracer, wrapped, instance, args, kwargs):
_set_span_attribute(
span, SpanAttributes.LLM_REQUEST_MODEL, assistants[assistant_id]["model"]
)
_set_span_attribute(
span,
SpanAttributes.LLM_SYSTEM,
"openai",
)
_set_span_attribute(
span,
SpanAttributes.LLM_RESPONSE_MODEL,
Expand All @@ -195,7 +264,7 @@ def runs_create_and_stream_wrapper(tracer, wrapped, instance, args, kwargs):
)

kwargs["event_handler"] = EventHandleWrapper(
original_handler=kwargs["event_handler"], span=span
original_handler=kwargs["event_handler"], span=span,
)

response = wrapped(*args, **kwargs)
Expand Down