Skip to content

ref(openai): Move input handling code into API-specific functions #5687

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

Draft
alexander-alderman-webb wants to merge 5 commits into
base: master
Choose a base branch
from
Draft

ref(openai): Move input handling code into API-specific functions #5687

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4917b6a
ref(openai): Move input handling code into API-specific functions
alexander-alderman-webb Mar 18, 2026
4261546
keep set_data_normalized
alexander-alderman-webb Mar 18, 2026
1f6149e
.
alexander-alderman-webb Mar 18, 2026
36842e0
.
alexander-alderman-webb Mar 18, 2026
532fa65
.
alexander-alderman-webb Mar 18, 2026
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
155 changes: 116 additions & 39 deletions sentry_sdk/integrations/openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -219,28 +219,13 @@
)


def _commmon_set_input_data(
def _set_responses_api_input_data(
span: "Span",
kwargs: "dict[str, Any]",
integration: "OpenAIIntegration",
) -> None:
# Input attributes: Common
set_data_normalized(span, SPANDATA.GEN_AI_SYSTEM, "openai")

# Input attributes: Optional
kwargs_keys_to_attributes = {
"model": SPANDATA.GEN_AI_REQUEST_MODEL,
"stream": SPANDATA.GEN_AI_RESPONSE_STREAMING,
"max_tokens": SPANDATA.GEN_AI_REQUEST_MAX_TOKENS,
"presence_penalty": SPANDATA.GEN_AI_REQUEST_PRESENCE_PENALTY,
"frequency_penalty": SPANDATA.GEN_AI_REQUEST_FREQUENCY_PENALTY,
"temperature": SPANDATA.GEN_AI_REQUEST_TEMPERATURE,
"top_p": SPANDATA.GEN_AI_REQUEST_TOP_P,
}
for key, attribute in kwargs_keys_to_attributes.items():
value = kwargs.get(key)

if value is not None and _is_given(value):
set_data_normalized(span, attribute, value)
explicit_instructions: "Union[Optional[str], Omit]" = kwargs.get("instructions")
messages: "Optional[Union[str, ResponseInputParam]]" = kwargs.get("input")

# Input attributes: Tools
tools = kwargs.get("tools")
Expand All @@ -249,18 +234,40 @@
span, SPANDATA.GEN_AI_REQUEST_AVAILABLE_TOOLS, safe_serialize(tools)
)

model = kwargs.get("model")
if model is not None and _is_given(model):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_MODEL, model)

def _set_responses_api_input_data(
span: "Span",
kwargs: "dict[str, Any]",
integration: "OpenAIIntegration",
) -> None:
explicit_instructions: "Union[Optional[str], Omit]" = kwargs.get("instructions")
messages: "Optional[Union[str, ResponseInputParam]]" = kwargs.get("input")
stream = kwargs.get("stream")
if stream is not None and _is_given(stream):
set_data_normalized(span, SPANDATA.GEN_AI_RESPONSE_STREAMING, stream)

max_tokens = kwargs.get("max_tokens")
if max_tokens is not None and _is_given(max_tokens):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_MAX_TOKENS, max_tokens)

presence_penalty = kwargs.get("presence_penalty")
if presence_penalty is not None and _is_given(presence_penalty):
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_PRESENCE_PENALTY, presence_penalty
)

frequency_penalty = kwargs.get("frequency_penalty")
if frequency_penalty is not None and _is_given(frequency_penalty):
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_FREQUENCY_PENALTY, frequency_penalty
)

temperature = kwargs.get("temperature")
if temperature is not None and _is_given(temperature):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_TEMPERATURE, temperature)

top_p = kwargs.get("top_p")
if top_p is not None and _is_given(top_p):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_TOP_P, top_p)

if not should_send_default_pii() or not integration.include_prompts:
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")
_commmon_set_input_data(span, kwargs)
return

if (
Expand All @@ -281,12 +288,10 @@
)

set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")
_commmon_set_input_data(span, kwargs)
return

if messages is None:
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")
_commmon_set_input_data(span, kwargs)
return

instructions_text_parts: "list[TextPart]" = []
Expand Down Expand Up @@ -319,7 +324,6 @@
)

set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")
_commmon_set_input_data(span, kwargs)
return

non_system_messages = [
Expand All @@ -335,7 +339,6 @@
)

set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")
_commmon_set_input_data(span, kwargs)


def _set_completions_api_input_data(
Expand All @@ -347,14 +350,56 @@
"messages"
)

# Input attributes: Tools
tools = kwargs.get("tools")
if tools is not None and _is_given(tools) and len(tools) > 0:
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_AVAILABLE_TOOLS, safe_serialize(tools)
)

model = kwargs.get("model")
if model is not None and _is_given(model):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_MODEL, model)

stream = kwargs.get("stream")
if stream is not None and _is_given(stream):
set_data_normalized(span, SPANDATA.GEN_AI_RESPONSE_STREAMING, stream)

max_tokens = kwargs.get("max_tokens")
if max_tokens is not None and _is_given(max_tokens):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_MAX_TOKENS, max_tokens)

presence_penalty = kwargs.get("presence_penalty")
if presence_penalty is not None and _is_given(presence_penalty):
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_PRESENCE_PENALTY, presence_penalty
)

frequency_penalty = kwargs.get("frequency_penalty")
if frequency_penalty is not None and _is_given(frequency_penalty):
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_FREQUENCY_PENALTY, frequency_penalty
)

temperature = kwargs.get("temperature")
if temperature is not None and _is_given(temperature):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_TEMPERATURE, temperature)

top_p = kwargs.get("top_p")
if top_p is not None and _is_given(top_p):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_TOP_P, top_p)

if not should_send_default_pii() or not integration.include_prompts:
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")

Check failure on line 393 in sentry_sdk/integrations/openai.py

View check run for this annotation

@sentry/warden / warden: code-review

[8XJ-QM9] Embeddings function incorrectly sets operation name to 'responses' instead of 'embeddings' (additional location) 

In `_set_embeddings_input_data`, when PII is disabled or prompts are excluded, the operation name is incorrectly set to 'responses' (line 489) instead of 'embeddings'. This was likely a copy-paste error from `_set_responses_api_input_data`. This will result in incorrect telemetry data being reported for embedding operations.

Check warning on line 393 in sentry_sdk/integrations/openai.py

View check run for this annotation

@sentry/warden / warden: find-bugs

Completions API sets incorrect operation name 'responses' instead of 'chat' 

In `_set_completions_api_input_data`, when PII is not enabled or prompts are not included, the span's operation name is incorrectly set to 'responses' instead of 'chat'. This causes incorrect telemetry data, as the completions API operations will be mislabeled as responses API operations.
return

if (
not should_send_default_pii()
or not integration.include_prompts
or messages is None
):
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "chat")
_commmon_set_input_data(span, kwargs)
return

Check warning on line 402 in sentry_sdk/integrations/openai.py

View check run for this annotation

@sentry/warden / warden: code-review

Unreachable code due to duplicate early return 

In `_set_completions_api_input_data`, the condition at lines 396-400 checks `not should_send_default_pii() or not integration.include_prompts or messages is None`, but the first two conditions are already checked and return early at lines 392-394. This makes the first two conditions in the second check unreachable. If `messages is None` was intended to be checked, it should be a separate check for just `messages is None`.

if isinstance(messages, str):
normalized_messages = normalize_message_roles([messages]) # type: ignore
Expand All @@ -365,13 +410,11 @@
span, SPANDATA.GEN_AI_REQUEST_MESSAGES, messages_data, unpack=False
)
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "chat")
_commmon_set_input_data(span, kwargs)
return

# dict special case following https://github.com/openai/openai-python/blob/3e0c05b84a2056870abf3bd6a5e7849020209cc3/src/openai/_utils/_transform.py#L194-L197
if not isinstance(messages, Iterable) or isinstance(messages, dict):
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "chat")
_commmon_set_input_data(span, kwargs)
return

messages = list(messages)
Expand Down Expand Up @@ -399,7 +442,6 @@
)

set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "chat")
_commmon_set_input_data(span, kwargs)


def _set_embeddings_input_data(
Expand All @@ -411,19 +453,53 @@
"input"
)

model = kwargs.get("model")
if model is not None and _is_given(model):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_MODEL, model)

stream = kwargs.get("stream")
if stream is not None and _is_given(stream):
set_data_normalized(span, SPANDATA.GEN_AI_RESPONSE_STREAMING, stream)

max_tokens = kwargs.get("max_tokens")
if max_tokens is not None and _is_given(max_tokens):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_MAX_TOKENS, max_tokens)

presence_penalty = kwargs.get("presence_penalty")
if presence_penalty is not None and _is_given(presence_penalty):
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_PRESENCE_PENALTY, presence_penalty
)

frequency_penalty = kwargs.get("frequency_penalty")
if frequency_penalty is not None and _is_given(frequency_penalty):
set_data_normalized(
span, SPANDATA.GEN_AI_REQUEST_FREQUENCY_PENALTY, frequency_penalty
)

temperature = kwargs.get("temperature")
if temperature is not None and _is_given(temperature):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_TEMPERATURE, temperature)

top_p = kwargs.get("top_p")
if top_p is not None and _is_given(top_p):
set_data_normalized(span, SPANDATA.GEN_AI_REQUEST_TOP_P, top_p)

if not should_send_default_pii() or not integration.include_prompts:
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "responses")

Check failure on line 489 in sentry_sdk/integrations/openai.py

View check run for this annotation

@sentry/warden / warden: code-review

Embeddings function incorrectly sets operation name to 'responses' instead of 'embeddings' 

In `_set_embeddings_input_data`, when PII is disabled or prompts are excluded, the operation name is incorrectly set to 'responses' (line 489) instead of 'embeddings'. This was likely a copy-paste error from `_set_responses_api_input_data`. This will result in incorrect telemetry data being reported for embedding operations.
return

Check warning on line 490 in sentry_sdk/integrations/openai.py

View check run for this annotation

@sentry/warden / warden: find-bugs

[DBV-8E9] Completions API sets incorrect operation name 'responses' instead of 'chat' (additional location) 

In `_set_completions_api_input_data`, when PII is not enabled or prompts are not included, the span's operation name is incorrectly set to 'responses' instead of 'chat'. This causes incorrect telemetry data, as the completions API operations will be mislabeled as responses API operations.

if (
not should_send_default_pii()
or not integration.include_prompts
or messages is None
):
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "embeddings")
_commmon_set_input_data(span, kwargs)

return

if isinstance(messages, str):
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "embeddings")
_commmon_set_input_data(span, kwargs)

normalized_messages = normalize_message_roles([messages]) # type: ignore
scope = sentry_sdk.get_current_scope()
Expand All @@ -440,7 +516,6 @@
# dict special case following https://github.com/openai/openai-python/blob/3e0c05b84a2056870abf3bd6a5e7849020209cc3/src/openai/_utils/_transform.py#L194-L197
if not isinstance(messages, Iterable) or isinstance(messages, dict):
set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "embeddings")
_commmon_set_input_data(span, kwargs)
return

messages = list(messages)
Expand All @@ -458,7 +533,6 @@
)

set_data_normalized(span, SPANDATA.GEN_AI_OPERATION_NAME, "embeddings")
_commmon_set_input_data(span, kwargs)


def _set_common_output_data(
Expand Down Expand Up @@ -551,6 +625,7 @@
)
span.__enter__()

span.set_data(SPANDATA.GEN_AI_SYSTEM, "openai")
_set_completions_api_input_data(span, kwargs, integration)

start_time = time.perf_counter()
Expand Down Expand Up @@ -945,6 +1020,7 @@
name=f"embeddings {model}",
origin=OpenAIIntegration.origin,
) as span:
span.set_data(SPANDATA.GEN_AI_SYSTEM, "openai")
_set_embeddings_input_data(span, kwargs, integration)

response = yield f, args, kwargs
Expand Down Expand Up @@ -1036,6 +1112,7 @@
)
span.__enter__()

span.set_data(SPANDATA.GEN_AI_SYSTEM, "openai")
_set_responses_api_input_data(span, kwargs, integration)

start_time = time.perf_counter()
Expand Down
14 changes: 14 additions & 0 deletions tests/integrations/openai/test_openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -161,6 +161,7 @@ def test_nonstreaming_chat_completion_no_prompts(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert SPANDATA.GEN_AI_SYSTEM_INSTRUCTIONS not in span["data"]
assert SPANDATA.GEN_AI_REQUEST_MESSAGES not in span["data"]
Expand Down Expand Up @@ -240,6 +241,7 @@ def test_nonstreaming_chat_completion(sentry_init, capture_events, messages, req
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

param_id = request.node.callspec.id
if "blocks" in param_id:
Expand Down Expand Up @@ -306,6 +308,7 @@ async def test_nonstreaming_chat_completion_async_no_prompts(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert SPANDATA.GEN_AI_SYSTEM_INSTRUCTIONS not in span["data"]
assert SPANDATA.GEN_AI_REQUEST_MESSAGES not in span["data"]
Expand Down Expand Up @@ -385,6 +388,7 @@ async def test_nonstreaming_chat_completion_async(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

param_id = request.node.callspec.id
if "blocks" in param_id:
Expand Down Expand Up @@ -503,6 +507,7 @@ def test_streaming_chat_completion_no_prompts(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert span["data"][SPANDATA.GEN_AI_RESPONSE_MODEL] == "model-id"

Expand Down Expand Up @@ -630,6 +635,7 @@ def test_streaming_chat_completion(sentry_init, capture_events, messages, reques
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

param_id = request.node.callspec.id
if "blocks" in param_id:
Expand Down Expand Up @@ -758,6 +764,7 @@ async def test_streaming_chat_completion_async_no_prompts(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert span["data"][SPANDATA.GEN_AI_RESPONSE_MODEL] == "model-id"

Expand Down Expand Up @@ -895,6 +902,7 @@ async def test_streaming_chat_completion_async(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.chat"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert span["data"][SPANDATA.GEN_AI_RESPONSE_MODEL] == "model-id"

Expand Down Expand Up @@ -1036,6 +1044,7 @@ def test_embeddings_create_no_pii(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.embeddings"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert SPANDATA.GEN_AI_EMBEDDINGS_INPUT not in span["data"]

Expand Down Expand Up @@ -1116,6 +1125,7 @@ def test_embeddings_create(sentry_init, capture_events, input, request):
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.embeddings"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

param_id = request.node.callspec.id
if param_id == "string":
Expand Down Expand Up @@ -1187,6 +1197,7 @@ async def test_embeddings_create_async_no_pii(
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.embeddings"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert SPANDATA.GEN_AI_EMBEDDINGS_INPUT not in span["data"]

Expand Down Expand Up @@ -1270,6 +1281,7 @@ async def test_embeddings_create_async(sentry_init, capture_events, input, reque
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "gen_ai.embeddings"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

param_id = request.node.callspec.id
if param_id == "string":
Expand Down Expand Up @@ -2772,6 +2784,7 @@ def test_streaming_responses_api(
(transaction,) = events
(span,) = transaction["spans"]
assert span["op"] == "gen_ai.responses"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert span["data"][SPANDATA.GEN_AI_RESPONSE_MODEL] == "response-model-id"

Expand Down Expand Up @@ -2829,6 +2842,7 @@ async def test_streaming_responses_api_async(
(transaction,) = events
(span,) = transaction["spans"]
assert span["op"] == "gen_ai.responses"
assert span["data"][SPANDATA.GEN_AI_SYSTEM] == "openai"

assert span["data"][SPANDATA.GEN_AI_RESPONSE_MODEL] == "response-model-id"

Expand Down
Loading