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(python-sdk): add Groq audio instrumentation (#1326)

Aditya Jethanib7dee1ca06d1
diff --git a/assets/pricing.json b/assets/pricing.json
index 6aeabdcb..692cecf3 100644
--- a/assets/pricing.json
+++ b/assets/pricing.json
@@ -122,7 +122,11 @@
"eleven_english_sts_v2": 0.24,
"eleven_multilingual_sts_v2": 0.24,
"best": 0.00010277777,
- "nano": 0.00003333333
+ "nano": 0.00003333333,
+ "whisper-large-v3": 0.00003083333,
+ "whisper-large-v3-turbo": 0.00001111111,
+ "canopylabs/orpheus-v1-english": 0.022,
+ "canopylabs/orpheus-arabic-saudi": 0.040
},
"chat": {
"gpt-4.1-nano-2025-04-14": {
diff --git a/sdk/python/src/openlit/instrumentation/groq/__init__.py b/sdk/python/src/openlit/instrumentation/groq/__init__.py
index d3db33f3..06b7831c 100644
--- a/sdk/python/src/openlit/instrumentation/groq/__init__.py
+++ b/sdk/python/src/openlit/instrumentation/groq/__init__.py
@@ -1,5 +1,6 @@
"""Initializer of Auto Instrumentation of Groq Functions"""
+import logging
from typing import Collection
import importlib.metadata
from opentelemetry import _logs, trace
@@ -7,12 +8,34 @@ from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
from wrapt import wrap_function_wrapper
from openlit._config import OpenlitConfig
-from openlit.instrumentation.groq.groq import chat
-from openlit.instrumentation.groq.async_groq import async_chat
+from openlit.instrumentation.groq.groq import (
+ chat,
+ audio_create,
+ audio_transcription,
+ audio_translation,
+)
+from openlit.instrumentation.groq.async_groq import (
+ async_chat,
+ async_audio_create,
+ async_audio_transcription,
+ async_audio_translation,
+)
+
+logger = logging.getLogger(__name__)
_instruments = ("groq >= 0.5.0",)
+def _safe_wrap(module, class_method, wrapper):
+ """Wrap a function, silently skipping if the target isn't present in this SDK version."""
+ try:
+ wrap_function_wrapper(module, class_method, wrapper)
+ except (ModuleNotFoundError, AttributeError):
+ logger.debug(
+ "Skipping %s.%s — not available in this groq version", module, class_method
+ )
+
+
class GroqInstrumentor(BaseInstrumentor):
"""
An instrumentor for Groq client library.
@@ -66,5 +89,107 @@ class GroqInstrumentor(BaseInstrumentor):
),
)
+ # Audio speech (text-to-speech)
+ _safe_wrap(
+ "groq.resources.audio.speech",
+ "Speech.create",
+ audio_create(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider,
+ ),
+ )
+
+ # Audio speech (text-to-speech)
+ _safe_wrap(
+ "groq.resources.audio.speech",
+ "AsyncSpeech.create",
+ async_audio_create(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider,
+ ),
+ )
+
+ # Audio transcriptions (speech-to-text)
+ _safe_wrap(
+ "groq.resources.audio.transcriptions",
+ "Transcriptions.create",
+ audio_transcription(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider,
+ ),
+ )
+
+ # Audio transcriptions (speech-to-text)
+ _safe_wrap(
+ "groq.resources.audio.transcriptions",
+ "AsyncTranscriptions.create",
+ async_audio_transcription(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider,
+ ),
+ )
+
+ # Audio translations (speech-to-text)
+ _safe_wrap(
+ "groq.resources.audio.translations",
+ "Translations.create",
+ audio_translation(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider,
+ ),
+ )
+
+ # Audio translations (speech-to-text)
+ _safe_wrap(
+ "groq.resources.audio.translations",
+ "AsyncTranslations.create",
+ async_audio_translation(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider,
+ ),
+ )
+
def _uninstrument(self, **kwargs):
pass
diff --git a/sdk/python/src/openlit/instrumentation/groq/async_groq.py b/sdk/python/src/openlit/instrumentation/groq/async_groq.py
index 18e742fc..c741a9b1 100644
--- a/sdk/python/src/openlit/instrumentation/groq/async_groq.py
+++ b/sdk/python/src/openlit/instrumentation/groq/async_groq.py
@@ -14,6 +14,8 @@ from openlit.instrumentation.groq.utils import (
process_chunk,
process_streaming_chat_response,
process_chat_response,
+ process_audio_response,
+ process_transcription_response,
)
from openlit.semcov import SemanticConvention
@@ -191,3 +193,199 @@ def async_chat(
return response
return wrapper
+
+
+def async_audio_create(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider=None,
+):
+ """
+ Generates a telemetry wrapper for GenAI audio creation function call
+ """
+
+ async def wrapper(wrapped, instance, args, kwargs):
+ """
+ Wraps the GenAI audio creation function call.
+ """
+
+ server_address, server_port = set_server_address_and_port(
+ instance, "api.groq.com", 443
+ )
+ request_model = kwargs.get("model", "canopylabs/orpheus-v1-english")
+
+ span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO} {request_model}"
+
+ with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+ start_time = time.time()
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ handle_exception(span, e)
+ raise
+ end_time = time.time()
+
+ try:
+ response = process_audio_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,
+ end_time=end_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)
+
+ return response
+
+ return wrapper
+
+
+def async_audio_transcription(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider=None,
+):
+ """
+ Generates a telemetry wrapper for GenAI audio transcription function call
+ """
+
+ async def wrapper(wrapped, instance, args, kwargs):
+ """
+ Wraps the GenAI audio transcription function call.
+ """
+
+ server_address, server_port = set_server_address_and_port(
+ instance, "api.groq.com", 443
+ )
+ request_model = kwargs.get("model", "whisper-large-v3")
+
+ span_name = (
+ f"{SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT} {request_model}"
+ )
+
+ with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+ start_time = time.time()
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ handle_exception(span, e)
+ raise
+ end_time = time.time()
+
+ try:
+ response = process_transcription_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,
+ end_time=end_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)
+
+ return response
+
+ return wrapper
+
+
+def async_audio_translation(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider=None,
+):
+ """
+ Generates a telemetry wrapper for GenAI audio translation function call
+ """
+
+ async def wrapper(wrapped, instance, args, kwargs):
+ """
+ Wraps the GenAI audio translation function call.
+ """
+
+ server_address, server_port = set_server_address_and_port(
+ instance, "api.groq.com", 443
+ )
+ request_model = kwargs.get("model", "whisper-large-v3")
+
+ span_name = (
+ f"{SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT} {request_model}"
+ )
+
+ with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+ start_time = time.time()
+ try:
+ response = await wrapped(*args, **kwargs)
+ except Exception as e:
+ handle_exception(span, e)
+ raise
+ end_time = time.time()
+
+ try:
+ response = process_transcription_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,
+ end_time=end_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)
+
+ return response
+
+ return wrapper
diff --git a/sdk/python/src/openlit/instrumentation/groq/groq.py b/sdk/python/src/openlit/instrumentation/groq/groq.py
index 7d53c075..380e2e24 100644
--- a/sdk/python/src/openlit/instrumentation/groq/groq.py
+++ b/sdk/python/src/openlit/instrumentation/groq/groq.py
@@ -14,6 +14,8 @@ from openlit.instrumentation.groq.utils import (
process_chunk,
process_streaming_chat_response,
process_chat_response,
+ process_audio_response,
+ process_transcription_response,
)
from openlit.semcov import SemanticConvention
@@ -191,3 +193,199 @@ def chat(
return response
return wrapper
+
+
+def audio_create(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider=None,
+):
+ """
+ Generates a telemetry wrapper for GenAI audio creation function call
+ """
+
+ def wrapper(wrapped, instance, args, kwargs):
+ """
+ Wraps the GenAI audio creation function call.
+ """
+
+ server_address, server_port = set_server_address_and_port(
+ instance, "api.groq.com", 443
+ )
+ request_model = kwargs.get("model", "canopylabs/orpheus-v1-english")
+
+ span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO} {request_model}"
+
+ with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+ start_time = time.time()
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ handle_exception(span, e)
+ raise
+ end_time = time.time()
+
+ try:
+ response = process_audio_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,
+ end_time=end_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)
+
+ return response
+
+ return wrapper
+
+
+def audio_transcription(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider=None,
+):
+ """
+ Generates a telemetry wrapper for GenAI audio transcription function call
+ """
+
+ def wrapper(wrapped, instance, args, kwargs):
+ """
+ Wraps the GenAI audio transcription function call.
+ """
+
+ server_address, server_port = set_server_address_and_port(
+ instance, "api.groq.com", 443
+ )
+ request_model = kwargs.get("model", "whisper-large-v3")
+
+ span_name = (
+ f"{SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT} {request_model}"
+ )
+
+ with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+ start_time = time.time()
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ handle_exception(span, e)
+ raise
+ end_time = time.time()
+
+ try:
+ response = process_transcription_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,
+ end_time=end_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)
+
+ return response
+
+ return wrapper
+
+
+def audio_translation(
+ version,
+ environment,
+ application_name,
+ tracer,
+ pricing_info,
+ capture_message_content,
+ metrics,
+ disable_metrics,
+ event_provider=None,
+):
+ """
+ Generates a telemetry wrapper for GenAI audio translation function call
+ """
+
+ def wrapper(wrapped, instance, args, kwargs):
+ """
+ Wraps the GenAI audio translation function call.
+ """
+
+ server_address, server_port = set_server_address_and_port(
+ instance, "api.groq.com", 443
+ )
+ request_model = kwargs.get("model", "whisper-large-v3")
+
+ span_name = (
+ f"{SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT} {request_model}"
+ )
+
+ with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+ start_time = time.time()
+ try:
+ response = wrapped(*args, **kwargs)
+ except Exception as e:
+ handle_exception(span, e)
+ raise
+ end_time = time.time()
+
+ try:
+ response = process_transcription_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,
+ end_time=end_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)
+
+ return response
+
+ return wrapper
diff --git a/sdk/python/src/openlit/instrumentation/groq/utils.py b/sdk/python/src/openlit/instrumentation/groq/utils.py
index 93c90060..e8047500 100644
--- a/sdk/python/src/openlit/instrumentation/groq/utils.py
+++ b/sdk/python/src/openlit/instrumentation/groq/utils.py
@@ -18,8 +18,10 @@ from openlit.__helpers import (
response_as_dict,
calculate_tbt,
general_tokens,
+ get_audio_model_cost,
get_chat_model_cost,
common_span_attributes,
+ record_audio_metrics,
record_completion_metrics,
otel_event,
set_span_reasoning_content,
@@ -674,3 +676,363 @@ def process_chat_response(
)
return response
+
+
+def handle_not_given(value, default=None):
+ """
+ Handle Groq's NotGiven/Omit sentinel values and None by converting them to defaults.
+ """
+ if hasattr(value, "__class__") and value.__class__.__name__ in ("NotGiven", "Omit"):
+ return default
+ if value is None:
+ return default
+ return value
+
+
+def common_audio_logic(
+ scope,
+ request_model,
+ pricing_info,
+ environment,
+ application_name,
+ metrics,
+ capture_message_content,
+ disable_metrics,
+ version,
+ event_provider=None,
+):
+ """
+ Common logic for processing audio operations.
+ """
+
+ # Calculate cost
+ input_text = scope._kwargs.get("input", "")
+ cost = get_audio_model_cost(request_model, pricing_info, input_text)
+
+ # Common Span Attributes
+ common_span_attributes(
+ scope,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ SemanticConvention.GEN_AI_SYSTEM_GROQ,
+ scope._server_address,
+ scope._server_port,
+ request_model,
+ request_model,
+ environment,
+ application_name,
+ False,
+ scope._tbt,
+ scope._ttft,
+ version,
+ )
+
+ # Span Attributes for Request parameters
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_REQUEST_AUDIO_VOICE,
+ handle_not_given(scope._kwargs.get("voice"), "troy"),
+ )
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_REQUEST_AUDIO_RESPONSE_FORMAT,
+ handle_not_given(scope._kwargs.get("response_format"), "wav"),
+ )
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_REQUEST_AUDIO_SPEED,
+ handle_not_given(scope._kwargs.get("speed"), 1.0),
+ )
+
+ # Span Attributes for Cost
+ scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
+
+ # Span Attributes for Content (OTel: array structure)
+ if capture_message_content:
+ input_text = scope._kwargs.get("input", "")
+ input_msgs = (
+ [{"role": "user", "parts": [{"type": "text", "content": input_text}]}]
+ if input_text
+ else []
+ )
+ output_msgs = [
+ {
+ "role": "assistant",
+ "parts": [{"type": "text", "content": "[audio generated]"}],
+ "finish_reason": "stop",
+ }
+ ]
+ _set_span_messages_as_array(scope._span, input_msgs, output_msgs)
+
+ if event_provider:
+ try:
+ emit_inference_event(
+ event_provider=event_provider,
+ operation_name=SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ request_model=request_model,
+ response_model=request_model,
+ input_messages=input_msgs,
+ output_messages=output_msgs,
+ tool_definitions=None,
+ server_address=scope._server_address,
+ server_port=scope._server_port,
+ )
+ except Exception as e:
+ logger.warning("Failed to emit inference event: %s", e, exc_info=True)
+
+ scope._span.set_status(Status(StatusCode.OK))
+
+ # Record metrics
+ if not disable_metrics:
+ record_audio_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_AUDIO,
+ SemanticConvention.GEN_AI_SYSTEM_GROQ,
+ scope._server_address,
+ scope._server_port,
+ request_model,
+ request_model,
+ environment,
+ application_name,
+ scope._start_time,
+ scope._end_time,
+ cost,
+ )
+
+
+def process_audio_response(
+ response,
+ request_model,
+ pricing_info,
+ server_port,
+ server_address,
+ environment,
+ application_name,
+ metrics,
+ start_time,
+ end_time,
+ span,
+ capture_message_content=False,
+ disable_metrics=False,
+ version="1.0.0",
+ event_provider=None,
+ **kwargs,
+):
+ """
+ Process audio generation response and generate telemetry.
+ """
+
+ scope = type("GenericScope", (), {})()
+
+ scope._start_time = start_time
+ scope._end_time = end_time
+ scope._span = span
+ scope._timestamps = []
+ scope._ttft, scope._tbt = scope._end_time - scope._start_time, 0
+ scope._server_address, scope._server_port = server_address, server_port
+ scope._kwargs = kwargs
+
+ common_audio_logic(
+ scope,
+ request_model,
+ pricing_info,
+ environment,
+ application_name,
+ metrics,
+ capture_message_content,
+ disable_metrics,
+ version,
+ event_provider=event_provider,
+ )
+
+ return response
+
+
+def common_transcription_logic(
+ scope,
+ request_model,
+ pricing_info,
+ environment,
+ application_name,
+ metrics,
+ capture_message_content,
+ disable_metrics,
+ version,
+ event_provider=None,
+):
+ """Common logic for processing audio transcription/translation operations."""
+
+ cost = get_audio_model_cost(
+ request_model, pricing_info, None, duration=getattr(scope, "_duration", 0)
+ )
+
+ common_span_attributes(
+ scope,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ SemanticConvention.GEN_AI_SYSTEM_GROQ,
+ scope._server_address,
+ scope._server_port,
+ request_model,
+ request_model,
+ environment,
+ application_name,
+ False,
+ scope._tbt,
+ scope._ttft,
+ version,
+ )
+
+ language = scope._kwargs.get("language")
+ if language:
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_REQUEST_AUDIO_LANGUAGE, language
+ )
+ response_format = scope._kwargs.get("response_format")
+ if response_format:
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_REQUEST_AUDIO_RESPONSE_FORMAT,
+ str(response_format),
+ )
+ temperature = handle_not_given(scope._kwargs.get("temperature"))
+ if temperature is not None:
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, temperature
+ )
+
+ if hasattr(scope, "_duration") and scope._duration:
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_RESPONSE_AUDIO_DURATION, scope._duration
+ )
+
+ if hasattr(scope, "_input_tokens") and scope._input_tokens:
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS, scope._input_tokens
+ )
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, scope._output_tokens
+ )
+ scope._span.set_attribute(
+ SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE,
+ scope._input_tokens + scope._output_tokens,
+ )
+
+ scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
+
+ if capture_message_content:
+ file_obj = scope._kwargs.get("file")
+ if isinstance(file_obj, (tuple, list)) and file_obj:
+ # Groq accepts (filename, fileobj[, content_type]) tuples for uploads.
+ file_name = file_obj[0]
+ elif file_obj is not None:
+ file_name = getattr(file_obj, "name", "unknown")
+ else:
+ file_name = "unknown"
+ input_msgs = [
+ {
+ "role": "user",
+ "parts": [{"type": "text", "content": f"[audio file: {file_name}]"}],
+ }
+ ]
+ output_msgs = None
+ if hasattr(scope, "_transcription_text") and scope._transcription_text:
+ output_msgs = [
+ {
+ "role": "assistant",
+ "parts": [{"type": "text", "content": scope._transcription_text}],
+ "finish_reason": "stop",
+ }
+ ]
+ _set_span_messages_as_array(scope._span, input_msgs, output_msgs)
+
+ if event_provider:
+ try:
+ emit_inference_event(
+ event_provider=event_provider,
+ operation_name=SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ request_model=request_model,
+ response_model=request_model,
+ input_messages=input_msgs,
+ output_messages=output_msgs,
+ server_address=scope._server_address,
+ server_port=scope._server_port,
+ )
+ except Exception as e:
+ logger.warning("Failed to emit inference event: %s", e, exc_info=True)
+
+ scope._span.set_status(Status(StatusCode.OK))
+
+ if not disable_metrics:
+ record_audio_metrics(
+ metrics,
+ SemanticConvention.GEN_AI_OPERATION_TYPE_SPEECH_TO_TEXT,
+ SemanticConvention.GEN_AI_SYSTEM_GROQ,
+ scope._server_address,
+ scope._server_port,
+ request_model,
+ request_model,
+ environment,
+ application_name,
+ scope._start_time,
+ scope._end_time,
+ cost,
+ )
+
+
+def process_transcription_response(
+ response,
+ request_model,
+ pricing_info,
+ server_port,
+ server_address,
+ environment,
+ application_name,
+ metrics,
+ start_time,
+ end_time,
+ span,
+ capture_message_content=False,
+ disable_metrics=False,
+ version="1.0.0",
+ event_provider=None,
+ **kwargs,
+):
+ """Process audio transcription/translation response and generate telemetry."""
+
+ scope = type("GenericScope", (), {})()
+ response_dict = response_as_dict(response)
+
+ scope._start_time = start_time
+ scope._end_time = end_time
+ scope._span = span
+ scope._timestamps = []
+ scope._ttft, scope._tbt = end_time - start_time, 0
+ scope._server_address, scope._server_port = server_address, server_port
+ scope._kwargs = kwargs
+
+ usage = response_dict.get("usage", {})
+ if isinstance(usage, dict):
+ usage_type = usage.get("type", "duration")
+ if usage_type == "tokens":
+ scope._input_tokens = usage.get("input_tokens", 0)
+ scope._output_tokens = usage.get("output_tokens", 0)
+ else:
+ scope._input_tokens = 0
+ scope._output_tokens = 0
+ else:
+ scope._input_tokens = 0
+ scope._output_tokens = 0
+
+ scope._transcription_text = response_dict.get("text", "")
+ scope._language = response_dict.get("language", "")
+ scope._duration = response_dict.get("duration", 0)
+
+ common_transcription_logic(
+ scope,
+ request_model,
+ pricing_info,
+ environment,
+ application_name,
+ metrics,
+ capture_message_content,
+ disable_metrics,
+ version,
+ event_provider=event_provider,
+ )
+
+ return response
diff --git a/sdk/python/tests/test_groq.py b/sdk/python/tests/test_groq.py
index e5491dc9..89e49832 100644
--- a/sdk/python/tests/test_groq.py
+++ b/sdk/python/tests/test_groq.py
@@ -86,3 +86,167 @@ async def test_async_groq_chat():
print("Rate Limited:", e)
else:
raise
+
+
+def _sample_audio_file():
+ """
+ Generates a small in-memory WAV file for speech-to-text tests.
+ """
+
+ import io
+ import struct
+ import wave
+
+ buffer = io.BytesIO()
+ with wave.open(buffer, "wb") as wav_file:
+ wav_file.setnchannels(1)
+ wav_file.setsampwidth(2)
+ wav_file.setframerate(16000)
+ wav_file.writeframes(struct.pack("<h", 0) * 16000)
+ buffer.seek(0)
+ return ("sample.wav", buffer)
+
+
+def test_sync_groq_transcription():
+ """
+ Tests synchronous Audio Transcriptions.
+
+ Raises:
+ AssertionError: If the Audio Transcriptions response object is not as expected.
+ """
+
+ try:
+ transcription_resp = sync_client.audio.transcriptions.create(
+ file=_sample_audio_file(),
+ model="whisper-large-v3",
+ )
+ assert isinstance(transcription_resp.text, str)
+
+ # pylint: disable=broad-exception-caught
+ except Exception as e:
+ if "rate limit" in str(e).lower():
+ print("Rate Limited:", e)
+ else:
+ raise
+
+
+def test_sync_groq_translation():
+ """
+ Tests synchronous Audio Translations.
+
+ Raises:
+ AssertionError: If the Audio Translations response object is not as expected.
+ """
+
+ try:
+ translation_resp = sync_client.audio.translations.create(
+ file=_sample_audio_file(),
+ model="whisper-large-v3",
+ )
+ assert isinstance(translation_resp.text, str)
+
+ # pylint: disable=broad-exception-caught
+ except Exception as e:
+ if "rate limit" in str(e).lower():
+ print("Rate Limited:", e)
+ else:
+ raise
+
+
+def test_sync_groq_speech():
+ """
+ Tests synchronous Audio Speech (text-to-speech).
+
+ Raises:
+ AssertionError: If the Audio Speech response object is not as expected.
+ """
+
+ try:
+ speech_resp = sync_client.audio.speech.create(
+ model="canopylabs/orpheus-v1-english",
+ voice="troy",
+ input="Monitor LLM Applications",
+ response_format="wav",
+ )
+ assert speech_resp.read()
+
+ # pylint: disable=broad-exception-caught
+ except Exception as e:
+ if "rate limit" in str(e).lower() or "terms acceptance" in str(e).lower():
+ print("Skipped:", e)
+ else:
+ raise
+
+
+async def test_async_groq_transcription():
+ """
+ Tests asynchronous Audio Transcriptions.
+
+ Raises:
+ AssertionError: If the Audio Transcriptions response object is not as expected.
+ """
+
+ try:
+ transcription_resp = await async_client.audio.transcriptions.create(
+ file=_sample_audio_file(),
+ model="whisper-large-v3",
+ )
+ assert isinstance(transcription_resp.text, str)
+
+ # pylint: disable=broad-exception-caught
+ except Exception as e:
+ if "rate limit" in str(e).lower():
+ print("Rate Limited:", e)
+ else:
+ raise
+
+
+async def test_async_groq_translation():
+ """
+ Tests asynchronous Audio Translations.
+
+ Raises:
+ AssertionError: If the Audio Translations response object is not as expected.
+ """
+
+ try:
+ translation_resp = await async_client.audio.translations.create(
+ file=_sample_audio_file(),
+ model="whisper-large-v3",
+ )
+ assert isinstance(translation_resp.text, str)
+
+ # pylint: disable=broad-exception-caught
+ except Exception as e:
+ if "rate limit" in str(e).lower():
+ print("Rate Limited:", e)
+ else:
+ raise
+
+
+async def test_async_groq_speech():
+ """
+ Tests asynchronous Audio Speech (text-to-speech).
+
+ Raises:
+ AssertionError: If the Audio Speech response object is not as expected.
+ """
+
+ try:
+ speech_resp = await async_client.audio.speech.create(
+ model="canopylabs/orpheus-v1-english",
+ voice="troy",
+ input="Monitor LLM Applications",
+ response_format="wav",
+ )
+ assert await speech_resp.read()
+
+ # pylint: disable=broad-exception-caught
+ except Exception as e:
+ if "rate limit" in str(e).lower() or "terms acceptance" in str(e).lower():
+ print("Skipped:", e)
+ else:
+ raise