opentelemetry-instrumentation-groq 0.38.6__py3-none-any.whl → 0.38.8__py3-none-any.whl

opentelemetry/instrumentation/groq/__init__.py

@@ -21,7 +21,9 @@ from opentelemetry.instrumentation.groq.version import __version__
 from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
 from opentelemetry.instrumentation.utils import _SUPPRESS_INSTRUMENTATION_KEY, unwrap
 from opentelemetry.metrics import Counter, Histogram, Meter, get_meter
-from opentelemetry.semconv._incubating.attributes.gen_ai_attributes import GEN_AI_RESPONSE_ID
+from opentelemetry.semconv._incubating.attributes.gen_ai_attributes import (
+    GEN_AI_RESPONSE_ID,
+)
 from opentelemetry.semconv_ai import (
     SUPPRESS_LANGUAGE_MODEL_INSTRUMENTATION_KEY,
     LLMRequestTypeValues,
@@ -97,7 +99,9 @@ def _set_input_attributes(span, kwargs):
     set_span_attribute(
         span, SpanAttributes.LLM_PRESENCE_PENALTY, kwargs.get("presence_penalty")
     )
-    set_span_attribute(span, SpanAttributes.LLM_IS_STREAMING, kwargs.get("stream") or False)
+    set_span_attribute(
+        span, SpanAttributes.LLM_IS_STREAMING, kwargs.get("stream") or False
+    )
 
     if should_send_prompts():
         if kwargs.get("prompt") is not None:
@@ -124,9 +128,7 @@ def _set_completions(span, choices):
     for choice in choices:
         index = choice.get("index")
         prefix = f"{SpanAttributes.LLM_COMPLETIONS}.{index}"
-        set_span_attribute(
-            span, f"{prefix}.finish_reason", choice.get("finish_reason")
-        )
+        set_span_attribute(span, f"{prefix}.finish_reason", choice.get("finish_reason"))
 
         if choice.get("content_filter_results"):
             set_span_attribute(
@@ -181,26 +183,38 @@ def _set_completions(span, choices):
 
 
 @dont_throw
-def _set_response_attributes(span, response):
+def _set_response_attributes(span, response, token_histogram):
     response = model_as_dict(response)
-
     set_span_attribute(span, SpanAttributes.LLM_RESPONSE_MODEL, response.get("model"))
     set_span_attribute(span, GEN_AI_RESPONSE_ID, response.get("id"))
 
-    usage = response.get("usage")
+    usage = response.get("usage") or {}
+    prompt_tokens = usage.get("prompt_tokens")
+    completion_tokens = usage.get("completion_tokens")
     if usage:
         set_span_attribute(
             span, SpanAttributes.LLM_USAGE_TOTAL_TOKENS, usage.get("total_tokens")
         )
         set_span_attribute(
             span,
-            SpanAttributes.LLM_USAGE_COMPLETION_TOKENS,
-            usage.get("completion_tokens"),
+            SpanAttributes.LLM_USAGE_COMPLETION_TOKENS, completion_tokens
         )
         set_span_attribute(
-            span, SpanAttributes.LLM_USAGE_PROMPT_TOKENS, usage.get("prompt_tokens")
+            span, SpanAttributes.LLM_USAGE_PROMPT_TOKENS, prompt_tokens
         )
 
+    if isinstance(prompt_tokens, int) and prompt_tokens >= 0 and token_histogram is not None:
+        token_histogram.record(prompt_tokens, attributes={
+            SpanAttributes.LLM_TOKEN_TYPE: "input",
+            SpanAttributes.LLM_RESPONSE_MODEL: response.get("model")
+        })
+
+    if isinstance(completion_tokens, int) and completion_tokens >= 0 and token_histogram is not None:
+        token_histogram.record(completion_tokens, attributes={
+            SpanAttributes.LLM_TOKEN_TYPE: "output",
+            SpanAttributes.LLM_RESPONSE_MODEL: response.get("model")
+        })
+
     choices = response.get("choices")
     if should_send_prompts() and choices:
         _set_completions(span, choices)
@@ -268,6 +282,96 @@ def _create_metrics(meter: Meter):
     return token_histogram, choice_counter, duration_histogram
 
 
+def _process_streaming_chunk(chunk):
+    """Extract content, finish_reason and usage from a streaming chunk."""
+    if not chunk.choices:
+        return None, None, None
+
+    delta = chunk.choices[0].delta
+    content = delta.content if hasattr(delta, "content") else None
+    finish_reason = chunk.choices[0].finish_reason
+
+    # Extract usage from x_groq if present in the final chunk
+    usage = None
+    if hasattr(chunk, "x_groq") and chunk.x_groq and chunk.x_groq.usage:
+        usage = chunk.x_groq.usage
+
+    return content, finish_reason, usage
+
+
+def _set_streaming_response_attributes(
+    span, accumulated_content, finish_reason=None, usage=None
+):
+    """Set span attributes for accumulated streaming response."""
+    if not span.is_recording():
+        return
+
+    prefix = f"{SpanAttributes.LLM_COMPLETIONS}.0"
+    set_span_attribute(span, f"{prefix}.role", "assistant")
+    set_span_attribute(span, f"{prefix}.content", accumulated_content)
+    if finish_reason:
+        set_span_attribute(span, f"{prefix}.finish_reason", finish_reason)
+
+    if usage:
+        set_span_attribute(
+            span, SpanAttributes.LLM_USAGE_COMPLETION_TOKENS, usage.completion_tokens
+        )
+        set_span_attribute(
+            span, SpanAttributes.LLM_USAGE_PROMPT_TOKENS, usage.prompt_tokens
+        )
+        set_span_attribute(
+            span, SpanAttributes.LLM_USAGE_TOTAL_TOKENS, usage.total_tokens
+        )
+
+
+def _create_stream_processor(response, span):
+    """Create a generator that processes a stream while collecting telemetry."""
+    accumulated_content = ""
+    finish_reason = None
+    usage = None
+
+    for chunk in response:
+        content, chunk_finish_reason, chunk_usage = _process_streaming_chunk(chunk)
+        if content:
+            accumulated_content += content
+        if chunk_finish_reason:
+            finish_reason = chunk_finish_reason
+        if chunk_usage:
+            usage = chunk_usage
+        yield chunk
+
+    if span.is_recording():
+        _set_streaming_response_attributes(
+            span, accumulated_content, finish_reason, usage
+        )
+        span.set_status(Status(StatusCode.OK))
+    span.end()
+
+
+async def _create_async_stream_processor(response, span):
+    """Create an async generator that processes a stream while collecting telemetry."""
+    accumulated_content = ""
+    finish_reason = None
+    usage = None
+
+    async for chunk in response:
+        content, chunk_finish_reason, chunk_usage = _process_streaming_chunk(chunk)
+        if content:
+            accumulated_content += content
+        if chunk_finish_reason:
+            finish_reason = chunk_finish_reason
+        if chunk_usage:
+            usage = chunk_usage
+        yield chunk
+
+    if span.is_recording():
+        _set_streaming_response_attributes(
+            span, accumulated_content, finish_reason, usage
+        )
+        span.set_status(Status(StatusCode.OK))
+    span.end()
+
+
 @_with_chat_telemetry_wrapper
 def _wrap(
     tracer: Tracer,
@@ -315,8 +419,16 @@ def _wrap(
     end_time = time.time()
 
     if is_streaming_response(response):
-        # TODO: implement streaming
-        pass
+        try:
+            return _create_stream_processor(response, span)
+        except Exception as ex:
+            logger.warning(
+                "Failed to process streaming response for groq span, error: %s",
+                str(ex),
+            )
+            span.set_status(Status(StatusCode.ERROR))
+            span.end()
+            raise
     elif response:
         try:
             metric_attributes = shared_metrics_attributes(response)
@@ -329,7 +441,7 @@ def _wrap(
                 )
 
             if span.is_recording():
-                _set_response_attributes(span, response)
+                _set_response_attributes(span, response, token_histogram)
 
         except Exception as ex:  # pylint: disable=broad-except
             logger.warning(
@@ -391,9 +503,19 @@ async def _awrap(
 
         raise e
 
+    end_time = time.time()
+
     if is_streaming_response(response):
-        # TODO: implement streaming
-        pass
+        try:
+            return await _create_async_stream_processor(response, span)
+        except Exception as ex:
+            logger.warning(
+                "Failed to process streaming response for groq span, error: %s",
+                str(ex),
+            )
+            span.set_status(Status(StatusCode.ERROR))
+            span.end()
+            raise
     elif response:
         metric_attributes = shared_metrics_attributes(response)
 
@@ -405,7 +527,7 @@ async def _awrap(
             )
 
         if span.is_recording():
-            _set_response_attributes(span, response)
+            _set_response_attributes(span, response, token_histogram)
 
         if span.is_recording():
             span.set_status(Status(StatusCode.OK))

opentelemetry/instrumentation/groq/version.py

@@ -1 +1 @@
-__version__ = "0.38.6"
+__version__ = "0.38.8"

{opentelemetry_instrumentation_groq-0.38.6.dist-info → opentelemetry_instrumentation_groq-0.38.8.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: opentelemetry-instrumentation-groq
-Version: 0.38.6
+Version: 0.38.8
 Summary: OpenTelemetry Groq instrumentation
 License: Apache-2.0
 Author: Gal Kleinman

opentelemetry_instrumentation_groq-0.38.8.dist-info/RECORD

@@ -0,0 +1,8 @@
+opentelemetry/instrumentation/groq/__init__.py,sha256=zRi0qEDmT-iBS2zGKUecx53flOP4criJWS9mtBsEYRc,20313
+opentelemetry/instrumentation/groq/config.py,sha256=eN2YxQdWlAF-qWPwZZr0xFM-8tx9zUjmiparuB64jcU,170
+opentelemetry/instrumentation/groq/utils.py,sha256=1ESL4NCp8Mjww8cGEzQO_AEqGiSK4JSiMFYUhwBnuao,2151
+opentelemetry/instrumentation/groq/version.py,sha256=iWPMFCHqu7GKqG8CoHB9sVbeBazpddbqk566MeK8WnE,23
+opentelemetry_instrumentation_groq-0.38.8.dist-info/METADATA,sha256=tFRh8E5qAfhDrbuWFk1HWOs5fDJnYRINHbT90eMq2Cs,2117
+opentelemetry_instrumentation_groq-0.38.8.dist-info/WHEEL,sha256=XbeZDeTWKc1w7CSIyre5aMDU_-PohRwTQceYnisIYYY,88
+opentelemetry_instrumentation_groq-0.38.8.dist-info/entry_points.txt,sha256=uezQe06CpIK8xTZZSK0lF29nOKkz_w6VR4sQnb4IAFQ,87
+opentelemetry_instrumentation_groq-0.38.8.dist-info/RECORD,,

opentelemetry_instrumentation_groq-0.38.6.dist-info/RECORD

@@ -1,8 +0,0 @@
-opentelemetry/instrumentation/groq/__init__.py,sha256=RgXnJKztm990BbUtJvOcliYqU-c975JF1CRv9wSalcc,16048
-opentelemetry/instrumentation/groq/config.py,sha256=eN2YxQdWlAF-qWPwZZr0xFM-8tx9zUjmiparuB64jcU,170
-opentelemetry/instrumentation/groq/utils.py,sha256=1ESL4NCp8Mjww8cGEzQO_AEqGiSK4JSiMFYUhwBnuao,2151
-opentelemetry/instrumentation/groq/version.py,sha256=uJq0GABwRg0CsFCDLTKi9uMtcv-BL9qIEL5QqdBeXXE,23
-opentelemetry_instrumentation_groq-0.38.6.dist-info/METADATA,sha256=Yiphzu4bN1WwFwrnOc8vPDne4aiS3yKXe9mcuny6NKk,2117
-opentelemetry_instrumentation_groq-0.38.6.dist-info/WHEEL,sha256=XbeZDeTWKc1w7CSIyre5aMDU_-PohRwTQceYnisIYYY,88
-opentelemetry_instrumentation_groq-0.38.6.dist-info/entry_points.txt,sha256=uezQe06CpIK8xTZZSK0lF29nOKkz_w6VR4sQnb4IAFQ,87
-opentelemetry_instrumentation_groq-0.38.6.dist-info/RECORD,,