Skip to content

Commit

Permalink
feat: Updated semantic conventions based on otel community (#884)
Browse files Browse the repository at this point in the history 
Co-authored-by: Nir Gazit <nirga@users.noreply.github.com>
  • Loading branch information
@gyliu513 @nirga
gyliu513 and nirga authored Apr 29, 2024
1 parent e2a09ae commit 23eba2a
Show file tree
Hide file tree
Showing 69 changed files with 454 additions and 435 deletions.
10 changes: 5 additions & 5 deletions ...entelemetry-instrumentation-anthropic/opentelemetry/instrumentation/anthropic/__init__.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -110,8 +110,8 @@ def _set_input_attributes(span, kwargs):
set_span_attribute(
span, SpanAttributes.LLM_REQUEST_MAX_TOKENS, kwargs.get("max_tokens_to_sample")
)
set_span_attribute(span, SpanAttributes.LLM_TEMPERATURE, kwargs.get("temperature"))
set_span_attribute(span, SpanAttributes.LLM_TOP_P, kwargs.get("top_p"))
set_span_attribute(span, SpanAttributes.LLM_REQUEST_TEMPERATURE, kwargs.get("temperature"))
set_span_attribute(span, SpanAttributes.LLM_REQUEST_TOP_P, kwargs.get("top_p"))
set_span_attribute(
span, SpanAttributes.LLM_FREQUENCY_PENALTY, kwargs.get("frequency_penalty")
)
Expand Down Expand Up @@ -394,7 +394,7 @@ def _calculate_metrics_attributes(response):
if not isinstance(response, dict):
response = response.__dict__
return {
"llm.response.model": response.get("model"),
"gen_ai.response.model": response.get("model"),
}


Expand All @@ -420,7 +420,7 @@ def _wrap(
name,
kind=SpanKind.CLIENT,
attributes={
SpanAttributes.LLM_VENDOR: "Anthropic",
SpanAttributes.LLM_SYSTEM: "Anthropic",
SpanAttributes.LLM_REQUEST_TYPE: LLMRequestTypeValues.COMPLETION.value,
},
)
Expand Down Expand Up @@ -516,7 +516,7 @@ async def _awrap(
name,
kind=SpanKind.CLIENT,
attributes={
SpanAttributes.LLM_VENDOR: "Anthropic",
SpanAttributes.LLM_SYSTEM: "Anthropic",
SpanAttributes.LLM_REQUEST_TYPE: LLMRequestTypeValues.COMPLETION.value,
},
)
Expand Down
4 changes: 2 additions & 2 deletions ...ntelemetry-instrumentation-anthropic/opentelemetry/instrumentation/anthropic/streaming.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -122,7 +122,7 @@ def build_from_streaming_response(
_process_response_item(item, complete_response)

metric_attributes = {
"llm.response.model": complete_response.get("model"),
"gen_ai.response.model": complete_response.get("model"),
}

if duration_histogram:
Expand Down Expand Up @@ -205,7 +205,7 @@ async def abuild_from_streaming_response(
_process_response_item(item, complete_response)

metric_attributes = {
"llm.response.model": complete_response.get("model"),
"gen_ai.response.model": complete_response.get("model"),
}

if duration_histogram:
Expand Down
8 changes: 4 additions & 4 deletions packages/opentelemetry-instrumentation-anthropic/poetry.lock
View file

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion packages/opentelemetry-instrumentation-anthropic/pyproject.toml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@ python = ">=3.9,<4"
opentelemetry-api = "^1.24.0"
opentelemetry-instrumentation = "^0.45b0"
opentelemetry-semantic-conventions = "^0.45b0"
opentelemetry-semantic-conventions-ai = "0.1.1"
opentelemetry-semantic-conventions-ai = "0.2.0"

[tool.poetry.group.dev.dependencies]
autopep8 = "2.1.0"
Expand Down
94 changes: 47 additions & 47 deletions packages/opentelemetry-instrumentation-anthropic/tests/test_completion.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -26,10 +26,10 @@ def test_anthropic_completion(exporter, reader):

anthropic_span = spans[0]
assert (
anthropic_span.attributes["llm.prompts.0.user"]
anthropic_span.attributes["gen_ai.prompt.0.user"]
== f"{HUMAN_PROMPT}\nHello world\n{AI_PROMPT}"
)
assert anthropic_span.attributes.get("llm.completions.0.content")
assert anthropic_span.attributes.get("gen_ai.completion.0.content")

metrics_data = reader.get_metrics_data()
resource_metrics = metrics_data.resource_metrics
Expand All @@ -51,7 +51,7 @@ def test_anthropic_completion(exporter, reader):
"prompt",
]
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-instant-1.2"
)
assert data_point.value > 0
Expand All @@ -61,7 +61,7 @@ def test_anthropic_completion(exporter, reader):
for data_point in metric.data.data_points:
assert data_point.value >= 1
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-instant-1.2"
)

Expand All @@ -74,7 +74,7 @@ def test_anthropic_completion(exporter, reader):
data_point.sum > 0 for data_point in metric.data.data_points
)
assert all(
data_point.attributes.get("llm.response.model")
data_point.attributes.get("gen_ai.response.model")
== "claude-instant-1.2"
or data_point.attributes.get("error.type") == "TypeError"
for data_point in metric.data.data_points
Expand Down Expand Up @@ -117,18 +117,18 @@ def test_anthropic_message_create(exporter, reader):

anthropic_span = spans[0]
assert (
anthropic_span.attributes["llm.prompts.0.content"]
anthropic_span.attributes["gen_ai.prompt.0.content"]
== "Tell me a joke about OpenTelemetry"
)
assert (anthropic_span.attributes["llm.prompts.0.role"]) == "user"
assert (anthropic_span.attributes["gen_ai.prompt.0.role"]) == "user"
assert (
anthropic_span.attributes.get("llm.completions.0.content")
anthropic_span.attributes.get("gen_ai.completion.0.content")
== response.content[0].text
)
assert anthropic_span.attributes["llm.usage.prompt_tokens"] == 8
assert anthropic_span.attributes["gen_ai.usage.prompt_tokens"] == 8
assert (
anthropic_span.attributes["llm.usage.completion_tokens"]
+ anthropic_span.attributes["llm.usage.prompt_tokens"]
anthropic_span.attributes["gen_ai.usage.completion_tokens"]
+ anthropic_span.attributes["gen_ai.usage.prompt_tokens"]
== anthropic_span.attributes["llm.usage.total_tokens"]
)

Expand All @@ -152,7 +152,7 @@ def test_anthropic_message_create(exporter, reader):
"prompt",
]
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-opus-20240229"
)
assert data_point.value > 0
Expand All @@ -162,7 +162,7 @@ def test_anthropic_message_create(exporter, reader):
for data_point in metric.data.data_points:
assert data_point.value >= 1
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-opus-20240229"
)

Expand All @@ -175,7 +175,7 @@ def test_anthropic_message_create(exporter, reader):
data_point.sum > 0 for data_point in metric.data.data_points
)
assert all(
data_point.attributes.get("llm.response.model")
data_point.attributes.get("gen_ai.response.model")
== "claude-3-opus-20240229"
or data_point.attributes.get("error.type") == "TypeError"
for data_point in metric.data.data_points
Expand Down Expand Up @@ -225,7 +225,7 @@ def test_anthropic_multi_modal(exporter):
"anthropic.completion",
]
anthropic_span = spans[0]
assert anthropic_span.attributes["llm.prompts.0.content"] == json.dumps(
assert anthropic_span.attributes["gen_ai.prompt.0.content"] == json.dumps(
[
{"type": "text", "text": "What do you see?"},
{
Expand All @@ -238,15 +238,15 @@ def test_anthropic_multi_modal(exporter):
},
]
)
assert (anthropic_span.attributes["llm.prompts.0.role"]) == "user"
assert (anthropic_span.attributes["gen_ai.prompt.0.role"]) == "user"
assert (
anthropic_span.attributes.get("llm.completions.0.content")
anthropic_span.attributes.get("gen_ai.completion.0.content")
== response.content[0].text
)
assert anthropic_span.attributes["llm.usage.prompt_tokens"] == 1381
assert anthropic_span.attributes["gen_ai.usage.prompt_tokens"] == 1381
assert (
anthropic_span.attributes["llm.usage.completion_tokens"]
+ anthropic_span.attributes["llm.usage.prompt_tokens"]
anthropic_span.attributes["gen_ai.usage.completion_tokens"]
+ anthropic_span.attributes["gen_ai.usage.prompt_tokens"]
== anthropic_span.attributes["llm.usage.total_tokens"]
)

Expand Down Expand Up @@ -277,17 +277,17 @@ def test_anthropic_message_streaming(exporter, reader):
]
anthropic_span = spans[0]
assert (
anthropic_span.attributes["llm.prompts.0.content"]
anthropic_span.attributes["gen_ai.prompt.0.content"]
== "Tell me a joke about OpenTelemetry"
)
assert (anthropic_span.attributes["llm.prompts.0.role"]) == "user"
assert (anthropic_span.attributes["gen_ai.prompt.0.role"]) == "user"
assert (
anthropic_span.attributes.get("llm.completions.0.content") == response_content
anthropic_span.attributes.get("gen_ai.completion.0.content") == response_content
)
assert anthropic_span.attributes["llm.usage.prompt_tokens"] == 8
assert anthropic_span.attributes["gen_ai.usage.prompt_tokens"] == 8
assert (
anthropic_span.attributes["llm.usage.completion_tokens"]
+ anthropic_span.attributes["llm.usage.prompt_tokens"]
anthropic_span.attributes["gen_ai.usage.completion_tokens"]
+ anthropic_span.attributes["gen_ai.usage.prompt_tokens"]
== anthropic_span.attributes["llm.usage.total_tokens"]
)

Expand All @@ -311,7 +311,7 @@ def test_anthropic_message_streaming(exporter, reader):
"prompt",
]
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-haiku-20240307"
)
assert data_point.value > 0
Expand All @@ -321,7 +321,7 @@ def test_anthropic_message_streaming(exporter, reader):
for data_point in metric.data.data_points:
assert data_point.value >= 1
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-haiku-20240307"
)

Expand All @@ -334,7 +334,7 @@ def test_anthropic_message_streaming(exporter, reader):
data_point.sum > 0 for data_point in metric.data.data_points
)
assert all(
data_point.attributes.get("llm.response.model")
data_point.attributes.get("gen_ai.response.model")
== "claude-3-haiku-20240307"
or data_point.attributes.get("error.type") == "TypeError"
for data_point in metric.data.data_points
Expand Down Expand Up @@ -372,18 +372,18 @@ async def test_async_anthropic_message_create(exporter, reader):
]
anthropic_span = spans[0]
assert (
anthropic_span.attributes["llm.prompts.0.content"]
anthropic_span.attributes["gen_ai.prompt.0.content"]
== "Tell me a joke about OpenTelemetry"
)
assert (anthropic_span.attributes["llm.prompts.0.role"]) == "user"
assert (anthropic_span.attributes["gen_ai.prompt.0.role"]) == "user"
assert (
anthropic_span.attributes.get("llm.completions.0.content")
anthropic_span.attributes.get("gen_ai.completion.0.content")
== response.content[0].text
)
assert anthropic_span.attributes["llm.usage.prompt_tokens"] == 8
assert anthropic_span.attributes["gen_ai.usage.prompt_tokens"] == 8
assert (
anthropic_span.attributes["llm.usage.completion_tokens"]
+ anthropic_span.attributes["llm.usage.prompt_tokens"]
anthropic_span.attributes["gen_ai.usage.completion_tokens"]
+ anthropic_span.attributes["gen_ai.usage.prompt_tokens"]
== anthropic_span.attributes["llm.usage.total_tokens"]
)

Expand All @@ -407,7 +407,7 @@ async def test_async_anthropic_message_create(exporter, reader):
"prompt",
]
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-opus-20240229"
)
assert data_point.value > 0
Expand All @@ -417,7 +417,7 @@ async def test_async_anthropic_message_create(exporter, reader):
for data_point in metric.data.data_points:
assert data_point.value >= 1
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-opus-20240229"
)

Expand All @@ -430,7 +430,7 @@ async def test_async_anthropic_message_create(exporter, reader):
data_point.sum > 0 for data_point in metric.data.data_points
)
assert all(
data_point.attributes.get("llm.response.model")
data_point.attributes.get("gen_ai.response.model")
== "claude-3-opus-20240229"
or data_point.attributes.get("error.type") == "TypeError"
for data_point in metric.data.data_points
Expand Down Expand Up @@ -474,17 +474,17 @@ async def test_async_anthropic_message_streaming(exporter, reader):
]
anthropic_span = spans[0]
assert (
anthropic_span.attributes["llm.prompts.0.content"]
anthropic_span.attributes["gen_ai.prompt.0.content"]
== "Tell me a joke about OpenTelemetry"
)
assert (anthropic_span.attributes["llm.prompts.0.role"]) == "user"
assert (anthropic_span.attributes["gen_ai.prompt.0.role"]) == "user"
assert (
anthropic_span.attributes.get("llm.completions.0.content") == response_content
anthropic_span.attributes.get("gen_ai.completion.0.content") == response_content
)
assert anthropic_span.attributes["llm.usage.prompt_tokens"] == 8
assert anthropic_span.attributes["gen_ai.usage.prompt_tokens"] == 8
assert (
anthropic_span.attributes["llm.usage.completion_tokens"]
+ anthropic_span.attributes["llm.usage.prompt_tokens"]
anthropic_span.attributes["gen_ai.usage.completion_tokens"]
+ anthropic_span.attributes["gen_ai.usage.prompt_tokens"]
== anthropic_span.attributes["llm.usage.total_tokens"]
)

Expand All @@ -508,7 +508,7 @@ async def test_async_anthropic_message_streaming(exporter, reader):
"prompt",
]
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-haiku-20240307"
)
assert data_point.value > 0
Expand All @@ -518,7 +518,7 @@ async def test_async_anthropic_message_streaming(exporter, reader):
for data_point in metric.data.data_points:
assert data_point.value >= 1
assert (
data_point.attributes["llm.response.model"]
data_point.attributes["gen_ai.response.model"]
== "claude-3-haiku-20240307"
)

Expand All @@ -531,7 +531,7 @@ async def test_async_anthropic_message_streaming(exporter, reader):
data_point.sum > 0 for data_point in metric.data.data_points
)
assert all(
data_point.attributes.get("llm.response.model")
data_point.attributes.get("gen_ai.response.model")
== "claude-3-haiku-20240307"
or data_point.attributes.get("error.type") == "TypeError"
for data_point in metric.data.data_points
Expand Down
Loading

0 comments on commit 23eba2a

Please sign in to comment.