openlit 0.0.1__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- openlit/__helpers.py +143 -0
- openlit/__init__.py +184 -0
- openlit/instrumentation/anthropic/__init__.py +50 -0
- openlit/instrumentation/anthropic/anthropic.py +291 -0
- openlit/instrumentation/anthropic/async_anthropic.py +291 -0
- openlit/instrumentation/chroma/__init__.py +86 -0
- openlit/instrumentation/chroma/chroma.py +197 -0
- openlit/instrumentation/cohere/__init__.py +51 -0
- openlit/instrumentation/cohere/cohere.py +397 -0
- openlit/instrumentation/langchain/__init__.py +74 -0
- openlit/instrumentation/langchain/langchain.py +161 -0
- openlit/instrumentation/mistral/__init__.py +80 -0
- openlit/instrumentation/mistral/async_mistral.py +417 -0
- openlit/instrumentation/mistral/mistral.py +416 -0
- openlit/instrumentation/openai/__init__.py +335 -0
- openlit/instrumentation/openai/async_azure_openai.py +841 -0
- openlit/instrumentation/openai/async_openai.py +875 -0
- openlit/instrumentation/openai/azure_openai.py +840 -0
- openlit/instrumentation/openai/openai.py +891 -0
- openlit/instrumentation/pinecone/__init__.py +66 -0
- openlit/instrumentation/pinecone/pinecone.py +173 -0
- openlit/instrumentation/transformers/__init__.py +37 -0
- openlit/instrumentation/transformers/transformers.py +156 -0
- openlit/otel/metrics.py +109 -0
- openlit/otel/tracing.py +83 -0
- openlit/semcov/__init__.py +123 -0
- openlit-0.0.1.dist-info/LICENSE +201 -0
- openlit-0.0.1.dist-info/METADATA +113 -0
- openlit-0.0.1.dist-info/RECORD +30 -0
- openlit-0.0.1.dist-info/WHEEL +4 -0
@@ -0,0 +1,291 @@
|
|
1
|
+
# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument
|
2
|
+
"""
|
3
|
+
Module for monitoring Anthropic API calls.
|
4
|
+
"""
|
5
|
+
|
6
|
+
import logging
|
7
|
+
from opentelemetry.trace import SpanKind, Status, StatusCode
|
8
|
+
from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
|
9
|
+
from openlit.__helpers import get_chat_model_cost, handle_exception
|
10
|
+
from openlit.semcov import SemanticConvetion
|
11
|
+
|
12
|
+
# Initialize logger for logging potential issues and operations
|
13
|
+
logger = logging.getLogger(__name__)
|
14
|
+
|
15
|
+
def async_messages(gen_ai_endpoint, version, environment, application_name,
|
16
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics):
|
17
|
+
"""
|
18
|
+
Generates a telemetry wrapper for messages to collect metrics.
|
19
|
+
|
20
|
+
Args:
|
21
|
+
gen_ai_endpoint: Endpoint identifier for logging and tracing.
|
22
|
+
version: Version of the monitoring package.
|
23
|
+
environment: Deployment environment (e.g., production, staging).
|
24
|
+
application_name: Name of the application using the OpenAI API.
|
25
|
+
tracer: OpenTelemetry tracer for creating spans.
|
26
|
+
pricing_info: Information used for calculating the cost of OpenAI usage.
|
27
|
+
trace_content: Flag indicating whether to trace the actual content.
|
28
|
+
|
29
|
+
Returns:
|
30
|
+
A function that wraps the chat method to add telemetry.
|
31
|
+
"""
|
32
|
+
|
33
|
+
async def wrapper(wrapped, instance, args, kwargs):
|
34
|
+
"""
|
35
|
+
Wraps the 'messages' API call to add telemetry.
|
36
|
+
|
37
|
+
This collects metrics such as execution time, cost, and token usage, and handles errors
|
38
|
+
gracefully, adding details to the trace for observability.
|
39
|
+
|
40
|
+
Args:
|
41
|
+
wrapped: The original 'messages' method to be wrapped.
|
42
|
+
instance: The instance of the class where the original method is defined.
|
43
|
+
args: Positional arguments for the 'messages' method.
|
44
|
+
kwargs: Keyword arguments for the 'messages' method.
|
45
|
+
|
46
|
+
Returns:
|
47
|
+
The response from the original 'messages' method.
|
48
|
+
"""
|
49
|
+
|
50
|
+
# Check if streaming is enabled for the API call
|
51
|
+
streaming = kwargs.get("stream", False)
|
52
|
+
|
53
|
+
# pylint: disable=no-else-return
|
54
|
+
if streaming:
|
55
|
+
# Special handling for streaming response to accommodate the nature of data flow
|
56
|
+
async def stream_generator():
|
57
|
+
with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
|
58
|
+
# Placeholder for aggregating streaming response
|
59
|
+
llmresponse = ""
|
60
|
+
|
61
|
+
# Loop through streaming events capturing relevant details
|
62
|
+
async for event in await wrapped(*args, **kwargs):
|
63
|
+
|
64
|
+
# Collect message IDs and input token from events
|
65
|
+
if event.type == "message_start":
|
66
|
+
response_id = event.message.id
|
67
|
+
prompt_tokens = event.message.usage.input_tokens
|
68
|
+
|
69
|
+
# Aggregate response content
|
70
|
+
if event.type == "content_block_delta":
|
71
|
+
llmresponse += event.delta.text
|
72
|
+
|
73
|
+
# Collect output tokens and stop reason from events
|
74
|
+
if event.type == "message_delta":
|
75
|
+
completion_tokens = event.usage.output_tokens
|
76
|
+
finish_reason = event.delta.stop_reason
|
77
|
+
yield event
|
78
|
+
|
79
|
+
# Handling exception ensure observability without disrupting operation
|
80
|
+
try:
|
81
|
+
# Format 'messages' into a single string
|
82
|
+
message_prompt = kwargs.get("messages", "")
|
83
|
+
formatted_messages = []
|
84
|
+
for message in message_prompt:
|
85
|
+
role = message["role"]
|
86
|
+
content = message["content"]
|
87
|
+
|
88
|
+
if isinstance(content, list):
|
89
|
+
content_str = ", ".join(
|
90
|
+
# pylint: disable=line-too-long
|
91
|
+
f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
|
92
|
+
if "type" in item else f'text: {item["text"]}'
|
93
|
+
for item in content
|
94
|
+
)
|
95
|
+
formatted_messages.append(f"{role}: {content_str}")
|
96
|
+
else:
|
97
|
+
formatted_messages.append(f"{role}: {content}")
|
98
|
+
prompt = "\n".join(formatted_messages)
|
99
|
+
|
100
|
+
# Calculate cost of the operation
|
101
|
+
cost = get_chat_model_cost(
|
102
|
+
kwargs.get("model", "claude-3-sonnet-20240229"),
|
103
|
+
pricing_info, prompt_tokens, completion_tokens
|
104
|
+
)
|
105
|
+
|
106
|
+
# Set Span attributes
|
107
|
+
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
|
108
|
+
span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
|
109
|
+
SemanticConvetion.GEN_AI_SYSTEM_ANTHROPIC)
|
110
|
+
span.set_attribute(SemanticConvetion.GEN_AI_TYPE,
|
111
|
+
SemanticConvetion.GEN_AI_TYPE_CHAT)
|
112
|
+
span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
|
113
|
+
gen_ai_endpoint)
|
114
|
+
span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
|
115
|
+
response_id)
|
116
|
+
span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
|
117
|
+
environment)
|
118
|
+
span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
|
119
|
+
application_name)
|
120
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
|
121
|
+
kwargs.get("model", "claude-3-sonnet-20240229"))
|
122
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
|
123
|
+
kwargs.get("max_tokens", ""))
|
124
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
|
125
|
+
True)
|
126
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
|
127
|
+
kwargs.get("temperature", 1.0))
|
128
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
|
129
|
+
kwargs.get("top_p", ""))
|
130
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_K,
|
131
|
+
kwargs.get("top_k", ""))
|
132
|
+
span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
|
133
|
+
finish_reason)
|
134
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
|
135
|
+
prompt_tokens)
|
136
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COMPLETION_TOKENS,
|
137
|
+
completion_tokens)
|
138
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
|
139
|
+
prompt_tokens + completion_tokens)
|
140
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
|
141
|
+
cost)
|
142
|
+
if trace_content:
|
143
|
+
span.set_attribute(SemanticConvetion.GEN_AI_CONTENT_PROMPT,
|
144
|
+
prompt)
|
145
|
+
span.set_attribute(SemanticConvetion.GEN_AI_CONTENT_COMPLETION,
|
146
|
+
llmresponse)
|
147
|
+
|
148
|
+
span.set_status(Status(StatusCode.OK))
|
149
|
+
|
150
|
+
if disable_metrics is False:
|
151
|
+
attributes = {
|
152
|
+
TELEMETRY_SDK_NAME:
|
153
|
+
"openlit",
|
154
|
+
SemanticConvetion.GEN_AI_APPLICATION_NAME:
|
155
|
+
application_name,
|
156
|
+
SemanticConvetion.GEN_AI_SYSTEM:
|
157
|
+
SemanticConvetion.GEN_AI_SYSTEM_ANTHROPIC,
|
158
|
+
SemanticConvetion.GEN_AI_ENVIRONMENT:
|
159
|
+
environment,
|
160
|
+
SemanticConvetion.GEN_AI_TYPE:
|
161
|
+
SemanticConvetion.GEN_AI_TYPE_CHAT,
|
162
|
+
SemanticConvetion.GEN_AI_REQUEST_MODEL:
|
163
|
+
kwargs.get("model", "claude-3-sonnet-20240229")
|
164
|
+
}
|
165
|
+
|
166
|
+
metrics["genai_requests"].add(1, attributes)
|
167
|
+
metrics["genai_total_tokens"].add(
|
168
|
+
prompt_tokens + completion_tokens, attributes
|
169
|
+
)
|
170
|
+
metrics["genai_completion_tokens"].add(completion_tokens, attributes)
|
171
|
+
metrics["genai_prompt_tokens"].add(prompt_tokens, attributes)
|
172
|
+
metrics["genai_cost"].record(cost, attributes)
|
173
|
+
|
174
|
+
except Exception as e:
|
175
|
+
handle_exception(span, e)
|
176
|
+
logger.error("Error in trace creation: %s", e)
|
177
|
+
|
178
|
+
return stream_generator()
|
179
|
+
|
180
|
+
# Handling for non-streaming responses
|
181
|
+
else:
|
182
|
+
with tracer.start_as_current_span(gen_ai_endpoint, kind=SpanKind.CLIENT) as span:
|
183
|
+
response = await wrapped(*args, **kwargs)
|
184
|
+
|
185
|
+
try:
|
186
|
+
# Format 'messages' into a single string
|
187
|
+
message_prompt = kwargs.get("messages", "")
|
188
|
+
formatted_messages = []
|
189
|
+
for message in message_prompt:
|
190
|
+
role = message["role"]
|
191
|
+
content = message["content"]
|
192
|
+
|
193
|
+
if isinstance(content, list):
|
194
|
+
content_str = ", ".join(
|
195
|
+
# pylint: disable=line-too-long
|
196
|
+
f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
|
197
|
+
if "type" in item else f'text: {item["text"]}'
|
198
|
+
for item in content
|
199
|
+
)
|
200
|
+
formatted_messages.append(f"{role}: {content_str}")
|
201
|
+
else:
|
202
|
+
formatted_messages.append(f"{role}: {content}")
|
203
|
+
prompt = "\n".join(formatted_messages)
|
204
|
+
|
205
|
+
# Calculate cost of the operation
|
206
|
+
cost = get_chat_model_cost(kwargs.get("model", "claude-3-sonnet-20240229"),
|
207
|
+
pricing_info, response.usage.input_tokens,
|
208
|
+
response.usage.output_tokens)
|
209
|
+
|
210
|
+
# Set Span attribues
|
211
|
+
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
|
212
|
+
span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
|
213
|
+
SemanticConvetion.GEN_AI_SYSTEM_ANTHROPIC)
|
214
|
+
span.set_attribute(SemanticConvetion.GEN_AI_TYPE,
|
215
|
+
SemanticConvetion.GEN_AI_TYPE_CHAT)
|
216
|
+
span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
|
217
|
+
gen_ai_endpoint)
|
218
|
+
span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
|
219
|
+
response.id)
|
220
|
+
span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
|
221
|
+
environment)
|
222
|
+
span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
|
223
|
+
application_name)
|
224
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
|
225
|
+
kwargs.get("model", "claude-3-sonnet-20240229"))
|
226
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
|
227
|
+
kwargs.get("max_tokens", ""))
|
228
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
|
229
|
+
False)
|
230
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
|
231
|
+
kwargs.get("temperature", 1.0))
|
232
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
|
233
|
+
kwargs.get("top_p", ""))
|
234
|
+
span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_K,
|
235
|
+
kwargs.get("top_k", ""))
|
236
|
+
span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
|
237
|
+
response.stop_reason)
|
238
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
|
239
|
+
response.usage.input_tokens)
|
240
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COMPLETION_TOKENS,
|
241
|
+
response.usage.output_tokens)
|
242
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
|
243
|
+
response.usage.input_tokens +
|
244
|
+
response.usage.output_tokens)
|
245
|
+
span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
|
246
|
+
cost)
|
247
|
+
if trace_content:
|
248
|
+
span.set_attribute(SemanticConvetion.GEN_AI_CONTENT_PROMPT,
|
249
|
+
prompt)
|
250
|
+
# pylint: disable=line-too-long
|
251
|
+
span.set_attribute(SemanticConvetion.GEN_AI_CONTENT_COMPLETION, response.content[0].text if response.content else "")
|
252
|
+
|
253
|
+
span.set_status(Status(StatusCode.OK))
|
254
|
+
|
255
|
+
if disable_metrics is False:
|
256
|
+
attributes = {
|
257
|
+
TELEMETRY_SDK_NAME:
|
258
|
+
"openlit",
|
259
|
+
SemanticConvetion.GEN_AI_APPLICATION_NAME:
|
260
|
+
application_name,
|
261
|
+
SemanticConvetion.GEN_AI_SYSTEM:
|
262
|
+
SemanticConvetion.GEN_AI_SYSTEM_ANTHROPIC,
|
263
|
+
SemanticConvetion.GEN_AI_ENVIRONMENT:
|
264
|
+
environment,
|
265
|
+
SemanticConvetion.GEN_AI_TYPE:
|
266
|
+
SemanticConvetion.GEN_AI_TYPE_CHAT,
|
267
|
+
SemanticConvetion.GEN_AI_REQUEST_MODEL:
|
268
|
+
kwargs.get("model", "claude-3-sonnet-20240229")
|
269
|
+
}
|
270
|
+
|
271
|
+
metrics["genai_requests"].add(1, attributes)
|
272
|
+
metrics["genai_total_tokens"].add(
|
273
|
+
response.usage.input_tokens +
|
274
|
+
response.usage.output_tokens, attributes)
|
275
|
+
metrics["genai_completion_tokens"].add(
|
276
|
+
response.usage.output_tokens, attributes)
|
277
|
+
metrics["genai_prompt_tokens"].add(
|
278
|
+
response.usage.input_tokens, attributes)
|
279
|
+
metrics["genai_cost"].record(cost, attributes)
|
280
|
+
|
281
|
+
# Return original response
|
282
|
+
return response
|
283
|
+
|
284
|
+
except Exception as e:
|
285
|
+
handle_exception(span, e)
|
286
|
+
logger.error("Error in trace creation: %s", e)
|
287
|
+
|
288
|
+
# Return original response
|
289
|
+
return response
|
290
|
+
|
291
|
+
return wrapper
|
@@ -0,0 +1,86 @@
|
|
1
|
+
# pylint: disable=useless-return, bad-staticmethod-argument, disable=duplicate-code
|
2
|
+
"""Initializer of Auto Instrumentation of Chromadb Functions"""
|
3
|
+
from typing import Collection
|
4
|
+
import importlib.metadata
|
5
|
+
from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
|
6
|
+
from wrapt import wrap_function_wrapper
|
7
|
+
|
8
|
+
from openlit.instrumentation.chroma.chroma import general_wrap
|
9
|
+
|
10
|
+
_instruments = ("chromadb >= 0.4.0",)
|
11
|
+
|
12
|
+
class ChromaInstrumentor(BaseInstrumentor):
|
13
|
+
"""An instrumentor for Chromadb's client library."""
|
14
|
+
|
15
|
+
def instrumentation_dependencies(self) -> Collection[str]:
|
16
|
+
return _instruments
|
17
|
+
|
18
|
+
def _instrument(self, **kwargs):
|
19
|
+
application_name = kwargs.get("application_name")
|
20
|
+
environment = kwargs.get("environment")
|
21
|
+
tracer = kwargs.get("tracer")
|
22
|
+
metrics = kwargs.get("metrics_dict")
|
23
|
+
pricing_info = kwargs.get("pricing_info")
|
24
|
+
trace_content = kwargs.get("trace_content")
|
25
|
+
disable_metrics = kwargs.get("disable_metrics")
|
26
|
+
version = importlib.metadata.version("chromadb")
|
27
|
+
|
28
|
+
wrap_function_wrapper(
|
29
|
+
"chromadb.db",
|
30
|
+
"DB.create_collection",
|
31
|
+
general_wrap("chroma.create_collection", version, environment, application_name,
|
32
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
33
|
+
)
|
34
|
+
|
35
|
+
wrap_function_wrapper(
|
36
|
+
"chromadb",
|
37
|
+
"Collection.add",
|
38
|
+
general_wrap("chroma.add", version, environment, application_name,
|
39
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
40
|
+
)
|
41
|
+
|
42
|
+
wrap_function_wrapper(
|
43
|
+
"chromadb",
|
44
|
+
"Collection.get",
|
45
|
+
general_wrap("chroma.get", version, environment, application_name,
|
46
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
47
|
+
)
|
48
|
+
|
49
|
+
wrap_function_wrapper(
|
50
|
+
"chromadb",
|
51
|
+
"Collection.peek",
|
52
|
+
general_wrap("chroma.peek", version, environment, application_name,
|
53
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
54
|
+
)
|
55
|
+
|
56
|
+
wrap_function_wrapper(
|
57
|
+
"chromadb",
|
58
|
+
"Collection.query",
|
59
|
+
general_wrap("chroma.query", version, environment, application_name,
|
60
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
61
|
+
)
|
62
|
+
|
63
|
+
wrap_function_wrapper(
|
64
|
+
"chromadb",
|
65
|
+
"Collection.update",
|
66
|
+
general_wrap("chroma.update", version, environment, application_name,
|
67
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
68
|
+
)
|
69
|
+
|
70
|
+
wrap_function_wrapper(
|
71
|
+
"chromadb",
|
72
|
+
"Collection.upsert",
|
73
|
+
general_wrap("chroma.upsert", version, environment, application_name,
|
74
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
75
|
+
)
|
76
|
+
wrap_function_wrapper(
|
77
|
+
"chromadb",
|
78
|
+
"Collection.delete",
|
79
|
+
general_wrap("chroma.delete", version, environment, application_name,
|
80
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
81
|
+
)
|
82
|
+
|
83
|
+
|
84
|
+
@staticmethod
|
85
|
+
def _uninstrument(self, **kwargs):
|
86
|
+
pass
|
@@ -0,0 +1,197 @@
|
|
1
|
+
# pylint: disable=duplicate-code, broad-exception-caught, too-many-statements, unused-argument
|
2
|
+
"""
|
3
|
+
Module for monitoring ChromaDB.
|
4
|
+
"""
|
5
|
+
|
6
|
+
import logging
|
7
|
+
from opentelemetry.trace import SpanKind, Status, StatusCode
|
8
|
+
from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
|
9
|
+
from openlit.__helpers import handle_exception
|
10
|
+
from openlit.semcov import SemanticConvetion
|
11
|
+
|
12
|
+
# Initialize logger for logging potential issues and operations
|
13
|
+
logger = logging.getLogger(__name__)
|
14
|
+
|
15
|
+
def object_count(obj):
|
16
|
+
"""
|
17
|
+
Counts Length of object if it exists, Else returns None
|
18
|
+
"""
|
19
|
+
|
20
|
+
if obj:
|
21
|
+
return len(obj)
|
22
|
+
|
23
|
+
return None
|
24
|
+
|
25
|
+
def general_wrap(gen_ai_endpoint, version, environment, application_name,
|
26
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics):
|
27
|
+
"""
|
28
|
+
Creates a wrapper around a function call to trace and log its execution metrics.
|
29
|
+
|
30
|
+
This function wraps any given function to measure its execution time,
|
31
|
+
log its operation, and trace its execution using OpenTelemetry.
|
32
|
+
|
33
|
+
Parameters:
|
34
|
+
- gen_ai_endpoint (str): A descriptor or name for the endpoint being traced.
|
35
|
+
- version (str): The version of the Langchain application.
|
36
|
+
- environment (str): The deployment environment (e.g., 'production', 'development').
|
37
|
+
- application_name (str): Name of the Langchain application.
|
38
|
+
- tracer (opentelemetry.trace.Tracer): The tracer object used for OpenTelemetry tracing.
|
39
|
+
- pricing_info (dict): Information about the pricing for internal metrics (currently not used).
|
40
|
+
- trace_content (bool): Flag indicating whether to trace the content of the response.
|
41
|
+
|
42
|
+
Returns:
|
43
|
+
- function: A higher-order function that takes a function 'wrapped' and returns
|
44
|
+
a new function that wraps 'wrapped' with additional tracing and logging.
|
45
|
+
"""
|
46
|
+
|
47
|
+
def wrapper(wrapped, instance, args, kwargs):
|
48
|
+
"""
|
49
|
+
An inner wrapper function that executes the wrapped function, measures execution
|
50
|
+
time, and records trace data using OpenTelemetry.
|
51
|
+
|
52
|
+
Parameters:
|
53
|
+
- wrapped (Callable): The original function that this wrapper will execute.
|
54
|
+
- instance (object): The instance to which the wrapped function belongs. This
|
55
|
+
is used for instance methods. For static and classmethods,
|
56
|
+
this may be None.
|
57
|
+
- args (tuple): Positional arguments passed to the wrapped function.
|
58
|
+
- kwargs (dict): Keyword arguments passed to the wrapped function.
|
59
|
+
|
60
|
+
Returns:
|
61
|
+
- The result of the wrapped function call.
|
62
|
+
|
63
|
+
The wrapper initiates a span with the provided tracer, sets various attributes
|
64
|
+
on the span based on the function's execution and response, and ensures
|
65
|
+
errors are handled and logged appropriately.
|
66
|
+
"""
|
67
|
+
with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
|
68
|
+
response = wrapped(*args, **kwargs)
|
69
|
+
|
70
|
+
try:
|
71
|
+
span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
|
72
|
+
span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
|
73
|
+
gen_ai_endpoint)
|
74
|
+
span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
|
75
|
+
environment)
|
76
|
+
span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
|
77
|
+
application_name)
|
78
|
+
span.set_attribute(SemanticConvetion.GEN_AI_TYPE,
|
79
|
+
SemanticConvetion.GEN_AI_TYPE_VECTORDB)
|
80
|
+
span.set_attribute(SemanticConvetion.DB_SYSTEM,
|
81
|
+
SemanticConvetion.DB_SYSTEM_CHROMA)
|
82
|
+
span.set_attribute(SemanticConvetion.DB_COLLECTION_NAME,
|
83
|
+
instance.name)
|
84
|
+
|
85
|
+
if gen_ai_endpoint == "chroma.add":
|
86
|
+
db_operation = SemanticConvetion.DB_OPERATION_ADD
|
87
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
88
|
+
SemanticConvetion.DB_OPERATION_GET)
|
89
|
+
span.set_attribute(SemanticConvetion.DB_ID_COUNT,
|
90
|
+
object_count(kwargs.get("ids")))
|
91
|
+
span.set_attribute(SemanticConvetion.DB_VECTOR_COUNT,
|
92
|
+
object_count(kwargs.get("embeddings")))
|
93
|
+
span.set_attribute(SemanticConvetion.DB_VECTOR_COUNT,
|
94
|
+
object_count(kwargs.get("metadatas")))
|
95
|
+
span.set_attribute(SemanticConvetion.DB_DOCUMENTS_COUNT,
|
96
|
+
object_count(kwargs.get("documents")))
|
97
|
+
|
98
|
+
elif gen_ai_endpoint == "chroma.get":
|
99
|
+
db_operation = SemanticConvetion.DB_OPERATION_GET
|
100
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
101
|
+
SemanticConvetion.DB_OPERATION_GET)
|
102
|
+
span.set_attribute(SemanticConvetion.DB_ID_COUNT,
|
103
|
+
object_count(kwargs.get("ids")))
|
104
|
+
span.set_attribute(SemanticConvetion.DB_QUERY_LIMIT,
|
105
|
+
kwargs.get("limit"))
|
106
|
+
span.set_attribute(SemanticConvetion.DB_OFFSET,
|
107
|
+
kwargs.get("offset"))
|
108
|
+
span.set_attribute(SemanticConvetion.DB_WHERE_DOCUMENT,
|
109
|
+
str(kwargs.get("where_document", "")))
|
110
|
+
|
111
|
+
elif gen_ai_endpoint == "chroma.query":
|
112
|
+
db_operation = SemanticConvetion.DB_OPERATION_QUERY
|
113
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
114
|
+
SemanticConvetion.DB_OPERATION_QUERY)
|
115
|
+
span.set_attribute(SemanticConvetion.DB_STATEMENT,
|
116
|
+
str(kwargs.get("query_texts")))
|
117
|
+
span.set_attribute(SemanticConvetion.DB_N_RESULTS,
|
118
|
+
kwargs.get("n_results", ""))
|
119
|
+
span.set_attribute(SemanticConvetion.DB_FILTER,
|
120
|
+
str(kwargs.get("where", "")))
|
121
|
+
span.set_attribute(SemanticConvetion.DB_WHERE_DOCUMENT,
|
122
|
+
str(kwargs.get("where_document", "")))
|
123
|
+
|
124
|
+
elif gen_ai_endpoint == "chroma.update":
|
125
|
+
db_operation = SemanticConvetion.DB_OPERATION_UPDATE
|
126
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
127
|
+
SemanticConvetion.DB_OPERATION_UPDATE)
|
128
|
+
span.set_attribute(SemanticConvetion.DB_VECTOR_COUNT,
|
129
|
+
object_count(kwargs.get("embeddings")))
|
130
|
+
span.set_attribute(SemanticConvetion.DB_VECTOR_COUNT,
|
131
|
+
object_count(kwargs.get("metadatas")))
|
132
|
+
span.set_attribute(SemanticConvetion.DB_ID_COUNT,
|
133
|
+
object_count(kwargs.get("ids")))
|
134
|
+
span.set_attribute(SemanticConvetion.DB_DOCUMENTS_COUNT,
|
135
|
+
object_count(kwargs.get("documents")))
|
136
|
+
|
137
|
+
elif gen_ai_endpoint == "chroma.upsert":
|
138
|
+
db_operation = SemanticConvetion.DB_OPERATION_UPSERT
|
139
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
140
|
+
SemanticConvetion.DB_OPERATION_UPSERT)
|
141
|
+
span.set_attribute(SemanticConvetion.DB_VECTOR_COUNT,
|
142
|
+
object_count(kwargs.get("embeddings")))
|
143
|
+
span.set_attribute(SemanticConvetion.DB_VECTOR_COUNT,
|
144
|
+
object_count(kwargs.get("metadatas")))
|
145
|
+
span.set_attribute(SemanticConvetion.DB_ID_COUNT,
|
146
|
+
object_count(kwargs.get("ids")))
|
147
|
+
span.set_attribute(SemanticConvetion.DB_DOCUMENTS_COUNT,
|
148
|
+
object_count(kwargs.get("documents")))
|
149
|
+
|
150
|
+
elif gen_ai_endpoint == "chroma.delete":
|
151
|
+
db_operation = SemanticConvetion.DB_OPERATION_DELETE
|
152
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
153
|
+
SemanticConvetion.DB_OPERATION_DELETE)
|
154
|
+
span.set_attribute(SemanticConvetion.DB_ID_COUNT,
|
155
|
+
object_count(kwargs.get("ids")))
|
156
|
+
span.set_attribute(SemanticConvetion.DB_FILTER,
|
157
|
+
str(kwargs.get("where", "")))
|
158
|
+
span.set_attribute(SemanticConvetion.DB_DELETE_ALL,
|
159
|
+
kwargs.get("delete_all", False))
|
160
|
+
span.set_attribute(SemanticConvetion.DB_WHERE_DOCUMENT,
|
161
|
+
str(kwargs.get("where_document", "")))
|
162
|
+
|
163
|
+
elif gen_ai_endpoint == "chroma.peek":
|
164
|
+
db_operation = SemanticConvetion.DB_OPERATION_PEEK
|
165
|
+
span.set_attribute(SemanticConvetion.DB_OPERATION,
|
166
|
+
SemanticConvetion.DB_OPERATION_PEEK)
|
167
|
+
|
168
|
+
span.set_status(Status(StatusCode.OK))
|
169
|
+
|
170
|
+
if disable_metrics is False:
|
171
|
+
attributes = {
|
172
|
+
TELEMETRY_SDK_NAME:
|
173
|
+
"openlit",
|
174
|
+
SemanticConvetion.GEN_AI_APPLICATION_NAME:
|
175
|
+
application_name,
|
176
|
+
SemanticConvetion.DB_SYSTEM:
|
177
|
+
SemanticConvetion.DB_SYSTEM_CHROMA,
|
178
|
+
SemanticConvetion.GEN_AI_ENVIRONMENT:
|
179
|
+
environment,
|
180
|
+
SemanticConvetion.GEN_AI_TYPE:
|
181
|
+
SemanticConvetion.GEN_AI_TYPE_VECTORDB,
|
182
|
+
SemanticConvetion.DB_OPERATION:
|
183
|
+
db_operation
|
184
|
+
}
|
185
|
+
|
186
|
+
metrics["db_requests"].add(1, attributes)
|
187
|
+
|
188
|
+
return response
|
189
|
+
|
190
|
+
except Exception as e:
|
191
|
+
handle_exception(span, e)
|
192
|
+
logger.error("Error in trace creation: %s", e)
|
193
|
+
|
194
|
+
# Return original response
|
195
|
+
return response
|
196
|
+
|
197
|
+
return wrapper
|
@@ -0,0 +1,51 @@
|
|
1
|
+
# pylint: disable=useless-return, bad-staticmethod-argument, disable=duplicate-code
|
2
|
+
"""Initializer of Auto Instrumentation of Cohere Functions"""
|
3
|
+
from typing import Collection
|
4
|
+
import importlib.metadata
|
5
|
+
from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
|
6
|
+
from wrapt import wrap_function_wrapper
|
7
|
+
|
8
|
+
from openlit.instrumentation.cohere.cohere import chat, chat_stream, embed
|
9
|
+
|
10
|
+
_instruments = ("cohere >= 5.0.0",)
|
11
|
+
|
12
|
+
class CohereInstrumentor(BaseInstrumentor):
|
13
|
+
"""An instrumentor for Cohere's client library."""
|
14
|
+
|
15
|
+
def instrumentation_dependencies(self) -> Collection[str]:
|
16
|
+
return _instruments
|
17
|
+
|
18
|
+
def _instrument(self, **kwargs):
|
19
|
+
application_name = kwargs.get("application_name")
|
20
|
+
environment = kwargs.get("environment")
|
21
|
+
tracer = kwargs.get("tracer")
|
22
|
+
metrics = kwargs.get("metrics_dict")
|
23
|
+
pricing_info = kwargs.get("pricing_info")
|
24
|
+
trace_content = kwargs.get("trace_content")
|
25
|
+
disable_metrics = kwargs.get("disable_metrics")
|
26
|
+
version = importlib.metadata.version("cohere")
|
27
|
+
|
28
|
+
wrap_function_wrapper(
|
29
|
+
"cohere.client",
|
30
|
+
"Client.chat",
|
31
|
+
chat("cohere.chat", version, environment, application_name,
|
32
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
33
|
+
)
|
34
|
+
|
35
|
+
wrap_function_wrapper(
|
36
|
+
"cohere.client",
|
37
|
+
"Client.chat_stream",
|
38
|
+
chat_stream("cohere.chat", version, environment, application_name,
|
39
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
40
|
+
)
|
41
|
+
|
42
|
+
wrap_function_wrapper(
|
43
|
+
"cohere.client",
|
44
|
+
"Client.embed",
|
45
|
+
embed("cohere.embed", version, environment, application_name,
|
46
|
+
tracer, pricing_info, trace_content, metrics, disable_metrics),
|
47
|
+
)
|
48
|
+
|
49
|
+
@staticmethod
|
50
|
+
def _uninstrument(self, **kwargs):
|
51
|
+
pass
|