openlit 1.29.1__py3-none-any.whl → 1.30.2__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.
Files changed (21) hide show
  1. openlit/__helpers.py +14 -0
  2. openlit/__init__.py +11 -3
  3. openlit/evals/utils.py +5 -2
  4. openlit/guard/utils.py +5 -1
  5. openlit/instrumentation/azure_ai_inference/async_azure_ai_inference.py +0 -1
  6. openlit/instrumentation/azure_ai_inference/azure_ai_inference.py +0 -1
  7. openlit/instrumentation/crewai/__init__.py +50 -0
  8. openlit/instrumentation/crewai/crewai.py +149 -0
  9. openlit/instrumentation/google_ai_studio/google_ai_studio.py +0 -1
  10. openlit/instrumentation/litellm/__init__.py +54 -0
  11. openlit/instrumentation/litellm/async_litellm.py +406 -0
  12. openlit/instrumentation/litellm/litellm.py +406 -0
  13. openlit/instrumentation/openai/async_openai.py +236 -170
  14. openlit/instrumentation/openai/openai.py +208 -156
  15. openlit/otel/metrics.py +6 -3
  16. openlit/otel/tracing.py +5 -2
  17. openlit/semcov/__init__.py +21 -0
  18. {openlit-1.29.1.dist-info → openlit-1.30.2.dist-info}/METADATA +11 -11
  19. {openlit-1.29.1.dist-info → openlit-1.30.2.dist-info}/RECORD +21 -16
  20. {openlit-1.29.1.dist-info → openlit-1.30.2.dist-info}/LICENSE +0 -0
  21. {openlit-1.29.1.dist-info → openlit-1.30.2.dist-info}/WHEEL +0 -0
openlit/instrumentation/openai/openai.py CHANGED
@@ -6,8 +6,15 @@ Module for monitoring OpenAI API calls.
6
6
  import logging
7
7
  from opentelemetry.trace import SpanKind, Status, StatusCode
8
8
  from opentelemetry.sdk.resources import TELEMETRY_SDK_NAME
9
- from openlit.__helpers import get_chat_model_cost, get_embed_model_cost, get_audio_model_cost
10
- from openlit.__helpers import get_image_model_cost, openai_tokens, handle_exception
9
+ from openlit.__helpers import (
10
+ get_chat_model_cost,
11
+ get_embed_model_cost,
12
+ get_audio_model_cost,
13
+ get_image_model_cost,
14
+ openai_tokens,
15
+ handle_exception,
16
+ response_as_dict,
17
+ )
11
18
  from openlit.semcov import SemanticConvetion
12
19
 
13
20
  # Initialize logger for logging potential issues and operations
@@ -31,10 +38,184 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
31
38
  A function that wraps the chat completions method to add telemetry.
32
39
  """
33
40
 
41
+ class TracedSyncStream:
42
+ """
43
+ Wrapper for streaming responses to collect metrics and trace data.
44
+ Wraps the 'openai.AsyncStream' response to collect message IDs and aggregated response.
45
+
46
+ This class implements the '__aiter__' and '__anext__' methods that
47
+ handle asynchronous streaming responses.
48
+
49
+ This class also implements '__aenter__' and '__aexit__' methods that
50
+ handle asynchronous context management protocol.
51
+ """
52
+ def __init__(
53
+ self,
54
+ wrapped,
55
+ span,
56
+ kwargs,
57
+ **args,
58
+ ):
59
+ self.__wrapped__ = wrapped
60
+ self._span = span
61
+ # Placeholder for aggregating streaming response
62
+ self._llmresponse = ""
63
+ self._response_id = ""
64
+
65
+ self._args = args
66
+ self._kwargs = kwargs
67
+
68
+ def __enter__(self):
69
+ self.__wrapped__.__enter__()
70
+ return self
71
+
72
+ def __exit__(self, exc_type, exc_value, traceback):
73
+ self.__wrapped__.__exit__(exc_type, exc_value, traceback)
74
+
75
+ def __iter__(self):
76
+ return self
77
+
78
+ def __getattr__(self, name):
79
+ """Delegate attribute access to the wrapped object."""
80
+ return getattr(self.__wrapped__, name)
81
+
82
+ def __next__(self):
83
+ try:
84
+ chunk = self.__wrapped__.__next__()
85
+ chunked = response_as_dict(chunk)
86
+ # Collect message IDs and aggregated response from events
87
+ if (len(chunked.get('choices')) > 0 and ('delta' in chunked.get('choices')[0] and
88
+ 'content' in chunked.get('choices')[0].get('delta'))):
89
+
90
+ content = chunked.get('choices')[0].get('delta').get('content')
91
+ if content:
92
+ self._llmresponse += content
93
+ self._response_id = chunked.get('id')
94
+ return chunk
95
+ except StopIteration:
96
+ # Handling exception ensure observability without disrupting operation
97
+ try:
98
+ # Format 'messages' into a single string
99
+ message_prompt = self._kwargs.get("messages", "")
100
+ formatted_messages = []
101
+ for message in message_prompt:
102
+ role = message["role"]
103
+ content = message["content"]
104
+
105
+ if isinstance(content, list):
106
+ content_str = ", ".join(
107
+ # pylint: disable=line-too-long
108
+ f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
109
+ if "type" in item else f'text: {item["text"]}'
110
+ for item in content
111
+ )
112
+ formatted_messages.append(f"{role}: {content_str}")
113
+ else:
114
+ formatted_messages.append(f"{role}: {content}")
115
+ prompt = "\n".join(formatted_messages)
116
+
117
+ # Calculate tokens using input prompt and aggregated response
118
+ prompt_tokens = openai_tokens(prompt,
119
+ self._kwargs.get("model", "gpt-3.5-turbo"))
120
+ completion_tokens = openai_tokens(self._llmresponse,
121
+ self._kwargs.get("model", "gpt-3.5-turbo"))
122
+
123
+ # Calculate cost of the operation
124
+ cost = get_chat_model_cost(self._kwargs.get("model", "gpt-3.5-turbo"),
125
+ pricing_info, prompt_tokens,
126
+ completion_tokens)
127
+
128
+ # Set Span attributes
129
+ self._span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
130
+ self._span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
131
+ SemanticConvetion.GEN_AI_SYSTEM_OPENAI)
132
+ self._span.set_attribute(SemanticConvetion.GEN_AI_TYPE,
133
+ SemanticConvetion.GEN_AI_TYPE_CHAT)
134
+ self._span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
135
+ gen_ai_endpoint)
136
+ self._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
137
+ self._response_id)
138
+ self._span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
139
+ environment)
140
+ self._span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
141
+ application_name)
142
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
143
+ self._kwargs.get("model", "gpt-3.5-turbo"))
144
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
145
+ self._kwargs.get("user", ""))
146
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
147
+ self._kwargs.get("top_p", 1.0))
148
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
149
+ self._kwargs.get("max_tokens", -1))
150
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
151
+ self._kwargs.get("temperature", 1.0))
152
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
153
+ self._kwargs.get("presence_penalty", 0.0))
154
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
155
+ self._kwargs.get("frequency_penalty", 0.0))
156
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
157
+ self._kwargs.get("seed", ""))
158
+ self._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
159
+ True)
160
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
161
+ prompt_tokens)
162
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COMPLETION_TOKENS,
163
+ completion_tokens)
164
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
165
+ prompt_tokens + completion_tokens)
166
+ self._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
167
+ cost)
168
+ if trace_content:
169
+ self._span.add_event(
170
+ name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
171
+ attributes={
172
+ SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
173
+ },
174
+ )
175
+ self._span.add_event(
176
+ name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
177
+ attributes={
178
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: self._llmresponse,
179
+ },
180
+ )
181
+
182
+ self._span.set_status(Status(StatusCode.OK))
183
+
184
+ if disable_metrics is False:
185
+ attributes = {
186
+ TELEMETRY_SDK_NAME:
187
+ "openlit",
188
+ SemanticConvetion.GEN_AI_APPLICATION_NAME:
189
+ application_name,
190
+ SemanticConvetion.GEN_AI_SYSTEM:
191
+ SemanticConvetion.GEN_AI_SYSTEM_OPENAI,
192
+ SemanticConvetion.GEN_AI_ENVIRONMENT:
193
+ environment,
194
+ SemanticConvetion.GEN_AI_TYPE:
195
+ SemanticConvetion.GEN_AI_TYPE_CHAT,
196
+ SemanticConvetion.GEN_AI_REQUEST_MODEL:
197
+ self._kwargs.get("model", "gpt-3.5-turbo")
198
+ }
199
+
200
+ metrics["genai_requests"].add(1, attributes)
201
+ metrics["genai_total_tokens"].add(
202
+ prompt_tokens + completion_tokens, attributes
203
+ )
204
+ metrics["genai_completion_tokens"].add(completion_tokens, attributes)
205
+ metrics["genai_prompt_tokens"].add(prompt_tokens, attributes)
206
+ metrics["genai_cost"].record(cost, attributes)
207
+
208
+ except Exception as e:
209
+ handle_exception(self._span, e)
210
+ logger.error("Error in trace creation: %s", e)
211
+ finally:
212
+ self._span.end()
213
+ raise
214
+
34
215
  def wrapper(wrapped, instance, args, kwargs):
35
216
  """
36
217
  Wraps the 'chat.completions' API call to add telemetry.
37
-
218
+
38
219
  This collects metrics such as execution time, cost, and token usage, and handles errors
39
220
  gracefully, adding details to the trace for observability.
40
221
 
@@ -54,141 +235,10 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
54
235
  # pylint: disable=no-else-return
55
236
  if streaming:
56
237
  # Special handling for streaming response to accommodate the nature of data flow
57
- def stream_generator():
58
- with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
59
- # Placeholder for aggregating streaming response
60
- llmresponse = ""
61
-
62
-
63
- # Loop through streaming events capturing relevant details
64
- for chunk in wrapped(*args, **kwargs):
65
- # Collect message IDs and aggregated response from events
66
- if len(chunk.choices) > 0:
67
- # pylint: disable=line-too-long
68
- if hasattr(chunk.choices[0], "delta") and hasattr(chunk.choices[0].delta, "content"):
69
- content = chunk.choices[0].delta.content
70
- if content:
71
- llmresponse += content
72
- yield chunk
73
- response_id = chunk.id
74
-
75
- # Handling exception ensure observability without disrupting operation
76
- try:
77
- # Format 'messages' into a single string
78
- message_prompt = kwargs.get("messages", "")
79
- formatted_messages = []
80
- for message in message_prompt:
81
- role = message["role"]
82
- content = message["content"]
83
-
84
- if isinstance(content, list):
85
- content_str = ", ".join(
86
- # pylint: disable=line-too-long
87
- f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
88
- if "type" in item else f'text: {item["text"]}'
89
- for item in content
90
- )
91
- formatted_messages.append(f"{role}: {content_str}")
92
- else:
93
- formatted_messages.append(f"{role}: {content}")
94
- prompt = "\n".join(formatted_messages)
95
-
96
- # Calculate tokens using input prompt and aggregated response
97
- prompt_tokens = openai_tokens(prompt,
98
- kwargs.get("model", "gpt-3.5-turbo"))
99
- completion_tokens = openai_tokens(llmresponse,
100
- kwargs.get("model", "gpt-3.5-turbo"))
101
-
102
- # Calculate cost of the operation
103
- cost = get_chat_model_cost(kwargs.get("model", "gpt-3.5-turbo"),
104
- pricing_info, prompt_tokens,
105
- completion_tokens)
106
-
107
- # Set Span attributes
108
- span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
109
- span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM,
110
- SemanticConvetion.GEN_AI_SYSTEM_OPENAI)
111
- span.set_attribute(SemanticConvetion.GEN_AI_TYPE,
112
- SemanticConvetion.GEN_AI_TYPE_CHAT)
113
- span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
114
- gen_ai_endpoint)
115
- span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
116
- response_id)
117
- span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
118
- environment)
119
- span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
120
- application_name)
121
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL,
122
- kwargs.get("model", "gpt-3.5-turbo"))
123
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
124
- kwargs.get("user", ""))
125
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P,
126
- kwargs.get("top_p", 1.0))
127
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS,
128
- kwargs.get("max_tokens", -1))
129
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE,
130
- kwargs.get("temperature", 1.0))
131
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
132
- kwargs.get("presence_penalty", 0.0))
133
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
134
- kwargs.get("frequency_penalty", 0.0))
135
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_SEED,
136
- kwargs.get("seed", ""))
137
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM,
138
- True)
139
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
140
- prompt_tokens)
141
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COMPLETION_TOKENS,
142
- completion_tokens)
143
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
144
- prompt_tokens + completion_tokens)
145
- span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
146
- cost)
147
- if trace_content:
148
- span.add_event(
149
- name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
150
- attributes={
151
- SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
152
- },
153
- )
154
- span.add_event(
155
- name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
156
- attributes={
157
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: llmresponse,
158
- },
159
- )
160
-
161
- span.set_status(Status(StatusCode.OK))
162
-
163
- if disable_metrics is False:
164
- attributes = {
165
- TELEMETRY_SDK_NAME:
166
- "openlit",
167
- SemanticConvetion.GEN_AI_APPLICATION_NAME:
168
- application_name,
169
- SemanticConvetion.GEN_AI_SYSTEM:
170
- SemanticConvetion.GEN_AI_SYSTEM_OPENAI,
171
- SemanticConvetion.GEN_AI_ENVIRONMENT:
172
- environment,
173
- SemanticConvetion.GEN_AI_TYPE:
174
- SemanticConvetion.GEN_AI_TYPE_CHAT,
175
- SemanticConvetion.GEN_AI_REQUEST_MODEL:
176
- kwargs.get("model", "gpt-3.5-turbo")
177
- }
178
-
179
- metrics["genai_requests"].add(1, attributes)
180
- metrics["genai_total_tokens"].add(
181
- prompt_tokens + completion_tokens, attributes
182
- )
183
- metrics["genai_completion_tokens"].add(completion_tokens, attributes)
184
- metrics["genai_prompt_tokens"].add(prompt_tokens, attributes)
185
- metrics["genai_cost"].record(cost, attributes)
238
+ awaited_wrapped = wrapped(*args, **kwargs)
239
+ span = tracer.start_span(gen_ai_endpoint, kind=SpanKind.CLIENT)
186
240
 
187
- except Exception as e:
188
- handle_exception(span, e)
189
- logger.error("Error in trace creation: %s", e)
190
-
191
- return stream_generator()
241
+ return TracedSyncStream(awaited_wrapped, span, kwargs)
192
242
 
193
243
  # Handling for non-streaming responses
194
244
  else:
@@ -196,6 +246,8 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
196
246
  with tracer.start_as_current_span(gen_ai_endpoint, kind= SpanKind.CLIENT) as span:
197
247
  response = wrapped(*args, **kwargs)
198
248
 
249
+ response_dict = response_as_dict(response)
250
+
199
251
  try:
200
252
  # Format 'messages' into a single string
201
253
  message_prompt = kwargs.get("messages", "")
@@ -225,7 +277,7 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
225
277
  span.set_attribute(SemanticConvetion.GEN_AI_ENDPOINT,
226
278
  gen_ai_endpoint)
227
279
  span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID,
228
- response.id)
280
+ response_dict.get("id"))
229
281
  span.set_attribute(SemanticConvetion.GEN_AI_ENVIRONMENT,
230
282
  environment)
231
283
  span.set_attribute(SemanticConvetion.GEN_AI_APPLICATION_NAME,
@@ -260,17 +312,17 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
260
312
  if "tools" not in kwargs:
261
313
  # Calculate cost of the operation
262
314
  cost = get_chat_model_cost(kwargs.get("model", "gpt-3.5-turbo"),
263
- pricing_info, response.usage.prompt_tokens,
264
- response.usage.completion_tokens)
315
+ pricing_info, response_dict.get('usage', {}).get('prompt_tokens', None),
316
+ response_dict.get('usage', {}).get('completion_tokens', None))
265
317
 
266
318
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
267
- response.usage.prompt_tokens)
319
+ response_dict.get('usage', {}).get('prompt_tokens', None))
268
320
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COMPLETION_TOKENS,
269
- response.usage.completion_tokens)
321
+ response_dict.get('usage', {}).get('completion_tokens', None))
270
322
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
271
- response.usage.total_tokens)
323
+ response_dict.get('usage', {}).get('total_tokens', None))
272
324
  span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON,
273
- [response.choices[0].finish_reason])
325
+ [response_dict.get('choices', [])[0].get('finish_reason', None)])
274
326
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
275
327
  cost)
276
328
 
@@ -280,7 +332,7 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
280
332
  span.add_event(
281
333
  name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
282
334
  attributes={
283
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.choices[0].message.content,
335
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response_dict.get('choices', [])[0].get("message").get("content"),
284
336
  },
285
337
  )
286
338
 
@@ -292,7 +344,7 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
292
344
  span.add_event(
293
345
  name=attribute_name,
294
346
  attributes={
295
- SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response.choices[i].message.content,
347
+ SemanticConvetion.GEN_AI_CONTENT_COMPLETION: response_dict.get('choices')[i].get("message").get("content"),
296
348
  },
297
349
  )
298
350
  i += 1
@@ -304,8 +356,8 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
304
356
  elif "tools" in kwargs:
305
357
  # Calculate cost of the operation
306
358
  cost = get_chat_model_cost(kwargs.get("model", "gpt-3.5-turbo"),
307
- pricing_info, response.usage.prompt_tokens,
308
- response.usage.completion_tokens)
359
+ pricing_info, response_dict.get('usage').get('prompt_tokens'),
360
+ response_dict.get('usage').get('completion_tokens'))
309
361
  span.add_event(
310
362
  name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
311
363
  attributes={
@@ -313,11 +365,11 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
313
365
  },
314
366
  )
315
367
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
316
- response.usage.prompt_tokens)
368
+ response_dict.get('usage').get('prompt_tokens'))
317
369
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COMPLETION_TOKENS,
318
- response.usage.completion_tokens)
370
+ response_dict.get('usage').get('completion_tokens'))
319
371
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_TOTAL_TOKENS,
320
- response.usage.total_tokens)
372
+ response_dict.get('usage').get('total_tokens'))
321
373
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST,
322
374
  cost)
323
375
 
@@ -340,9 +392,9 @@ def chat_completions(gen_ai_endpoint, version, environment, application_name,
340
392
  }
341
393
 
342
394
  metrics["genai_requests"].add(1, attributes)
343
- metrics["genai_total_tokens"].add(response.usage.total_tokens, attributes)
344
- metrics["genai_completion_tokens"].add(response.usage.completion_tokens, attributes)
345
- metrics["genai_prompt_tokens"].add(response.usage.prompt_tokens, attributes)
395
+ metrics["genai_total_tokens"].add(response_dict.get('usage').get('total_tokens'), attributes)
396
+ metrics["genai_completion_tokens"].add(response_dict.get('usage').get('completion_tokens'), attributes)
397
+ metrics["genai_prompt_tokens"].add(response_dict.get('usage').get('prompt_tokens'), attributes)
346
398
  metrics["genai_cost"].record(cost, attributes)
347
399
 
348
400
  # Return original response
@@ -416,8 +468,8 @@ def embedding(gen_ai_endpoint, version, environment, application_name,
416
468
  kwargs.get("model", "text-embedding-ada-002"))
417
469
  span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_EMBEDDING_FORMAT,
418
470
  kwargs.get("encoding_format", "float"))
419
- span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_EMBEDDING_DIMENSION,
420
- kwargs.get("dimensions", ""))
471
+ # span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_EMBEDDING_DIMENSION,
472
+ # kwargs.get("dimensions", "null"))
421
473
  span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_USER,
422
474
  kwargs.get("user", ""))
423
475
  span.set_attribute(SemanticConvetion.GEN_AI_USAGE_PROMPT_TOKENS,
openlit/otel/metrics.py CHANGED
@@ -1,4 +1,4 @@
1
- # pylint: disable=duplicate-code, line-too-long
1
+ # pylint: disable=duplicate-code, line-too-long, ungrouped-imports
2
2
  """
3
3
  Setups up OpenTelemetry Meter
4
4
  """
@@ -8,10 +8,13 @@ from opentelemetry.sdk.metrics import MeterProvider
8
8
  from opentelemetry.sdk.metrics.export import PeriodicExportingMetricReader, ConsoleMetricExporter
9
9
  from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
10
10
  from opentelemetry.sdk.resources import Resource
11
- from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
12
-
13
11
  from openlit.semcov import SemanticConvetion
14
12
 
13
+ if os.environ.get("OTEL_EXPORTER_OTLP_PROTOCOL") == "grpc":
14
+ from opentelemetry.exporter.otlp.proto.grpc.metric_exporter import OTLPMetricExporter
15
+ else:
16
+ from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
17
+
15
18
  # Global flag to check if the meter provider initialization is complete.
16
19
  METER_SET = False
17
20
 
openlit/otel/tracing.py CHANGED
@@ -1,4 +1,4 @@
1
- # pylint: disable=duplicate-code, line-too-long
1
+ # pylint: disable=duplicate-code, line-too-long, ungrouped-imports
2
2
  """
3
3
  Setups up OpenTelemetry tracer
4
4
  """
@@ -10,8 +10,11 @@ from opentelemetry.sdk.resources import Resource
10
10
  from opentelemetry.sdk.trace import TracerProvider
11
11
  from opentelemetry.sdk.trace.export import BatchSpanProcessor, SimpleSpanProcessor
12
12
  from opentelemetry.sdk.trace.export import ConsoleSpanExporter
13
- from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
14
13
 
14
+ if os.environ.get("OTEL_EXPORTER_OTLP_PROTOCOL") == "grpc":
15
+ from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
16
+ else:
17
+ from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
15
18
 
16
19
  # Global flag to check if the tracer provider initialization is complete.
17
20
  TRACER_SET = False
openlit/semcov/__init__.py CHANGED
@@ -88,6 +88,7 @@ class SemanticConvetion:
88
88
  GEN_AI_TYPE_FINETUNING = "fine_tuning"
89
89
  GEN_AI_TYPE_VECTORDB = "vectordb"
90
90
  GEN_AI_TYPE_FRAMEWORK = "framework"
91
+ GEN_AI_TYPE_AGENT = "agent"
91
92
 
92
93
  GEN_AI_SYSTEM_HUGGING_FACE = "huggingface"
93
94
  GEN_AI_SYSTEM_OPENAI = "openai"
@@ -108,6 +109,8 @@ class SemanticConvetion:
108
109
  GEN_AI_SYSTEM_LLAMAINDEX = "llama_index"
109
110
  GEN_AI_SYSTEM_HAYSTACK = "haystack"
110
111
  GEN_AI_SYSTEM_EMBEDCHAIN = "embedchain"
112
+ GEN_AI_SYSTEM_LITELLM = "litellm"
113
+ GEN_AI_SYSTEM_CREWAI = "crewai"
111
114
 
112
115
  # Vector DB
113
116
  DB_REQUESTS = "db.total.requests"
@@ -154,6 +157,24 @@ class SemanticConvetion:
154
157
  DB_SYSTEM_QDRANT = "qdrant"
155
158
  DB_SYSTEM_MILVUS = "milvus"
156
159
 
160
+ # Agents
161
+ GEN_AI_AGENT_ID = "gen_ai.agent.id"
162
+ GEN_AI_AGENT_TASK_ID = "gen_ai.agent.task.id"
163
+ GEN_AI_AGENT_ROLE = "gen_ai.agent.role"
164
+ GEN_AI_AGENT_GOAL = "gen_ai.agent.goal"
165
+ GEN_AI_AGENT_CONTEXT = "gen_ai.agent.context"
166
+ GEN_AI_AGENT_ENABLE_CACHE = "gen_ai.agent.enable_cache"
167
+ GEN_AI_AGENT_ALLOW_DELEGATION = "gen_ai.agent.allow_delegation"
168
+ GEN_AI_AGENT_ALLOW_CODE_EXECUTION = "gen_ai.agent.allow_code_execution"
169
+ GEN_AI_AGENT_MAX_RETRY_LIMIT = "gen_ai.agent.max_retry_limit"
170
+ GEN_AI_AGENT_TOOLS = "gen_ai.agent.tools"
171
+ GEN_AI_AGENT_TOOL_RESULTS = "gen_ai.agent.tool_results"
172
+ GEN_AI_AGENT_TASK = "gen_ai.agent.task"
173
+ GEN_AI_AGENT_EXPECTED_OUTPUT = "gen_ai.agent.expected_output"
174
+ GEN_AI_AGENT_ACTUAL_OUTPUT = "gen_ai.agent.actual_output"
175
+ GEN_AI_AGENT_HUMAN_INPUT = "gen_ai.agent.human_input"
176
+ GEN_AI_AGENT_TASK_ASSOCIATION = "gen_ai.agent.task_associations"
177
+
157
178
  # GPU
158
179
  GPU_INDEX = "gpu.index"
159
180
  GPU_UUID = "gpu.uuid"
{openlit-1.29.1.dist-info → openlit-1.30.2.dist-info}/METADATA RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.1
2
2
  Name: openlit
3
- Version: 1.29.1
3
+ Version: 1.30.2
4
4
  Summary: OpenTelemetry-native Auto instrumentation library for monitoring LLM Applications and GPUs, facilitating the integration of observability into your GenAI-driven projects
5
5
  Home-page: https://github.com/openlit/openlit/tree/main/openlit/python
6
6
  Keywords: OpenTelemetry,otel,otlp,llm,tracing,openai,anthropic,claude,cohere,llm monitoring,observability,monitoring,gpt,Generative AI,chatGPT,gpu
@@ -63,16 +63,16 @@ This project proudly follows and maintains the [Semantic Conventions](https://gi
63
63
 
64
64
  ## Auto Instrumentation Capabilities
65
65
 
66
- | LLMs | Vector DBs | Frameworks | GPUs |
67
- |--------------------------------------------------------------------------|----------------------------------------------|----------------------------------------------|---------------|
68
- | [✅ OpenAI](https://docs.openlit.io/latest/integrations/openai) | [✅ ChromaDB](https://docs.openlit.io/latest/integrations/chromadb) | [✅ Langchain](https://docs.openlit.io/latest/integrations/langchain) | [✅ NVIDIA](https://docs.openlit.io/latest/integrations/nvidia-gpu) |
69
- | [✅ Ollama](https://docs.openlit.io/latest/integrations/ollama) | [✅ Pinecone](https://docs.openlit.io/latest/integrations/pinecone) | [✅ LiteLLM](https://docs.openlit.io/latest/integrations/litellm) | [✅ AMD](#) |
70
- | [✅ Anthropic](https://docs.openlit.io/latest/integrations/anthropic) | [✅ Qdrant](https://docs.openlit.io/latest/integrations/qdrant) | [✅ LlamaIndex](https://docs.openlit.io/latest/integrations/llama-index) | |
71
- | [✅ GPT4All](https://docs.openlit.io/latest/integrations/gpt4all) | [✅ Milvus](https://docs.openlit.io/latest/integrations/milvus) | [✅ Haystack](https://docs.openlit.io/latest/integrations/haystack) | |
72
- | [✅ Cohere](https://docs.openlit.io/latest/integrations/cohere) | | [✅ EmbedChain](https://docs.openlit.io/latest/integrations/embedchain) | |
73
- | [✅ Mistral](https://docs.openlit.io/latest/integrations/mistral) | | [✅ Guardrails](https://docs.openlit.io/latest/integrations/guardrails) | |
74
- | [✅ Azure OpenAI](https://docs.openlit.io/latest/integrations/azure-openai) | | | |
75
- | [✅ Azure AI Inference](https://docs.openlit.io/latest/integrations/azure-ai-inference) | | | |
66
+ | LLMs | Vector DBs | Frameworks | GPUs |
67
+ |--------------------------------------------------------------------------|----------------------------------------------|-------------------------------------------------|---------------|
68
+ | [✅ OpenAI](https://docs.openlit.io/latest/integrations/openai) | [✅ ChromaDB](https://docs.openlit.io/latest/integrations/chromadb) | [✅ Langchain](https://docs.openlit.io/latest/integrations/langchain) | [✅ NVIDIA](https://docs.openlit.io/latest/integrations/nvidia-gpu) |
69
+ | [✅ Ollama](https://docs.openlit.io/latest/integrations/ollama) | [✅ Pinecone](https://docs.openlit.io/latest/integrations/pinecone) | [✅ LiteLLM](https://docs.openlit.io/latest/integrations/litellm) | [✅ AMD](#) |
70
+ | [✅ Anthropic](https://docs.openlit.io/latest/integrations/anthropic) | [✅ Qdrant](https://docs.openlit.io/latest/integrations/qdrant) | [✅ LlamaIndex](https://docs.openlit.io/latest/integrations/llama-index) | |
71
+ | [✅ GPT4All](https://docs.openlit.io/latest/integrations/gpt4all) | [✅ Milvus](https://docs.openlit.io/latest/integrations/milvus) | [✅ Haystack](https://docs.openlit.io/latest/integrations/haystack) | |
72
+ | [✅ Cohere](https://docs.openlit.io/latest/integrations/cohere) | | [✅ EmbedChain](https://docs.openlit.io/latest/integrations/embedchain) | |
73
+ | [✅ Mistral](https://docs.openlit.io/latest/integrations/mistral) | | [✅ Guardrails](https://docs.openlit.io/latest/integrations/guardrails) | |
74
+ | [✅ Azure OpenAI](https://docs.openlit.io/latest/integrations/azure-openai) | | [✅ CrewAI](https://docs.openlit.io/latest/integrations/crewai) | |
75
+ | [✅ Azure AI Inference](https://docs.openlit.io/latest/integrations/azure-ai-inference) | |
76
76
  | [✅ GitHub AI Models](https://docs.openlit.io/latest/integrations/github-models) | | | |
77
77
  | [✅ HuggingFace Transformers](https://docs.openlit.io/latest/integrations/huggingface) | | | |
78
78
  | [✅ Amazon Bedrock](https://docs.openlit.io/latest/integrations/bedrock) | | | |