1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
|
import itertools
import pytest
from huggingface_hub import (
InferenceClient,
)
from huggingface_hub.errors import OverloadedError
from sentry_sdk import start_transaction
from sentry_sdk.integrations.huggingface_hub import HuggingfaceHubIntegration
from unittest import mock # python 3.3 and above
@pytest.mark.parametrize(
"send_default_pii, include_prompts, details_arg",
itertools.product([True, False], repeat=3),
)
def test_nonstreaming_chat_completion(
sentry_init, capture_events, send_default_pii, include_prompts, details_arg
):
sentry_init(
integrations=[HuggingfaceHubIntegration(include_prompts=include_prompts)],
traces_sample_rate=1.0,
send_default_pii=send_default_pii,
)
events = capture_events()
client = InferenceClient("some-model")
if details_arg:
client.post = mock.Mock(
return_value=b"""[{
"generated_text": "the model response",
"details": {
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [],
"tokens": []
}
}]"""
)
else:
client.post = mock.Mock(
return_value=b'[{"generated_text": "the model response"}]'
)
with start_transaction(name="huggingface_hub tx"):
response = client.text_generation(
prompt="hello",
details=details_arg,
stream=False,
)
if details_arg:
assert response.generated_text == "the model response"
else:
assert response == "the model response"
tx = events[0]
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "ai.chat_completions.create.huggingface_hub"
if send_default_pii and include_prompts:
assert "hello" in span["data"]["ai.input_messages"]
assert "the model response" in span["data"]["ai.responses"]
else:
assert "ai.input_messages" not in span["data"]
assert "ai.responses" not in span["data"]
if details_arg:
assert span["measurements"]["ai_total_tokens_used"]["value"] == 10
@pytest.mark.parametrize(
"send_default_pii, include_prompts, details_arg",
itertools.product([True, False], repeat=3),
)
def test_streaming_chat_completion(
sentry_init, capture_events, send_default_pii, include_prompts, details_arg
):
sentry_init(
integrations=[HuggingfaceHubIntegration(include_prompts=include_prompts)],
traces_sample_rate=1.0,
send_default_pii=send_default_pii,
)
events = capture_events()
client = InferenceClient("some-model")
client.post = mock.Mock(
return_value=[
b"""data:{
"token":{"id":1, "special": false, "text": "the model "}
}""",
b"""data:{
"token":{"id":2, "special": false, "text": "response"},
"details":{"finish_reason": "length", "generated_tokens": 10, "seed": 0}
}""",
]
)
with start_transaction(name="huggingface_hub tx"):
response = list(
client.text_generation(
prompt="hello",
details=details_arg,
stream=True,
)
)
assert len(response) == 2
print(response)
if details_arg:
assert response[0].token.text + response[1].token.text == "the model response"
else:
assert response[0] + response[1] == "the model response"
tx = events[0]
assert tx["type"] == "transaction"
span = tx["spans"][0]
assert span["op"] == "ai.chat_completions.create.huggingface_hub"
if send_default_pii and include_prompts:
assert "hello" in span["data"]["ai.input_messages"]
assert "the model response" in span["data"]["ai.responses"]
else:
assert "ai.input_messages" not in span["data"]
assert "ai.responses" not in span["data"]
if details_arg:
assert span["measurements"]["ai_total_tokens_used"]["value"] == 10
def test_bad_chat_completion(sentry_init, capture_events):
sentry_init(integrations=[HuggingfaceHubIntegration()], traces_sample_rate=1.0)
events = capture_events()
client = InferenceClient("some-model")
client.post = mock.Mock(side_effect=OverloadedError("The server is overloaded"))
with pytest.raises(OverloadedError):
client.text_generation(prompt="hello")
(event,) = events
assert event["level"] == "error"
def test_span_origin(sentry_init, capture_events):
sentry_init(
integrations=[HuggingfaceHubIntegration()],
traces_sample_rate=1.0,
)
events = capture_events()
client = InferenceClient("some-model")
client.post = mock.Mock(
return_value=[
b"""data:{
"token":{"id":1, "special": false, "text": "the model "}
}""",
]
)
with start_transaction(name="huggingface_hub tx"):
list(
client.text_generation(
prompt="hello",
stream=True,
)
)
(event,) = events
assert event["contexts"]["trace"]["origin"] == "manual"
assert event["spans"][0]["origin"] == "auto.ai.huggingface_hub"
|
