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
|
import functools
import pytest
from devtools_testutils import AzureRecordedTestCase, EnvironmentVariableLoader, recorded_by_proxy
from azure.ai.language.conversations import ConversationAnalysisClient, AnalyzeConversationLROPoller
from azure.core.paging import ItemPaged
from azure.ai.language.conversations.models import (
# request models
AnalyzeConversationOperationInput,
MultiLanguageConversationInput,
TextConversation,
TextConversationItem,
PiiOperationAction,
ConversationPiiActionContent,
ConversationActions,
AnalyzeConversationOperationResult,
ConversationPiiOperationResult,
ConversationalPiiResult,
ConversationPiiItemResult,
NamedEntity,
InputWarning,
ConversationError,
AnalyzeConversationOperationAction,
CharacterMaskPolicyType,
RedactionCharacter,
)
from typing import cast, List
from azure.core.credentials import AzureKeyCredential
ConversationsPreparer = functools.partial(
EnvironmentVariableLoader,
"conversations",
conversations_endpoint="https://Sanitized.cognitiveservices.azure.com/",
conversations_key="fake_key",
)
class TestConversations(AzureRecordedTestCase):
# Start with any helper functions you might need, for example a client creation method:
def create_client(self, endpoint, key):
credential = AzureKeyCredential(key)
client = ConversationAnalysisClient(endpoint, credential)
return client
...
class TestConversationsCase(TestConversations):
@ConversationsPreparer()
@recorded_by_proxy
def test_conversation_pii_with_character_mask_policy(self, conversations_endpoint, conversations_key):
client = self.create_client(conversations_endpoint, conversations_key)
# Track redacted texts we verify
redacted_verified: List[str] = []
# ---- Redaction policy: mask with '*' ---------------------------------
redaction_policy = CharacterMaskPolicyType(redaction_character=RedactionCharacter.ASTERISK)
# ---- Build input -----------------------------------------------------
ml_input = MultiLanguageConversationInput(
conversations=[
TextConversation(
id="1",
language="en",
conversation_items=[
TextConversationItem(id="1", participant_id="Agent_1", text="Can you provide your name?"),
TextConversationItem(id="2", participant_id="Customer_1", text="Hi, my name is John Doe."),
TextConversationItem(
id="3",
participant_id="Agent_1",
text="Thank you John, that has been updated in our system.",
),
],
)
]
)
# Action with CharacterMaskPolicyType
pii_action: AnalyzeConversationOperationAction = PiiOperationAction(
action_content=ConversationPiiActionContent(redaction_policy=redaction_policy),
name="Conversation PII with Character Mask Policy",
)
actions: List[AnalyzeConversationOperationAction] = [pii_action]
operation_input = AnalyzeConversationOperationInput(
conversation_input=ml_input,
actions=actions,
)
# ---- Begin LRO -------------------------------------------------------
poller: AnalyzeConversationLROPoller[ItemPaged[ConversationActions]] = client.begin_analyze_conversation_job(
body=operation_input
)
print(f"Operation ID: {poller.details.get('operation_id')}")
paged_actions: ItemPaged[ConversationActions] = poller.result()
d = poller.details
print(f"Job ID: {d.get('job_id')}")
print(f"Status: {d.get('status')}")
# ---- Iterate results and verify redaction ----------------------------
for actions_page in paged_actions:
for action_result in actions_page.task_results or []:
ar = cast(AnalyzeConversationOperationResult, action_result)
if isinstance(ar, ConversationPiiOperationResult):
for conversation in ar.results.conversations or []:
conversation = cast(ConversationalPiiResult, conversation)
for item in conversation.conversation_items or []:
item = cast(ConversationPiiItemResult, item)
redacted_text = (getattr(item.redacted_content, "text", None) or "").strip()
if not redacted_text:
continue
# Only verify when there are detected entities in the original item
if item.entities:
# Ensure original PII text is NOT present and '*' is present
for entity in item.entities:
ent_text = cast(NamedEntity, entity).text or ""
assert (
ent_text not in redacted_text
), f"Expected entity '{ent_text}' to be redacted but found in: {redacted_text}"
assert (
"*" in redacted_text
), f"Expected redacted text to contain '*', got: {redacted_text}"
redacted_verified.append(redacted_text)
# ---- Assertions -------------------------------------------------------
assert (d.get("status") or "").lower() in {"succeeded", "partiallysucceeded"}
assert len(redacted_verified) > 0, "Expected at least one redacted line to be verified."
|
