0
# Azure AI Content Safety
1

2
A comprehensive Python client library for Azure AI Content Safety services that enables developers to detect harmful user-generated and AI-generated content in applications and services. It provides both text and image analysis APIs that classify content across four harm categories (sexual, violence, hate, and self-harm) with multi-severity levels, includes text blocklist management for screening custom terms, and supports both Azure Key Credential and Microsoft Entra ID authentication methods.
3

4
## Package Information
5

6
- **Package Name**: azure-ai-contentsafety
7
- **Package Type**: pypi
8
- **Language**: Python
9
- **Installation**: `pip install azure-ai-contentsafety`
10

11
## Core Imports
12

13
Synchronous clients:
14
```python
15
from azure.ai.contentsafety import ContentSafetyClient, BlocklistClient
16
```
17

18
Asynchronous clients:
19
```python
20
from azure.ai.contentsafety.aio import ContentSafetyClient, BlocklistClient
21
```
22

23
Authentication:
24
```python
25
from azure.core.credentials import AzureKeyCredential
26
from azure.ai.contentsafety import ContentSafetyClient
27

28
# Using API key authentication
29
client = ContentSafetyClient(
30
    endpoint="https://your-resource.cognitiveservices.azure.com",
31
    credential=AzureKeyCredential("your-api-key")
32
)
33
```
34

35
## Basic Usage
36

37
```python
38
from azure.ai.contentsafety import ContentSafetyClient
39
from azure.ai.contentsafety.models import AnalyzeTextOptions
40
from azure.core.credentials import AzureKeyCredential
41

42
# Initialize the client
43
client = ContentSafetyClient(
44
    endpoint="https://your-resource.cognitiveservices.azure.com",
45
    credential=AzureKeyCredential("your-api-key")
46
)
47

48
# Analyze text content
49
request = AnalyzeTextOptions(text="Some text to analyze")
50
response = client.analyze_text(request)
51

52
# Check results
53
for result in response.categories_analysis:
54
    print(f"Category: {result.category}, Severity: {result.severity}")
55

56
# Close the client
57
client.close()
58
```
59

60
## Architecture
61

62
The Azure AI Content Safety client library is organized around two main client classes:
63

64
- **ContentSafetyClient**: Handles content analysis for text and images, providing severity scores across four harm categories
65
- **BlocklistClient**: Manages custom text blocklists for screening domain-specific prohibited terms
66

67
Both clients support:
68
- **Synchronous and Asynchronous Operations**: Full async/await support with corresponding aio clients
69
- **Flexible Authentication**: Azure Key Credential or Microsoft Entra ID token-based authentication
70
- **Enterprise Features**: Built-in error handling, logging, retry policies, and request/response pipeline customization
71
- **Context Management**: Automatic resource cleanup with Python context managers
72

73
## Capabilities
74

75
### Content Analysis
76

77
Analyze text and image content for harmful material across four categories (hate, self-harm, sexual, violence) with configurable severity levels and custom blocklist integration.
78

79
```python { .api }
80
def analyze_text(self, options: AnalyzeTextOptions, **kwargs) -> AnalyzeTextResult: ...
81
def analyze_image(self, options: AnalyzeImageOptions, **kwargs) -> AnalyzeImageResult: ...
82
```
83

84
[Content Analysis](./content-analysis.md)
85

86
### Blocklist Management
87

88
Create and manage custom text blocklists to screen for domain-specific prohibited terms, with support for adding, updating, removing, and querying blocklist items.
89

90
```python { .api }
91
def create_or_update_text_blocklist(self, blocklist_name: str, options: TextBlocklist, **kwargs) -> TextBlocklist: ...
92
def add_or_update_blocklist_items(self, blocklist_name: str, options: AddOrUpdateTextBlocklistItemsOptions, **kwargs) -> AddOrUpdateTextBlocklistItemsResult: ...
93
def list_text_blocklists(self, **kwargs) -> Iterable[TextBlocklist]: ...
94
def delete_text_blocklist(self, blocklist_name: str, **kwargs) -> None: ...
95
```
96

97
[Blocklist Management](./blocklist-management.md)
98

99
## Core Types
100

101
```python { .api }
102
class ContentSafetyClient:
103
    def __init__(
104
        self,
105
        endpoint: str,
106
        credential: Union[AzureKeyCredential, TokenCredential],
107
        **kwargs
108
    ): ...
109
    def close(self) -> None: ...
110
    def __enter__(self) -> "ContentSafetyClient": ...
111
    def __exit__(self, *exc_details: Any) -> None: ...
112

113
class BlocklistClient:
114
    def __init__(
115
        self,
116
        endpoint: str,
117
        credential: Union[AzureKeyCredential, TokenCredential],
118
        **kwargs
119
    ): ...
120
    def close(self) -> None: ...
121
    def __enter__(self) -> "BlocklistClient": ...
122
    def __exit__(self, *exc_details: Any) -> None: ...
123

124
# Content categories
125
class TextCategory(str, Enum):
126
    HATE: str
127
    SELF_HARM: str
128
    SEXUAL: str
129
    VIOLENCE: str
130

131
class ImageCategory(str, Enum):
132
    HATE: str
133
    SELF_HARM: str
134
    SEXUAL: str
135
    VIOLENCE: str
136

137
# Output severity levels
138
class AnalyzeTextOutputType(str, Enum):
139
    FOUR_SEVERITY_LEVELS: str  # 0, 2, 4, 6
140
    EIGHT_SEVERITY_LEVELS: str  # 0, 1, 2, 3, 4, 5, 6, 7
141

142
class AnalyzeImageOutputType(str, Enum):
143
    FOUR_SEVERITY_LEVELS: str  # 0, 2, 4, 6
144
```