0
# Embeddings
1

2
Text embedding generation for converting text into numerical vectors for semantic similarity, search, and machine learning applications. Supports multiple providers through Portkey's unified interface.
3

4
## Capabilities
5

6
### Embedding Creation
7

8
Creates vector embeddings from input text using the specified model. Supports various embedding models from different providers with consistent interface.
9

10
```python { .api }
11
class Embeddings:
12
    def create(
13
        self,
14
        *,
15
        input: str,
16
        model: Optional[str] = "portkey-default",
17
        dimensions: Union[int, NotGiven] = NOT_GIVEN,
18
        encoding_format: Union[str, NotGiven] = NOT_GIVEN,
19
        user: Union[str, NotGiven] = NOT_GIVEN,
20
        **kwargs
21
    ) -> CreateEmbeddingResponse:
22
        """
23
        Create embeddings for the given input text.
24

25
        Args:
26
            input: Input text to create embeddings for
27
            model: Embedding model to use (defaults to "portkey-default")
28
            dimensions: Number of dimensions for the embedding vectors (model-dependent)
29
            encoding_format: Format for encoding the embeddings (e.g., "float", "base64")
30
            user: User identifier for tracking and analytics
31
            **kwargs: Additional provider-specific parameters
32

33
        Returns:
34
            CreateEmbeddingResponse: Response containing embedding vectors and metadata
35
        """
36

37
class AsyncEmbeddings:
38
    async def create(
39
        self,
40
        *,
41
        input: str,
42
        model: Optional[str] = "portkey-default",  
43
        dimensions: Union[int, NotGiven] = NOT_GIVEN,
44
        encoding_format: Union[str, NotGiven] = NOT_GIVEN,
45
        user: Union[str, NotGiven] = NOT_GIVEN,
46
        **kwargs
47
    ) -> CreateEmbeddingResponse:
48
        """Async version of create method."""
49
```
50

51
### Usage Examples
52

53
```python
54
from portkey_ai import Portkey
55

56
# Initialize client
57
portkey = Portkey(
58
    api_key="PORTKEY_API_KEY",
59
    virtual_key="VIRTUAL_KEY"  
60
)
61

62
# Create embeddings for single text
63
response = portkey.embeddings.create(
64
    input="Hello, world!",
65
    model="text-embedding-ada-002"
66
)
67

68
# Access the embedding vector
69
embedding = response.data[0].embedding
70
print(f"Embedding dimensions: {len(embedding)}")
71

72
# Create embeddings with specific dimensions
73
response = portkey.embeddings.create(
74
    input="Text to embed",
75
    model="text-embedding-3-small",
76
    dimensions=512
77
)
78

79
# Batch embedding creation (provider-dependent)
80
texts = [
81
    "First text to embed",
82
    "Second text to embed", 
83
    "Third text to embed"
84
]
85

86
for text in texts:
87
    response = portkey.embeddings.create(
88
        input=text,
89
        model="text-embedding-ada-002",
90
        user="user-123"
91
    )
92
    print(f"Embedding for '{text}': {len(response.data[0].embedding)} dimensions")
93
```
94

95
### Async Usage
96

97
```python
98
import asyncio
99
from portkey_ai import AsyncPortkey
100

101
async def create_embeddings():
102
    portkey = AsyncPortkey(
103
        api_key="PORTKEY_API_KEY",
104
        virtual_key="VIRTUAL_KEY"
105
    )
106
    
107
    response = await portkey.embeddings.create(
108
        input="Async embedding creation",
109
        model="text-embedding-ada-002"
110
    )
111
    
112
    return response.data[0].embedding
113

114
# Run async function
115
embedding = asyncio.run(create_embeddings())
116
```
117

118
## Types
119

120
```python { .api }
121
class CreateEmbeddingResponse:
122
    """Response from embedding creation request"""
123
    object: str  # "list"
124
    data: List[EmbeddingObject]
125
    model: str
126
    usage: EmbeddingUsage
127
    _headers: Optional[dict]  # Response headers for debugging
128

129
class EmbeddingObject:
130
    """Individual embedding object"""
131
    object: str  # "embedding"
132
    embedding: List[float]  # The embedding vector
133
    index: int  # Index in the input list
134

135
class EmbeddingUsage:
136
    """Token usage information"""
137
    prompt_tokens: int
138
    total_tokens: int
139
```