Skip to content

Embedding Models API

EmbeddingModel

Bases: SynalinksSaveable

An embedding model API wrapper.

Embedding models are a type of machine learning model used to convert high-dimensional data, such as text into lower-dimensional vector representations while preserving the semantic meaning and relationships within the data. These vector representations, known as embeddings, allow for more efficient and effective processing in various tasks.

Many providers are available like OpenAI, Azure OpenAI, Vertex AI or Ollama.

For the complete list of models, please refer to the providers documentation.

Using OpenAI models

import synalinks
import os

os.environ["OPENAI_API_KEY"] = "your-api-key"

embedding_model = synalinks.EmbeddingModel(
    model="openai/text-embedding-ada-002",
)

Using Azure OpenAI models

import synalinks
import os

os.environ["AZURE_API_KEY"] = "your-api-key"
os.environ["AZURE_API_BASE"] = "your-api-base"
os.environ["AZURE_API_VERSION"] = "your-api-version"

embedding_model = synalinks.EmbeddingModel(
    model="azure/<your_deployment_name>",
)

Using VertexAI models

import synalinks
import os

embedding_model = synalinks.EmbeddingModel(
    model="vertex_ai/text-embedding-004",
    vertex_project = "hardy-device-38811", # Your Project ID
    vertex_location = "us-central1",  # Project location
)

Using Ollama models

import synalinks

embedding_model = synalinks.EmbeddingModel(
    model="ollama/mxbai-embed-large",
)

Note: Obviously, use an .env file and .gitignore to avoid putting your API keys in the code or a config file that can lead to leackage when pushing it into repositories.

Parameters:

Name Type Description Default
model str

The model to use.

 None
api_base str

Optional. The endpoint to use.

 None
retry int

Optional. The number of retry.

 5
fallback EmbeddingModel

Optional. The embedding model to fallback if anything is wrong.

 None
Source code in synalinks/src/embedding_models/embedding_model.py 
 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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
@synalinks_export(
    ["synalinks.EmbeddingModel", "synalinks.embedding_models.EmbeddingModel"]
)
class EmbeddingModel(SynalinksSaveable):
    """An embedding model API wrapper.

    Embedding models are a type of machine learning model used to convert
    high-dimensional data, such as text into lower-dimensional vector
    representations while preserving the semantic meaning and relationships
    within the data. These vector representations, known as embeddings,
    allow for more efficient and effective processing in various tasks.

    Many providers are available like OpenAI, Azure OpenAI, Vertex AI or Ollama.

    For the complete list of models, please refer to the providers documentation.

    **Using OpenAI models**

    ```python
    import synalinks
    import os

    os.environ["OPENAI_API_KEY"] = "your-api-key"

    embedding_model = synalinks.EmbeddingModel(
        model="openai/text-embedding-ada-002",
    )
    ```

    **Using Azure OpenAI models**

    ```python
    import synalinks
    import os

    os.environ["AZURE_API_KEY"] = "your-api-key"
    os.environ["AZURE_API_BASE"] = "your-api-base"
    os.environ["AZURE_API_VERSION"] = "your-api-version"

    embedding_model = synalinks.EmbeddingModel(
        model="azure/<your_deployment_name>",
    )
    ```

    **Using VertexAI models**

    ```python
    import synalinks
    import os

    embedding_model = synalinks.EmbeddingModel(
        model="vertex_ai/text-embedding-004",
        vertex_project = "hardy-device-38811", # Your Project ID
        vertex_location = "us-central1",  # Project location
    )
    ```

    **Using Ollama models**

    ```python
    import synalinks

    embedding_model = synalinks.EmbeddingModel(
        model="ollama/mxbai-embed-large",
    )
    ```

    **Note**: Obviously, use an `.env` file and `.gitignore` to avoid
    putting your API keys in the code or a config file that can lead to
    leackage when pushing it into repositories.

    Args:
        model (str): The model to use.
        api_base (str): Optional. The endpoint to use.
        retry (int): Optional. The number of retry.
        fallback (EmbeddingModel): Optional. The embedding model to fallback
            if anything is wrong.
    """

    def __init__(
        self,
        model=None,
        api_base=None,
        retry=5,
        fallback=None,
    ):
        if model is None:
            raise ValueError(
                "You need to set the `model` argument for any EmbeddingModel"
            )
        self.model = model
        if self.model.startswith("ollama") and not api_base:
            self.api_base = "http://localhost:11434"
        else:
            self.api_base = api_base
        self.retry = retry
        self.fallback = fallback

    async def __call__(self, texts, **kwargs):
        """
        Call method to get dense embeddings vectors

        Args:
            texts (list): A list of texts to embed.

        Returns:
            (list): The list of corresponding vectors.
        """
        maybe_initialize_telemetry()

        for i in range(self.retry):
            try:
                if self.api_base:
                    response = litellm.embedding(
                        model=self.model,
                        input=texts,
                        api_base=self.api_base,
                        **kwargs,
                    )
                else:
                    response = litellm.embedding(
                        model=self.model,
                        input=texts,
                        **kwargs,
                    )
                vectors = []
                for data in response["data"]:
                    vectors.append(data["embedding"])
                return {"embeddings": vectors}
            except Exception as e:
                warnings.warn(f"Error occured while trying to call {self}: " + str(e))
                capture_exception(e)
        if self.fallback:
            return self.fallback(
                texts,
                **kwargs,
            )
        else:
            return None

    def _obj_type(self):
        return "EmbeddingModel"

    def get_config(self):
        config = {
            "model": self.model,
            "api_base": self.api_base,
            "retry": self.retry,
        }
        if self.fallback:
            fallback_config = {
                "fallback": serialization_lib.serialize_synalinks_object(
                    self.fallback,
                )
            }
            return {**fallback_config, **config}
        else:
            return config

    @classmethod
    def from_config(cls, config):
        if "fallback" in config:
            fallback = serialization_lib.deserialize_synalinks_object(
                config.pop("fallback")
            )
            return cls(fallback=fallback, **config)
        else:
            return cls(**config)

    def __repr__(self):
        api_base = f" api_base={self.api_base}" if self.api_base else ""
        return f"<EmbeddingModel model={self.model}{api_base}>"

__call__(texts, **kwargs) async

Call method to get dense embeddings vectors

Parameters:

Name Type Description Default
texts list

A list of texts to embed.

 required

Returns:

Type Description
list

The list of corresponding vectors.
Source code in synalinks/src/embedding_models/embedding_model.py 
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
async def __call__(self, texts, **kwargs):
    """
    Call method to get dense embeddings vectors

    Args:
        texts (list): A list of texts to embed.

    Returns:
        (list): The list of corresponding vectors.
    """
    maybe_initialize_telemetry()

    for i in range(self.retry):
        try:
            if self.api_base:
                response = litellm.embedding(
                    model=self.model,
                    input=texts,
                    api_base=self.api_base,
                    **kwargs,
                )
            else:
                response = litellm.embedding(
                    model=self.model,
                    input=texts,
                    **kwargs,
                )
            vectors = []
            for data in response["data"]:
                vectors.append(data["embedding"])
            return {"embeddings": vectors}
        except Exception as e:
            warnings.warn(f"Error occured while trying to call {self}: " + str(e))
            capture_exception(e)
    if self.fallback:
        return self.fallback(
            texts,
            **kwargs,
        )
    else:
        return None