AI_EMBEDDING

Uses a supported embedding model to convert text into high-dimensional vector representations. The generated vectors can be used for downstream tasks such as semantic search, text similarity computation, cluster analysis, and recommendation systems.

Syntax

AI_EMBEDDING(<model>, <input> [ , <model_parameters> ])

This function contains 2 required parameters and 1 optional parameter, supporting both positional and named parameter syntax.

Parameter Description

Required Parameters

model

Specifies the embedding model to invoke. Consistent with AI_COMPLETE, models come from two sources: API Gateway Endpoints and API Connection objects.

When calling via an API Gateway endpoint, use the endpoint prefix followed by the endpoint name:

'endpoint:<endpoint_name>'

When calling via an API Connection object, first create the connection using CREATE API CONNECTION, then reference it with the connection/ prefix:

'connection:<connection_object_name>'

input

The input text to be converted into a vector. This can be a single word, a sentence, a paragraph, or a value from a table column. The maximum length of the input text is limited by the context window of the selected model; exceeding the limit may cause text truncation or call failure.

Optional Parameters

model_parameters

Model hyperparameters passed in as a JSON object, used to control the behavior of the embedding model. Supported parameters may vary by model. Common parameters are as follows:

Parameter Example:

JSON '{
    "input": "text",
    "dimensions": "1024"
}'

Examples

SELECT AI_EMBEDDING(
    'endpoint:lis_openai_embedding',
    'What is the capital of China?',
    JSON '{
        "input": "text",
        "dimensions": "1024"
    }'
);