Invokes a registered model to run machine learning and natural language processing (NLP) tasks, such as sentiment analysis, question answering, translation, summarization, and text generation.
Syntax
jsonb AI_CallModel(text model_id, anyelement content);Parameters
| Parameter | Type | Description |
|---|---|---|
model_id | text | The unique name of the model to invoke. |
content | anyelement | The input data for the model. Supported data types include text, images, complex data structures, and records. Must match the data type specified in the model_in_transform_fn function used when creating the model. |
Return value
Returns a JSONB (JSON binary) object. Parse the JSONB object to extract results such as text or image embedding vectors.
Usage notes
Before calling
AI_CallModel, create a model using AI_CreateModel.The
contentparameter type must match the type defined in the function of the target model. For text input, cast the value explicitly:content::text.
Example
Convert text into embedding vectors using a registered model:
SELECT polar_ai.AI_CallModel('my_text_embedding_model', '风急天高猿啸哀'::text);Sample output:
[1.4284451007843018, -2.7454426288604736, 1.5957790613174438, -2.182790756225586, 3.9757323265075684, -3.0645480155944824, 3.613696336746216, -3.484774351119995, -2.174361228942871, 1.0907986164093018, 1.4526909589767456, 1.3686197996139526, 1.4130642414093018, -1.694422721862793, 2.053454637527466, 4.982985973358154, 1.7921278476715088, 2.6027371883392334, 0.1595267355442047, 1.7544844150543213, 2.0501770973205566]