To achieve semantic text search using the imported model with this feature, specify the imported model when defining the vectorization You can use the imported model for vectorization by specifying pgx_vectorizer.pgx_embedding_onnx in the embedding argument of pgx_vectorizer.pgx_create_vectorizer. Specify the model name and the number of dimensions of the generated vector in pgx_vectorizer.pgx_embedding_onnx.

If an unavailable model is specified when executing pgx_create_vectorizer, a warning message will be output. Follow the warning message to load the specified model.