This function calls a general text embedding model to convert text to embeddings.
Only PolarDB for PostgreSQL Standard Edition clusters in the China (Beijing) region support calling built-in models.
Syntax
real[] AI_Text_Embedding(text content);
real[] AI_Text_Embedding(text content, text model_id);Parameters
Parameter Name | Description |
content | The text to convert to an embedding. The text can be up to 2048 characters long. |
model_id | The ID of the model. If you do not specify this parameter, the built-in model |
Return value
Returns a vector of the real[] type.
Description
You can use this function directly without creating a model. By default, the function calls the built-in model
_dashscope/text_embedding/text_embedding_v2to convert text to an embedding.You can specify the model_id parameter to call other existing text embedding models, such as
_dashscope/text_embedding/text_embedding_v3.For more information about the built-in models
_dashscope/text_embedding/text_embedding_v2and_dashscope/text_embedding/text_embedding_v3, see Text and multimodal embeddings.
Preparations
The built-in models are integrated with the Natural Language Processing models provided by Alibaba Cloud Model Studio.
Set the model token: You can use the AI_SetModelToken function to attach your Alibaba Cloud Model Studio API key.
Modify the model call address: You can use the AI_AlterModel function to modify the call address of the built-in model. Replace the original address with the PrivateLink endpoint of Alibaba Cloud Model Studio. Change
https://dashscope.aliyuncs.comtohttp://vpc-cn-beijing.dashscope.aliyuncs.com.For example:
_dashscope/text_embedding/text_embedding_v2:SELECT polar_ai.AI_AlterModel('_dashscope/text_embedding/text_embedding_v2', model_url=>'http://vpc-cn-beijing.dashscope.aliyuncs.com/api/v1/services/embeddings/text-embedding/text-embedding');_dashscope/text_embedding/text_embedding_v3:SELECT polar_ai.AI_AlterModel('_dashscope/text_embedding/text_embedding_v3', model_url=>'http://vpc-cn-beijing.dashscope.aliyuncs.com/api/v1/services/embeddings/text-embedding/text-embedding');
Examples
In this example, the model_id parameter is not specified.
SELECT polar_ai.ai_text_embedding('The wind is sharp, the sky is high, apes shriek sadly.');The following result is returned:
--- {0.004930191827757042, -0.012394876394385515, 0.041976027360927766, -0.01943111099801429, 0.03970705290898573, -0.03082425930989348, ..., -0.008285377453501126, 0.001457333324851072, -0.01758455200255082, -0.03808980516132491, -0.03719669819619879, -0.008629344325205105 }In this example, the model_id parameter is set to the built-in model
_dashscope/text_embedding/text_embedding_v3.SELECT polar_ai.ai_text_embedding('The wind is swift and the sky is high.', '_dashscope/text_embedding/text_embedding_v3');The following result is returned:
--- {0.004930191827757042, -0.012394876394385515, 0.041976027360927766, -0.01943111099801429, 0.03970705290898573, -0.03082425930989348, ..., -0.008285377453501126, 0.001457333324851072, -0.01758455200255082, -0.03808980516132491, -0.03719669819619879, -0.008629344325205105 }