Skip to content

Latest commit

 

History

History

T-SQL

README.md

Get Embeddings from OpenAI from Azure SQL

In this sample you'll be creating a stored procedure to easily transform text into a vector using OpenAI embedding model.

Create the Embedding Model

Make sure you can access OpenAI service by following the documentation here: How do I get access to Azure OpenAI?.

Deploy an embedding model - for example the text-embedding-3-small - following the Create and deploy an Azure OpenAI Service resource. Please note that the sample assumes that the choose embedding model returns a 1536-dimensional vector, so if you choose another embedding model, you may need to adjust the sample accordingly.

Then retrieve the Azure OpenAI endpoint and key:

Azure OpenAI Endpoint and Key

Store the credentials into Azure SQL

Connect to Azure SQL database and run the 01-store-openai-credentials.sql to store the Azure OpenAI endpoint and secret so that it can be used for later use.

Create the get_embeddings Stored Procedure

Use the 02-create-get-embeddings-procedure.sql to create a stored procedure that will call the OpenAI embedding model you have deployed before. The stored procedure uses the sp_invoke_external_rest_endpoint

Transform text into embedding

Use the 03-get-embeddings.sql to call to OpenAI to transform sample text into embeddings. Make sure to use the deployed model name:

Deployed OpenAI Models

And then use

declare @king vector(1536);
exec dbo.get_embedding @deployedModelName = 'text-embedding-3-small', @inputText = 'King', @embedding = @king output;

the resulting vector will be stored into the @king variable. You can now compare it with another variable to find how similar they are:

select
    vector_distance('cosine', @king, @queen) as 'King vs Queen'