I’m using Azure suite, particularly Azure OpenAi, but I want to understand how GPT behaves with the connection to the SQL database. When you connect through Azure Search these 2 services (openAI and SQL), how does GPT use the information in the table in terms of tokens/chunks? How much would it be cost to get insights from a sql table of millions rows and hundreds columns, using gpt 4? I want to understand if they consider as token just one value of the table converted into a JSON