Is there a method to prioritize certain text arguments when creating embeddings with the (ADA 002) OpenAI model?

Question

I am developing the search API for a website where users can search for any car. My goal is to make it easier for them to find exactly what they are looking for by providing accurate and relevant results.

I am creating embeddings using the input string, which includes various parameters such as make, model, exterior, etc. There are specific priority arguments that users can use to refine their search.

In the current setup, when a user searches for "Honda Civic 2023 black FWD", my API will suggest cars that match "2023", "black", "FWD", but it will prioritize the initial query "Honda Civic" and then narrow down the results based on the additional details.

I use pg_vector, the PostgreSQL extension, to store embedding vectors and the (<=>) cosine similarity operator for matching the embeddings.

I aimed to create an intelligent search function where users can input their thoughts, and my system can recommend the precise results.