Course of Automatic Embeddings in Postgres

By Javier Ferrer y Rafa Gómez – Software Design and Architecture y Artificial Intelligence

Delegate embedding generation to the infrastructure.

⭐ Course in early access. Only available for Codely Pro Premium.

Currently, this course is only available if you have a Codely Pro Premium account. You can change yours to this modality from here by selecting Premium and using the same email with which you have an account on this platform.

Eventually, we will also publish it for users with standard access.

💸 First lesson available without registration 💸

If we want to make the most of the power of embeddings for semantic searches, recommendations, or AI systems, it is important to implement them efficiently and scalably.
But this is not always straightforward. Often we ask ourselves:

❓ Where do we generate them from?: Database or application?, synchronously or asynchronously?
🐌 Poor performance and high costs when generating and updating embeddings.
⚙️ Lack of robustness and testability in the embedding generation pipelines.

This makes us invest more time in solving infrastructure problems and less in developing the features that add value.

In this course, you will see different places to generate embeddings, which fits best for each case, and how to make it scalable. All thanks to the potential of Postgres. We will also see the recommended implementation of Supabase and the one generated by Timescale.