A semantic search on a website is the best way to make its content easily accessible to users because it interprets the meaning of words, and is in fact increasingly used with the growth of AI technologies.

The implementation of semantic search can be complex and many adopt the strategy of using dedicated vector databases, in addition to the main database, but this strategy has architectural and performance issues.

In this talk we will see a Pythonic way to implement semantic search on a website using a purely Open-Source AI stack (Python, Django, PostgreSQL, pgvector, Sentence Transformes). We’ll analyze some issues of using external vector databases with examples from my experience.

Through this talk you can learn how to add a semantic search on your website, based on Django and PostgreSQL, or you can learn how to update the semantic search function if you use other vector databases.