I wrote a short post demonstrating how vector search works by building a D&D spell finder.
Instead of using a vector DB immediately, I implemented the cosine similarity math manually in PHP to demystify how embeddings actually work under the hood.
The code compares query vectors against a JSON dataset of spells to find matches by intent rather than keywords.
I wrote a short post demonstrating how vector search works by building a D&D spell finder.
Instead of using a vector DB immediately, I implemented the cosine similarity math manually in PHP to demystify how embeddings actually work under the hood.
The code compares query vectors against a JSON dataset of spells to find matches by intent rather than keywords.