Another aspect to consider is the size and density of the vector dataset.

• IVFFlat is better for large, dense datasets.
• HNSW is better where initial data is sparse or data accumulates gradually (our use case).