Vector Frankl is a semantic search experiment built around vector embeddings and retrieval. Keyword search is great until you cannot remember the keyword. Then you want the computer to understand “that thing about the architecture decision with the weird tradeoff” and somehow not make you regret asking.

The related post on using a vector database covers the underlying idea: represent meaning as numbers, compare those numbers, and accept that “meaning as numbers” sounds fake right up until it is useful.