I don't think we need > 1 million vector yet. But we plan to target folder of local document such as obsidian vault or store web search results which could rack

I don't think we need > 1 million vector yet. But we plan to target folder of local document such as obsidian vault or store web search results which could rack up a large number of vectors. Persistence on disk is also a desired feature when looking at the choices because we don't want to reindex existing files. Benchmark is also not everything, easiness to embed and integrate with existing sqlite apps could make sqlite-vec stand out and help with adoption.