Una plataforma SaaS que unifique y simplifique la búsqueda vectorial en múltiples bases de datos SQL mediante una API única y adaptable.

https://github.com/mihirahuja1/vectorwrap Over the past few months I kept switching between pgvector in production and SQLite-VSS &#x2F; DuckDB in notebooks. Every swap meant spending time in the syntax land.<p>Vectorwrap hides those differences behind one tiny Python adaptor: create a collection, upsert, query — then change only the connection string when you switch databases.<p>Here, I have tried to solve my direct pain point, so curious if this has application beyond my usecase.<p>What’s inside<p>Backends: Postgres 16 + pgvector, MySQL 8.2 (Vector Store), SQLite VSS, DuckDB VSS<p>Same filter dict syntax on every engine (SQLite uses adaptive oversampling internally)<p>Repo <a href="https:&#x2F;&#x2F;github.com&#x2F;mihirahuja1&#x2F;vectorwrap" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;mihirahuja1&#x2F;vectorwrap</a><p>I’d love some feedback on API design, missing edge cases, or reasons you wouldn’t adopt something like this. Thanks for taking a look!