Vector Search on Object Storage Goes MIT-Licensed: No More DB Drama
A new open-source library brings vector search directly to object storage, ditching the need for costly specialized databases.

Developers at Opendata have released an open-source library that performs vector search directly on object storage. Yes, you heard that right: no need to haul a heavy vector database into your project — just dump your vectors into any S3-compatible storage and start searching.
How It Works and Why It Matters
The library uses coarse clustering and Product Quantization to efficiently search through millions of vectors. All under an MIT license, so you can use it in commercial projects without legal headaches. The devs claim performance is on par with heavyweights like FAISS, but without the extra infrastructure overhead.
Who Should Care
- Startups wanting semantic search without paying for a separate database.
- Teams already using object storage who want to avoid another single point of failure.
- Anyone tired of every new service requiring its own cluster and midnight deploys.
Sure, if you have billions of vectors, a specialized solution might still be faster. But for most use cases, this is a sweet spot between simplicity and performance.
METABYTE studio comment: We're all for skipping yet another heavyweight service. If your project needs vector search on a budget, this library might be the IKEA furniture that actually assembles without extra screws.
NEXT STEP
Liked the approach?
We apply the same principles to client projects: AI, automation, products that don't die after launch.