SvectorDB

Create a serverless vector database in under 120 seconds

No technical mumbo jumbo, just a few lines of code and you're good to go. Within minutes, you could have a fully managed serverless vector database up and running.

SvectorDB is a powerful serverless vector database designed to simplify the management and retrieval of high-dimensional vectors in various applications.

For a product to be truly serverless, it should be fully managed, scale automatically, and charge per request instead of 'scaling units' like capacity. SvectorDB meets all these criteria, making it a truly serverless product.

Yes, you can try SvectorDB for free, with up to 1,000 items in our sandbox tier.

SvectorDB supports the 3 most common types of indexes: Euclidean, Cosine, Dot Product.

We offer official clients for JavaScript and Python. We also have a publicly available OpenAPI spec, so you can use any language you like.

From a recent benchmark, queries had a mean latency of 9ms and a median of 8.48ms. The benchmark was conducted on a database with 10,000 items, with a vector dimension of 128, using a Euclidean index. Requests were sent in parallel with batches of 10 queries at a time.

Yes, SvectorDB has native CloudFormation support. To get started, check out our CloudFormation documentation.

While almost all limits are adjustable, the default limits are as follows: Reads per second: 100, Writes per second: 100, Items in a single database: 1,000,000.

SvectorDB is designed for developers who want to build applications that require the management and retrieval of high-dimensional vectors. It is particularly useful for applications such as: Recommendation systems, Document similarity search, Image similarity, RAG Chatbots, Embedding nearest neighbour search. Any application that needs to index and retrieve vectors by nearest neighbour can benefit from SvectorDB, whether high or low volume and dimensionality.

While SvectorDB is a powerful tool, it may not be the best fit for every use case. SvectorDB may not be the best fit for you if: You need to store and retrieve non-vector data, You need fine-grained control over algorithmic parameters such as efConstruction, M, etc, You need a fixed cost each month.