Last updated 07-24-2026
Category:
Reviews:
Join thousands of AI enthusiasts in the World of AI!
SvectorDB
SvectorDB is a serverless vector database built for AWS. It stores vectors, keys, and values, then runs similarity search and hybrid filters so you can ship recommendation engines, semantic search, and RAG without running your own database cluster.
The service uses pay-per-request pricing with no provisioning or scaling steps. Upserts and deletes show up immediately, and you can filter results with Lucene-style key-value queries alongside vector distance. Built-in text and image embedders cover common models, or you can bring your own vectors from any source.
Official clients ship for JavaScript/TypeScript and Python, with an OpenAPI spec for other languages. Sandbox databases let you start free with up to 10 indexes and 5,000 records each, and CloudFormation templates fit into existing AWS workflows.
Hybrid search mixes vector similarity with Lucene-style key-value filters
Upserts and deletes land instantly with no eventual consistency lag
Built-in embedders handle text and images, or bring your own vectors
Pay only per read, write, and stored GB with no minimum fees
Sandbox tier gives 10 free databases at 5,000 records each, no card needed
CloudFormation templates plug into existing AWS infrastructure setups
JavaScript, Python, and OpenAPI clients cover most integration paths
Sandbox tier lets you test with up to 10 free databases and no credit card.
Pay-per-request pricing avoids provisioning or scaling decisions upfront.
Hybrid search combines vector similarity with key-value metadata filters.
Official JavaScript, Python, and OpenAPI clients cover common integration paths.
Users cannot create their own database snapshots; only internal backups are maintained.
Standard databases default to a 1 million record cap unless support raises the limit.
Embeddings are free only while in beta, so future pricing may change.
Does SvectorDB have a free tier?
Yes. SvectorDB offers sandbox databases with free reads and writes. You can create up to 10 sandbox indexes with up to 5,000 records each, and no credit card is required to get started.
How does SvectorDB pricing work?
SvectorDB charges per request on standard databases: $5 per million queries, $20 per million writes, and about $0.25 per GB per month for storage. Sandbox databases are free with usage limits suited for testing.
What programming languages does SvectorDB support?
SvectorDB provides official JavaScript/TypeScript and Python clients. It also publishes an OpenAPI specification, so you can generate clients for other languages.
Can SvectorDB generate embeddings for text and images?
Yes. SvectorDB includes built-in embedders for text (all-MiniLM-L6-v2) and text plus images (CLIP ViT-Base-Patch32). Embeddings are free while in beta, and you can also upload your own vectors.
What are the record limits on SvectorDB databases?
Sandbox databases cap at 5,000 items. Standard databases default to 1 million records per database. Higher limits require contacting SvectorDB support.
Does SvectorDB support hybrid search?
Yes. SvectorDB supports hybrid queries that combine vector similarity with Lucene or ElasticSearch-style key-value filters, so you can narrow results by metadata alongside nearest-neighbor search.