dynavec
A serverless hybrid vector database Python client that fuses Amazon DynamoDB metadata storage with Amazon S3 Vectors approximate-nearest-neighbor search, running entirely inside a user's own AWS account.
dynavec combines Amazon DynamoDB, for single-digit-millisecond metadata and document storage, with Amazon S3 Vectors, for billion-scale approximate-nearest-neighbor search, into one Python client, positioned as a drop-in alternative to hosted vector databases such as Pinecone, Qdrant, Milvus, Weaviate and OpenSearch. Documents are hydrated from DynamoDB with a single batch get after the nearest-neighbor keys come back from S3 Vectors, and every call uses only the caller's own AWS credentials with no third-party data plane involved.
Because it runs on serverless primitives that scale to zero, idle cost is close to storage-only, and IAM is the only access boundary. dynavec adds the layers around the managed ANN engine that a user does control: a two-store hybrid design, metadata pre-filtering, reciprocal-rank-fusion hybrid search and maximal-marginal-relevance reranking, with an optional in-process hot tier for sub-10-millisecond hot-partition queries on the roadmap.