ScyllaDB is the First to Bring Native Vector Search to the DynamoDB API

SUNNYVALE, Calif., May 5, 2026 – ScyllaDB today announced Vector Search for its DynamoDB-compatible API (Alternator). For the first time, developers building on the DynamoDB API can run high-performance vector similarity search natively. This eliminates the need for managing both DynamoDB and OpenSearch – and using those two separate APIs for vector semantic search queries.

The Problem with DynamoDB + OpenSearch for Vector Search

Until now, DynamoDB users seeking vector search capabilities have been forced into a fragmented, multi-system architecture. Amazon’s recommended approach requires exporting data to S3, synchronizing it to OpenSearch via DynamoDB Streams, and then querying across two completely separate APIs. The reliance on an external OpenSearch cluster to compensate for DynamoDB’s lack of native vector support creates a high-friction architectural split.

While marketed as “Zero ETL,” this approach still requires the cost and operational complexity of a dedicated search cluster. Lambda sync failures cause data drift. DynamoDB deletions may still surface in OpenSearch search results. Schema changes require full re-indexing. The result is two systems to monitor, two APIs to manage, and two bills to pay.

“DynamoDB lacks a native solution,” said Dor Laor, ScyllaDB Co-Founder and CEO. “The bolted-on zero ETL job works, but brings cost up, latency up, and complexity up in the form of unnatural API. With native vector search in ScyllaDB’s DynamoDB-compatible Alternator API, teams can build semantic search, RAG pipelines, and real-time recommendation systems on the infrastructure they already run – without the complexity and cost of a bolt-on search cluster. ScyllaDB is 90-50% lower cost.”

Built for Real-Time AI Workloads

ScyllaDB’s Vector Search is especially attractive to DynamoDB users working with real-time AI applications such as:

• Semantic Product Search: Serve intent-based results at millisecond latency across millions of products.
• Retrieval-Augmented Generation (RAG): Retrieve the most relevant context for LLMs at scale, with 99.2% recall and sub-10ms latency.
• Semantic Deduplication: Scan millions of incoming vectors for near-duplicates in real time, at full ingestion speed.

Availability

ScyllaDB Vector Search for Alternator runs on ScyllaDB Cloud, the company’s fully managed database service. Code changes are not required to migrate existing DynamoDB applications.