vespa-engine/vespa

Performs vector, tensor, and structured data indexing with real-time inference at sub-100ms latency across distributed clusters, supporting continuous corpus updates. Built on a C++/Java hybrid architecture with automatic data partitioning, parallel query evaluation, and machine-learned model serving integrated into the query pipeline. Exposes REST and gRPC APIs for search, recommendation, and personalization workloads at scale.

6,820 stars. Actively maintained with 534 commits in the last 30 days.