Codenary Semantic Engine
High-performance vector search API designed for enterprise document retrieval and knowledge extraction.
Overview
Move beyond keyword matching. The Codenary Semantic Engine utilizes advanced transformer models to understand the contextual meaning of queries and documents. Built for sub-millisecond retrieval across billions of vectors, it natively supports hybrid search (dense + sparse) and custom reranking.
Key Features
Hybrid Retrieval
Combines BM25 lexical search with dense vector embeddings for maximum precision.
Multi-tenant architecture
Built-in namespace separation ensuring strict data boundaries for SaaS platforms.
Dynamic reranking
Cross-encoder models applied to top-k results to push the most relevant documents to the top.
Common Use Cases
- Internal knowledge bases and wikis
- E-commerce product discovery
- Automated customer support routing
Technical Specifications
Protocol
REST & gRPC
Latency
< 50ms at 95th percentile
Vector Store
Custom distributed HNSW index
Throughput
10,000+ QPS per node