January 2026 Summaries

Eisuke Izawa's article explores the challenges and solutions for improving retrieval quality in Japanese text using a hybrid search system that combines Sudachi for normalization, Zilliz Cloud's Milvus for vector storage, and AWS Bedrock for dense embeddings. The hybrid search pipeline addresses the language's orthographic variations and mixed scripts by integrating dense vector search for semantic similarity and keyword-based BM25 methods for exact matches. The system employs Reciprocal Rank Fusion (RRF) to merge results, ensuring accuracy and ease of use. The tutorial provided allows users to replicate the setup using Zilliz Cloud's free serverless tier and AWS Bedrock, demonstrating its applicability to scenarios such as internal policy searches and e-commerce product retrieval. This approach showcases robust retrieval capabilities while maintaining low operational overhead by leveraging Milvus's built-in functions for sparse vector generation.

Zilliz, a company founded by former Oracle database engineers, foresaw the potential of vector databases as a solution to handle unstructured data like text, images, and audio, which traditional databases struggled to process. Initially exploring GPU-accelerated databases, Zilliz pivoted to focus on vector representations, leading to the creation of Milvus, an open-source vector database designed for large-scale vector search. The evolution of vector databases from research to production was marked by the need for systems that could handle scalability, reliability, and daily operations alongside search quality. As AI-driven applications became more prevalent, the role of vector databases expanded, supporting tasks such as similarity-based retrieval and Retrieval-Augmented Generation (RAG). Zilliz distinguished itself by building Milvus as a distributed, cloud-native system, emphasizing stability, scalability, and cost-efficiency, and by balancing open-source community engagement with commercial growth. The company's strategic focus on robust infrastructure and operational reliability in vector databases has positioned it to meet the growing demands of AI applications, with an eye toward sustainable operations in the future.

Milvus 2.6.x is now generally available on Zilliz Cloud, offering enhanced vector search capabilities that are faster, smarter, and more cost-efficient for production AI applications. This release brings significant improvements in data ingestion, performance, and cost efficiency, alongside cloud-only optimizations that enhance stability and reliability. Milvus 2.6.x introduces a three-layer tiered storage system that reduces storage costs by up to 87% and automatically adapts to data access patterns, making it ideal for large-scale, cost-sensitive applications. It also offers an Index Build Level feature to balance search accuracy with storage costs, expanded data type support for more AI use cases, and improved JSON handling for faster metadata filtering. Additionally, the platform integrates BM25-optimized full-text search, providing up to 7× faster performance than Elasticsearch, and enhances multilingual support for global applications. With these capabilities, Zilliz Cloud positions itself as a leading fully managed vector database service, supporting the development and scaling of AI applications without operational complexities.