Milvus offers frequently used similarity metrics, including Euclidean distance, inner product, Hamming distance, Jaccard distance, etc, allowing you to explore vector similarity in the most effective and efficient way possible.
Milvus is built on top of multiple optimized Approximate Nearest Neighbor Search (ANNS) indexing libraries, such as faiss, annoy, and hnswlib, ensuring that you always get the best performance across various scenarios.
No longer troubled by static data, you can operate data with insertion, deletion, search and update whenever needed.
Data is available for search almost immediately after being inserted and updated. Milvus does the heavy lifting in your best interests in terms of both result accuracy and data consistency.
Milvus harnesses the parallelism of modern processors and enables billion-scale similarity searches in milliseconds on a single off-the-shelf server.
Milvus supports various data types for fields in a record. You can also use advanced search methods, such as filtering, sorting and aggregation for one or multiple fields.
You can deploy Milvus in a distributed environment. To increase the capacity and reliability of a Milvus cluster, you can simply add more nodes.
We make it easy for you to run Milvus on public cloud, private cloud, or anywhere in between.
Milvus provides easy-to-use SDKs in Python, Java, Go and C++, as well as RESTful APIs.