Search & Rerank
Zilliz Cloud implements vector-based similarity searches and provides multiple optimizations to improve search efficiency and performance. In this chapter, you will learn the principles and tutorials of various vector search methods available on Zilliz Cloud.
Basic Vector Search [READ MORE]
Based on an index file recording the sorted order of vector embeddings, the Approximate Nearest Neighbor (ANN) search locates a subset of vector embeddings based on the query vector carried in a received search request, compares the query vector with those in the subgroup, and returns the most similar results. With ANN search, Zilliz Cloud provides an efficient search experience. This page helps you to learn how to conduct basic ANN searches.
Tune Recall Rate [READ MORE]
Zilliz Cloud introduces a search parameter `level` to allow users to balance search recall and performance. It also provides another search parameter, `enablerecallcalculation`, to give users the estimated recall rate of the current search. You can combine these two parameters to tune the recall rate of vector searches.
Filtered Search [READ MORE]
An ANN search finds vector embeddings most similar to specified vector embeddings. However, the search results may not always be correct. You can include filtering conditions in a search request so that Zilliz Cloud conducts metadata filtering before conducting ANN searches, reducing the search scope from the whole collection to only the entities matching the specified filtering conditions.
Range Search [READ MORE]
A range search improves search result relevancy by restricting the distance or score of the returned entities within a specific range. This page helps you understand what range search is and the procedures to conduct a range search.
Grouping Search [READ MORE]
A grouping search allows Zilliz Cloud to group the search results by the values in a specified field to aggregate data at a higher level. For example, you can use a basic ANN search to find books similar to the one at hand, but you can use a grouping search to find the book categories that may involve the topics discussed in that book. This topic describes how to use Grouping Search along with key considerations.
Hybrid Search [READ MORE]
In many applications, data is represented using various fields like title and description, or through different modalities such as text, images, and audio. Hybrid search enhances accuracy by combining searches across these diverse fields. Zilliz Cloud supports this feature with multiple vector fields, conducting several approximate nearest neighbor (ANN) searches simultaneously. Multi-vector hybrid search typically yields better results than a single search. This approach is especially valuable in domains that demand a complex understanding of content, such as e-commerce search, academic paper search, and recommendation systems.
Query [READ MORE]
In addition to ANN searches, Zilliz Cloud also supports metadata filtering through queries. This page introduces how to use Query, Get, and QueryIterators to perform metadata filtering.
Filtering [READ MORE]
This chapter explains how to perform filtering on scalar fields and applicable filtering operators.
Full Text Search [READ MORE]
Full text search is a feature that retrieves documents containing specific terms or phrases in text datasets, then ranking the results based on relevance. This feature overcomes semantic search limitations, which might overlook precise terms, ensuring you receive the most accurate and contextually relevant results. Additionally, it simplifies vector searches by accepting raw text input, automatically converting your text data into sparse embeddings without the need to manually generate vector embeddings.
Text Match [READ MORE]
Text match in Zilliz Cloud enables precise document retrieval based on specific terms. This feature is primarily used for filtered search to satisfy specific conditions and can incorporate scalar filtering to refine query results, allowing similarity searches within vectors that meet scalar criteria.
Elasticsearch Queries to Milvus [READ MORE]
Elasticsearch, built on Apache Lucene, is a leading open-source search engine. However, it faces challenges in modern AI applications, including high update costs, poor real-time performance, inefficient shard management, a non-cloud-native design, and excessive resource demands. As a cloud-native vector database, Milvus overcomes these issues with decoupled storage and computing, efficient indexing for high-dimensional data, and seamless integration with modern infrastructures. It offers superior performance and scalability for AI workloads.
Search Iterator [READ MORE]
The ANN Search has a maximum limit on the number of entities that can be recalled in a single query, and simply using basic ANN Search may not meet the demands of large-scale retrieval. For ANN Search requests where topK exceeds 16,384, it is advisable to consider using the SearchIterator. This section will introduce how to use the SearchIterator and related considerations.
Use Partition Key [READ MORE]
The Partition Key is a search optimization solution based on partitions. By designating a specific scalar field as the Partition Key and specifying filtering conditions based on the Partition Key during the search, the search scope can be narrowed down to several partitions, thereby improving search efficiency. This article will introduce how to use the Partition Key and related considerations.
Use mmap [READ MORE]
Memory mapping (Mmap) enables direct memory access to large files on disk, allowing Zilliz Cloud to store indexes and data in both memory and hard drives. This approach helps optimize data placement policy based on access frequency, expanding storage capacity for collections without impacting search performance. This page helps you understand how Zilliz Cloud uses mmap to enable fast and efficient data storage and retrieval.
Consistency Level [READ MORE]
As a distributed vector database, Zilliz Cloud offers multiple levels of consistency to ensure that each node or replica can access the same data during read and write operations. Currently, the supported levels of consistency include Strong, Bounded, Eventually, and Session, with Bounded being the default level of consistency used.
Metric Types [READ MORE]
Similarity metrics are used to measure similarities among vectors. Choosing an appropriate distance metric helps improve classification and clustering performance significantly.
Reranking [READ MORE]
Hybrid Search achieves more precise search results through multiple simultaneous ANN searches. Multiple searches return several sets of results, which require a reranking strategy to help merge and reorder the results and return a single set of results. This guide will introduce the reranking strategies supported by Zilliz Cloud and provide tips for selecting the appropriate reranking strategy.