Any search, One database
Hyperspace native lexical search is built from the ground up to leverage the exact term matching of keyword search with the semantic understanding of vector search to achieve superior relevancy, all in a simple and intuitive Python-like syntax.
Optimize search relevancy by combining classic and vector search, achieving breadth and depth in information retrieval.
High precision queries in no time
Easily create complex queries with an intuitive unification of classic and vector search using Python like syntax.
Boost accuracy and precision
Power up queries using a range of functionalities that allow sophisticated logic to achieve the most relevant results.
The power of genuine multi model search
Easily convert business logic to a query with no limitation on the number of keyword filters or logic complexity.
Search without limits
Easily convert business logic to a query with no limitation on the number of keyword filters or logic complexity.
Work seamlessly on a single database
Unify your search capabilities in one solution that include both structured and unstructured data being able to query them both simultaneously.
Genuine multi model search
Hyperspace allows the use of multiple vectors in each document and each query, as well as to use complex score arithmetics and filtering based on any combination of vector search or keyword based scores. This allows to improve recall and precision, by including all relevant information sources in the query, with no need to compromise on functionality.
Intuitive Hybrid Search
Enjoy the best of both worlds, classic and vector search in a single query at a single database.
Rich semantic search with business logic and metadata filtering
Hyperspace allows to integrate rule-based logic into vector search, with various metadata filters. These filters can enforce specific policies, target business units, locations, users, and more. Reducing the search space decreases the load during the scoring phase, significantly minimizing latency and costs.
Pre vs. Post filtering
Pre-filtering – Filters applied first (e.g. by country) and k-NN is running over the remaining vector space.
Post-filtering – Vector search is performed over the entire space using an approximation algorithm (e.g. HNSW), and the filters applied over the top K results.
Superior relevancy with hybrid scoring
Modern information retrieval demands sophisticated ranking capabilities. Traditional methods like TF/IDF provide accurate lexical scoring, while vector search offers versatile, context-aware fuzzy scoring. Hyperspace eliminates the need to choose between these methods by combining the strengths of both in a single query.
Lightning fast hybrid indexing
Don't let software limitations slow you down. With Hyperspace's domain-specific compute architecture, you can index billions of documents, each containing a wide range of vectors and keywords, in a cost-effective way. Retrieve them faster than ever before.
.webp)
