Relevance as a vector platform

Important problems that can be solved with vectors:

  • Semantic & unstructured data search
  • Recommendation systems
  • Data deduplication & matching
  • Topic modelling
  • User clustering
  • Zero shot classification
  • K-nearest neighbors similarity-based regression
  • Semantic operation
  • and many more

Relevance as a vector platform

In the vector workflow, we decided to focus heavily on the foundation of all good solutions - the experimentation stage. Our experimentation-first approach helps users experiment, tune and prototype various vector weightings, configurations, data structures and vector search methods to improve their vectors.

Relevance AI Workflow PhasesRelevance AI Workflow Phases
Relevance AI's Workflow

Key features as a vector platform:

  • Fully managed API, so that you don't have to manage infrastructure or DevOps.
  • Flexible multi-vector search, flexible search methods that allow you to easily add and weight multiple different vectors into your search query to query against different data structures such as chunks. To find the best vectors and search methods for each problem.
  • Filter and traditional keyword search, combine traditional methods with vector search to maximize search relevance.
  • Real-time vector index, no index rebuilding or constant retraining. Vectors become searchable as soon as they are inserted.
  • Hybrid vector database, flexible data structures that allow for storing of multiple vectors and metadata in 1 dataset so that you don't have to manage multiple nearest neighbor indices or 3rd party metadata store for every dataset.
  • Visualize and interpret vectors, visualize the biases of your vectors in multi-dimensional space with our embeddings projector and vector comparator graphs.
  • Operations beyond vector search, clustering for topic modelling, vector averaging against a category to create category vectors.
  • Production-grade, once you have finished experimenting, and decided on the vectors and vector methods for search, recommendations or more. Deploy into either our enterprise production-grade environment for low latency vector search or your own FAISS or Elasticsearch with ease.