Overview

The R2R Indices API manages vector index structures that enable high-speed similarity search across chunks. It supports multiple index methods, configurable similarity measures, and concurrent building for optimized RAG retrieval performance.

An Index in R2R represents a vector index structure designed to optimize similarity search operations across document chunks. Indices are a core component of R2R’s Retrieval-Augmented Generation (RAG) architecture, enabling fast and scalable semantic retrieval.

By organizing vector embeddings efficiently, indices make it possible to perform high-performance vector searches across large datasets using different similarity metrics and indexing strategies.

Key Capabilities

Indices in R2R provide:

• Fast similarity search for vector-based retrieval.
• Multiple index methods, including HNSW (Hierarchical Navigable Small World) and IVF-Flat.
• Configurable similarity measures such as cosine similarity or inner product.
• Concurrent index building to improve throughput and scalability.
• Performance optimization for large-scale vector operations.

API Endpoints

Method	Endpoint	Description
GET	/indices	List available indices with pagination.
GET	/indices//	Retrieve details and configuration for a specific index.
DELETE	/indices//	Delete an existing index.