Weaviate Features

Allows semantic search service by organizing data.

Allows indexing of data for search and retrieval.

The feature aids in filtering queries by metadata thus achieving accurate search.

This feature integrates vectors and metadata indexes into a single index.

Orchestrates retrieval, reranking, and enrichment steps within RAG workflows

Optimizes chunk selection, context assembly, and grounding specifically for LLM consumption

Enables advanced tuning of lexical, semantic, and reranked retrieval strategies

Provides insights into query behavior, retrieval quality, and relevance performance

Ensures resilience through redundancy, failover, and recovery mechanisms

Manages embedding updates, re-indexing, and model version changes over time

Enables semantic search across text, images, audio, or video using embeddings

Allows teams to bring their own embedding models or LLM providers and change them without re-architecting

Supports near-real-time index updates as source data changes

Enriches content using chunking, metadata inference, entity extraction, or OCR during indexing

Enforces document-, field-, or metadata-level permissions during retrieval

Supports regional data controls, retention rules, and regulatory compliance

Tracks queries, retrieved content, and access decisions for compliance and debugging

Enables fine-grained control over ranking using rules, weights, feedback loops, or learning-to-rank models

Improves retrieval quality through query rewriting, semantic expansion, or intent detection

Supports multi-step retrieval pipelines where initial candidates are reranked using ML or LLM-based approaches

Adjusts retrieval and ranking based on user context, behavior, or application-level signals