Redis 7.2 Unified Release: New AI, Vector Search, and Programmable Engine Features

Redis 7.2.0 has been launched as the first Unified Redis Release , promising faster performance, easier developer experience, and a broad set of new capabilities.

AI Innovation Made Easier – The release adds vector database support with two index types (FLAT brute‑force and HNSW approximate) and three distance metrics (cosine, inner product, Euclidean). It also offers range queries, hybrid search (filter + semantic), and JSON object handling to power generative AI workloads.

Redis helps build and deploy LLM‑enabled applications through:

Retrieval‑Augmented Generation (RAG) : hybrid semantic search and external knowledge‑base integration for more accurate LLM responses.

LLM Semantic Cache : caches prior query results and retrieves semantically similar responses, reducing token usage and latency.

Recommendation Systems : real‑time, context‑aware product suggestions powered by LLM understanding.

Document Search : powerful discovery and retrieval across large document collections.

Integration with popular development frameworks such as LlamaIndex, LangChain, RelevanceAI, DocArray, MantiumAI, and ChatGPT retrieval plugins, as well as close collaboration with NVIDIA on AI Workflows, Tools, and RAPIDS indexing, expands the ecosystem.

Scalable Search Preview – A new preview of scalable search delivers up to 16× higher query throughput with low latency across clusters.

Enhanced Developer Ecosystem – Updated guidance for client libraries (Jedis, node‑redis, redis‑py, NRedisStack, Go‑Redis) and support for RESP3 in Redis Stack and Redis Enterprise.

Triggers and Functions (Public Preview) – A server‑side event‑driven engine lets developers run TypeScript/JavaScript directly in the database, with cross‑shard reads and cluster‑wide capabilities.

Geospatial Improvements – Polygon search enhancements simplify region‑based queries.

Sorted Set Performance – Optimizations deliver 30%‑100% faster operations, benefiting use‑cases like game leaderboards.

Redis Data Integration (RDI) – A tool for converting any dataset (e.g., Oracle, Postgres, Cassandra) into real‑time Redis data via a configuration UI, streaming changes and mapping them to JSON or Hash formats.

For full details, see the official blog: https://redis.com/blog/introducing-redis-7-2/ .