Redis 7.2 Unified Release: Boost AI, Vector Search, and Real‑Time Functions

Redis 7.2.0 has been released as the first “Unified Redis Release”, bringing a broad set of new features and significant investment in AI‑related capabilities.

The announcement describes the release as the most far‑reaching version, aiming to make Redis easier to use, faster, and more innovative for developers.

Making AI innovation easier

Redis supports generative AI workloads with vector database capabilities, offering two index types—FLAT (brute‑force) and HNSW (approximate)—and three distance metrics: cosine, inner product, and Euclidean. Additional features include range queries, hybrid search (filter + semantic), and JSON object support.

Redis helps build and deploy LLM‑powered applications through:

Retrieval‑augmented generation (RAG): Provides hybrid semantic search and can act as an external knowledge base, delivering relevant context to LLMs and reducing hallucinations.

LLM semantic caching: Stores previous query results and retrieves semantically similar cached responses, cutting token usage, latency, and cost.

Recommendation systems: Enables real‑time personalized product recommendations and conversational assistants.

Document search: Allows powerful discovery and retrieval of large document collections using semantic search.

In the past year Redis integrated with popular LLM frameworks such as LlamaIndex, LangChain, RelevanceAI, DocArray, MantiumAI, and the ChatGPT retrieval plugin, and collaborated with NVIDIA on AI Workflows, Tools, and RAPIDS RAFT indexing technology.

Scalable search preview

Redis Enterprise 7.2 introduces a preview of scalable search that can handle high QPS, low‑latency workloads and achieve up to 16× higher query throughput compared with previous search engines.

Deeper integration with developer ecosystems

Redis 7.2 adds new guidance for client libraries and works directly with maintainers of Jedis (Java), node‑redis (Node.js), redis‑py (Python), NRedisStack (.NET), and Go‑Redis (Go). RESP3 protocol support is added to Redis Stack and Redis Enterprise (cloud and software).

Powerful server‑side programmability

Public preview of Triggers and Functions brings a server‑side event‑driven engine that runs TypeScript/JavaScript code inside the database, supporting cross‑shard reads and enabling complex data operations consistently across clients.

Improved geospatial and sorted‑set performance

Polygon search in Redis Stack is enhanced for better geospatial queries, and sorted‑set performance is increased by 30‑100 % over Redis Enterprise Cloud 6.2, benefiting use cases such as game leaderboards.

Enterprise data integration

Redis Data Integration (RDI) is now publicly previewed, allowing data from sources like Oracle, Postgres, or Cassandra to be streamed into Redis, transformed into JSON or Hash formats, and kept in sync in real time without custom code.

For more details, see the official Redis blog.