Introducing LangCache and Vector Sets for AI Apps

Introduction

On April 8, 2025, Redis announced two significant advancements aimed at enhancing the development of AI applications: LangCache and vector sets. These innovations are designed to streamline the handling of complex data, enabling developers to create faster and more accurate AI applications.

Key Highlights

• Launch of LangCache, a fully-managed semantic caching service for AI applications.
• Introduction of vector sets, a new native data type for managing vector embeddings.
• Hybrid search capabilities combining full-text and vector similarity search.
• New tools and features on Redis Cloud to support real-time data architecture.
• Redis 8 release candidate now available with significant performance improvements.

LangCache: Fully Managed Semantic Caching

LangCache is a new managed service that provides semantic caching specifically for AI applications. Currently in private preview, it allows developers to integrate large language model (LLM) response caching seamlessly into their applications. The service features a REST API interface, making implementation straightforward.

Key benefits of LangCache include:

• Reduced Latency: By caching user queries and returning relevant stored prompts, LangCache significantly decreases response times.
• Improved Accuracy: Utilises custom fine-tuned models and configurable search criteria to enhance cache retrieval accuracy.
• Streamlined Management: Eliminates the need for separate management of models and API keys.
• Governed Responses: Ensures that applications only return data approved for the current user.

With LangCache, developers can accelerate the building of GenAI applications without the overhead of managing vector databases and cache invalidation.

Vector Sets: A New Native Data Type

Vector sets, developed by Redis creator Salvatore Sanfilippo, introduce a new native data type that simplifies the management of vector embeddings. This data type allows developers to store and query vectors in a manner similar to sorted sets, but with the added capability of semantic searching.

Key features of vector sets include:

• Ability to add items and retrieve subsets based on similarity to a specified vector.
• Enhanced query capabilities through hybrid search, which combines full-text search with vector similarity search.
• Support for quantization, allowing for memory-efficient int8 vector types that reduce memory usage by 75% while maintaining high accuracy.

Integration with LangGraph and Agent Memory

Redis is expanding its GenAI ecosystem with native integrations for LangGraph, enhancing the performance of agentic applications. By integrating Redis with LangGraph’s memory architecture, developers can leverage Redis as a fast data layer, facilitating both short-term and long-term memory management.

Additionally, Redis has released the Redis Agent Memory Server, an open-source service that manages memory for AI conversations. This service includes features such as automatic topic extraction and context summarisation, further enhancing the capabilities of AI agents.

Redis Cloud Enhancements

To support the growing demands of AI applications, Redis is introducing several new features on Redis Cloud:

• Redis Data Integration: A change data capture offering that syncs data between cache and database in real-time.
• Redis Flex: A rearchitected version of Redis that spans both RAM and SSD, allowing for increased data storage without additional costs.
• Redis Insight: A cloud-based tool for querying and managing data directly from a browser, now available in public preview.

Conclusion

With the introduction of LangCache and vector sets, Redis is poised to significantly enhance the development of AI applications. These tools provide developers with the necessary infrastructure to build faster, more efficient, and scalable AI solutions. As Redis continues to innovate, the upcoming release of Redis 8 promises even greater performance improvements and new features. Developers are encouraged to explore these offerings and consider how they can leverage them in their AI projects.