Redis Feature Trails

Feature trailActive

Context Engineering for AI

36 posts · 2026

Timeline

Feature trailActive

Vector Search Indexing and Operations

Redis Search indexes can now live on Flex tiered storage in Redis Cloud Pro, allowing for large-scale search on terabyte-scale datasets with a significantly reduced RAM footprint. This enables use cases like comprehensive fraud screening, full corpus RAG pipelines, and unified caching/search layers. The same Search API and client libraries are used, and users can dynamically adjust the RAM-to-Flash ratio without re-indexing. Query Performance Factor (QPF) is used to maintain acceptable latency. This post extends this by detailing how Redis Flex (RAM + SSD tiering) can be applied to agentic workloads and feature stores to manage increasing memory costs driven by AI, allowing terabyte-scale datasets to be managed with a lower RAM footprint by dynamically adjusting the RAM-to-SSD ratio.

8 posts · 2026

Timeline

Feature trailActive

FastAPI Redis SDK Integration

The official FastAPI Redis SDK provides native integration for FastAPI applications, simplifying connection management, dependency injection for sync and async operations, and enabling first-class HTTP caching and rate limiting behaviors. It integrates with FastAPI Cloud and supports various Redis configurations. This post expands on cache layer architecture, detailing patterns like cache-aside, read-through, write-through, write-behind, and write-around, and discusses failure modes such as cache stampedes, hot keys, cache invalidation, and cache avalanches, along with scaling strategies across nodes and regions.

2 posts · 2026

Timeline

Feature trailActive

Client-Side Geographic Failover

Redis Cloud Pro now offers Active-Active Redis, distributing data across multiple regions for high availability. This is complemented by the `MultiDBClient` in redis-py, which provides client-side geographic failover. The `MultiDBClient` manages health monitoring, weighted endpoint selection, failover, failback, and retries, abstracting these complexities from application developers. It utilizes a circuit-breaker pattern for failure detection and offers configurable health-check policies (HEALTHY_ALL, HEALTHY_ANY, HEALTHY_MAJORITY) to balance failover speed with tolerance for transient failures. Observability is enhanced through runtime notifications and Prometheus metrics.

1 post · 2026

Timeline

Feature trailActive

Atomic Slot Migration for Scaling

Redis Cloud Pro now offers Smooth Scaling, built on Atomic Slot Migration (ASM), to enable faster and less disruptive database scaling. ASM allows for the movement of individual slots directly between shards, minimizing data movement and reducing the impact on live applications. This feature is available for Redis 8.4+ databases using the default hashing policy and RAM-based configurations.

1 post · 2026

Timeline

Feature trailActive

Redis Data Integration Multi-Source and Multi-Pipeline

Redis Data Integration (RDI) in Redis Cloud on AWS is now generally available. RDI is a managed service for moving operational and analytical data into Redis in near real-time, keeping it in sync with source databases and data warehouses. It supports initial bulk hydration and ongoing CDC, transforming source records into Redis data structures optimized for sub-millisecond reads. Key improvements include better CDC latency and throughput, faster full hydration, self-service pipeline management with early validation, advanced source/processor configuration, support for snapshot SQL statements, new source support for Snowflake (preview) and MongoDB, and CAPI support for API and Terraform provisioning. RDI is a foundational building block for Redis Iris, enabling agents to retrieve real-time context from systems of record.

2 posts · 2026

Timeline

Feature trailActive

Bulk Hash Insertion

Redis 8.10 introduces HIMPORT for efficient bulk hash ingestion, reducing network bandwidth by avoiding repeated field name transmission. This command, along with HIMPORT PREPARE and DISCARD, allows for optimized data loading, especially when used with pipelining. Benchmarks show an 11% improvement in import speed for small field sets, with greater gains expected for larger ones. The redis-rb gem supports this new command, with considerations for connection pooling. The post on cache consistency details strategies to keep data fresh, including TTL timing, write-ordering races, multi-instance population races, cache-aside vs. write-through patterns, and event-driven sync using keyspace notifications and change data capture (CDC). It emphasizes that cache consistency is about shrinking the window where cached values disagree with the source database, and discusses trade-offs between synchronous and asynchronous updates.

2 posts · 2026

Timeline

Feature trailActive

Real-time Fraud Detection

This post details how real-time feature stores and sliding-window data structures fit into the scoring pipeline for fraud detection, covering scaling to billions of events and the importance of high availability. It emphasizes that fraud detection is primarily a latency problem, with tight constraints on scoring time. The post highlights the use of Redis for low-latency feature retrieval, employing sorted sets for exact velocity counts and probabilistic structures like Bloom filters, HyperLogLog, and Count-Min Sketch for memory-efficient counting. It also discusses the training-serving skew problem and the benefits of Redis as an in-memory feature store for online inference.

2 posts · 2026

Timeline

Feature trailActive

Redis Security Vulnerability Management

This post details Redis's response to the CVE-2026-81934 vulnerability, a use-after-free flaw in TLS pending-data processing. It outlines the remediation efforts, including expedited fixes for Redis Open Source and Redis Software, and patching of Redis Cloud Essentials. Remediation for Redis Cloud Pro is underway. The post also reiterates best practices for securing Redis instances, such as strong authentication, ACLs, and restricting network access. Redis's assessment of the severity is High, with a CVSS v4.0 score of 7.5. Redis has contacted the assigning organization to review the published score.

3 posts · 2026

Timeline

Feature trailActive

Redis Iris Context Engine

Redis Iris continues to evolve as a real-time context engine for AI agents, integrating memory, retrieval, operational data, and semantic caching. It provides a unified platform for short-term recall, long-term retrieval, and operational state management, with features like Redis Agent Memory for tiered memory management and Redis LangCache for cost-effective LLM inference. The goal is to consolidate agent data paths, reducing complexity and improving reliability, enabling agents to connect to operational data and external tools. This post clarifies the distinction between BI semantic layers and AI context layers, highlighting the need for a dedicated context layer to manage runtime information for AI agents, including retrieval pipelines, short-term and long-term memory, tool definitions, and permission filtering. It emphasizes that context assembly is infrastructure, not glue code, and that a unified context layer prevents context rot, fragmentation across data stores, and retrieval gaps.

17 posts · 2026

Timeline

Feature trailActive

Tail Latency Optimization

This post details how to build a robust integration layer around LLM APIs, focusing on practical engineering challenges beyond the initial API call. It covers cost and latency optimization through caching (prompt and semantic), effective retry strategies for different error types, managing rate limits, and handling session state. The post highlights Redis's role in providing low-latency caching and memory solutions (Redis Iris, Redis LangCache, Redis Agent Memory) for these challenges, and discusses the importance of monitoring latency as a measure of answer quality, not just speed. It details where latency hides in AI networks (retrieval, model inference, cross-service hops) and what to monitor (TTFT, ITL, end-to-end, tail latency) using OpenTelemetry GenAI conventions and SRE golden signals. A faster retrieval layer, through semantic caching and low-latency vector search, is presented as a key strategy for managing latency.

4 posts · 2026

Timeline

Feature trailActive

Agent-to-Agent (A2A) Protocol

This post analyzes five common agent architectures and their protocol needs, comparing the Model Context Protocol (MCP) for agent-to-tool communication and the Agent2Agent (A2A) Protocol for agent-to-agent collaboration. It emphasizes choosing protocols based on the specific boundary being crossed (agent-to-tool vs. agent-to-agent) and highlights security considerations for each. MCP is presented as a convenience layer for tool access with potential security gaps, while A2A focuses on agent identity and coordination. The post also discusses the need for a shared-state layer to manage context across different agent frameworks, as neither MCP nor A2A directly address state management. Redis Iris is presented as a solution for providing this shared state, offering memory, retrieval, and operational data synchronization for multi-framework agent systems.

1 post · 2026

Timeline

Feature trailActive

Model Quantization for Inference

This post provides a practical guide to model quantization for AI inference. It explains what quantization is, how it speeds up inference by reducing data movement and computation, and when it is most effective. The post details different quantization formats (FP32, FP16, INT8, INT4), their impact on accuracy and performance, and discusses scenarios where quantization might not be beneficial. It also contrasts quantization with other inference optimization techniques like batching and caching, highlighting the trade-offs between speed, memory usage, and accuracy. The post also touches on the role of hardware support and software overhead in determining the effectiveness of quantization, and how it complements other optimization techniques like caching.

1 post · 2026

Timeline

Feature trailActive

Dynamic Batching for Inference

This post details dynamic batching as a server-side mechanism to improve GPU utilization in inference servers. It explains how dynamic batching combines individual requests into batches at runtime, trading increased latency for higher throughput. The post contrasts dynamic batching with static batching, explains the role of timeout windows, and discusses the limitations of dynamic batching for autoregressive LLMs, leading into the concept of continuous batching. It also highlights Redis's role in providing low-latency caching and memory solutions (Redis Iris, Redis LangCache, Redis Agent Memory) for these challenges, and discusses semantic caching as a way to reduce requests before they reach the batch queue.

1 post · 2026

Timeline

Feature trailActive

Redis vs Memorystore Comparison

Redis is a unified real-time data platform that delivers sub-millisecond response times for caching, session management, streaming, and AI workloads like vector search and semantic caching, all in a single system. Redis 8 consolidated capabilities that were previously distributed as separate modules into a unified binary, and the Redis 8.x line keeps building on it. Memorystore is Google Cloud's fully managed in-memory service, optimized for caching and session storage inside Google Cloud, not as a unified platform for AI workloads, streaming, or operational data. The tradeoff is scope, version support, and portability. Memorystore's Redis engines are frozen at version 7.2, while Redis continues to evolve with features like vector search, hybrid retrieval, semantic caching, and agent memory under Redis Iris. Memorystore also has limitations in persistence, command availability, configuration control, and is Google Cloud only, whereas Redis offers multi-cloud deployment and active-active geo-distribution.

1 post · 2026

Timeline

Feature trailActive

Redis Cloud AWS Connectivity

This post details the comparison between AWS PrivateLink and VPC peering for connecting to Redis Cloud. PrivateLink is recommended as the default for most deployments due to its enhanced security isolation, operational simplicity across accounts and CIDR ranges, and scoped connectivity. VPC peering is suggested for extremely latency-sensitive workloads where minimal network overhead is critical and network isolation is not a primary concern. Benchmarks indicate a small, measurable latency increase for PrivateLink.

1 post · 2026

Timeline

Feature trailActive

Redis Array Data Structure

Redis 8.8 introduces a new general-purpose array data structure, offering index-addressable collections of string values. Arrays are dynamic, sparse-friendly, and compute-aware, supporting ring buffer semantics, server-side aggregation (SUM, MIN, MAX, AND, OR, XOR), and search capabilities. Benchmarks show arrays offer significantly better random element access performance compared to lists and competitive performance with hashes, with a moderate increase in memory usage. They are ideal for use cases requiring fast indexed access, sliding windows, server-side aggregation, or element searching. The post also details performance improvements in core data types and operations, a new window counter rate limiter, streams message NACKing, subkey notifications for hash fields, explicit control over JSON numeric array storage, multiple aggregators in a single time series query, and a new COUNT aggregator for sorted sets union and intersection.

2 posts · 2026

Timeline

Feature trailActive

Redis Core Performance Improvements

Redis 8.8 introduces significant performance enhancements across core data types and operations. Key improvements include up to 68% faster MGET with pipelining and I/O threads, up to 25% faster HGETALL on large hashes, up to 40% faster SCAN family operations, and up to 83% faster XREADGROUP for streams. Sorted set operations see up to 74% improvement due to faster float parsing, and bitmap operations are vectorized with AVX-512 for up to 28% gains on modern x86. Persistence and replication also benefit from faster full sync (up to 60%) and reduced BGSAVE overhead. These improvements are achieved through techniques like memory prefetching, batched dict bucket prefetching, stack-allocated vector containers for replies, specialized radix tree optimizations for streams, and AVX-512 vectorization for bitmap operations. Diskless full sync is significantly faster, and fork child overhead during snapshotting is reduced. Replication feed path is also optimized, and primary/replica clients are now serviced within I/O threads.

1 post · 2026

Timeline

Feature trailActive

Active-Active Database Replication with CRDTs

This post introduces Conflict-free Replicated Data Types (CRDTs) as a mechanism to enable safer active-active database replication. It explains the challenges of active-active replication, such as write conflicts and data loss, and contrasts CRDTs with traditional approaches like Last Write Wins (LWW) and synchronous consensus protocols. CRDTs are presented as data structures with built-in merge rules that allow concurrent updates to converge to a consistent state without application-level reconciliation. Redis Cloud and Redis Software apply CRDT semantics in Active-Active for strings, hashes, lists, sets, sorted sets, streams, JSON, and HyperLogLog. The merge behavior varies by type: counters and sets converge through commutative operations, hashes merge at the field level, and strings fall back to LWW. The more naturally your workload maps to one of those mergeable types, the more naturally a CRDT-based replication model fits. It isn't a universal answer, though. Data types outside that list don't get CRDT semantics in Active-Active, so a workload built around them may need a different replication strategy. Some workloads are also about preserving invariants that need tighter coordination, such as financial transfers or uniqueness constraints, where one update has to block, validate, or reject another before commit.

1 post · 2026

Timeline