Redis-inference-optimization is a Redis module designed for serving deep learning and machine learning models directly within Redis, aiming to maximize computation throughput and reduce latency through data locality. It simplifies the deployment and serving of ML graphs by leveraging Redis's robust infrastructure. The project was formerly known as RedisAI but was renamed in January 2025. It is targeted at developers and organizations seeking to integrate ML inference capabilities seamlessly into their data-intensive applications. However, it is important to note that this project is no longer actively maintained or supported.

How It Works

This module acts as a "workhorse" for model serving by providing out-of-the-box support for popular ML frameworks. It executes deep learning/machine learning models and manages their data by adhering to the principle of data locality, processing data where it resides within Redis. This approach minimizes data transfer overhead and enhances performance. The module simplifies the deployment and serving of ML graphs by integrating directly into Redis's production-proven infrastructure.

Quick Start & Requirements

The quickest way to try Redis-inference-optimization is via Docker.

• CPU: docker run -p 6379:6379 redislabs/redisai:1.2.7-cpu-bionic
• GPU: Requires Nvidia driver (CUDA 11.3, cuDNN 8.1+), nvidia-container-toolkit, and Docker 19.03+. docker run -p 6379:6379 --gpus all -it --rm redislabs/redisai:1.2.7-gpu-bionic

Building from source requires:

• Prerequisites: git, python3, make, wget, g++/clang.
• Dependencies: CMake 3.0+, Redis v6.0.0+, CUDA 11.3 and cuDNN 8.1+ (for GPU support).
• Build Process: Clone the repository (git clone --recursive), build backend dependencies (bash get_deps.sh or bash get_deps.sh gpu), and then compile the module (make -C opt clean ALL=1 and make -C opt or make -C opt GPU=1).

Highlighted Details

• Supports PyTorch (libtorch), TensorFlow, TensorFlow Lite, and ONNXRuntime as backends.
• Provides a version map crucial for ensuring model compatibility between Redis-inference-optimization and backend frameworks.
• Client libraries are available for Java (JredisAI, redis-modules-java), Python (redisAI-py, smartredis), Go (redisAI-go), Typescript/Javascript (redisAI-js), and C/C++/Fortran (redis-modules-sdk, smartredis).

Maintenance & Community

Redis-inference-optimization is no longer actively maintained or supported. The project acknowledges the community's past interest and support. Further information on Redis's current AI offerings can be found on the Redis website.

Licensing & Compatibility

The project is licensed under the choice of Redis Source Available License 2.0 (RSALv2) or Server Side Public License v1 (SSPLv1). These licenses may have restrictions on commercial use or linking with closed-source software.

Limitations & Caveats

The primary limitation is that Redis-inference-optimization is no longer actively maintained or supported, indicating a lack of ongoing development, bug fixes, or security updates. Users should be aware of potential compatibility issues with newer versions of Redis or ML frameworks, and the absence of community support for new deployments.