The openKylin Community’s RISC-V Special Interest Group (SIG) has achieved a major breakthrough in expanding Python and AI ecosystem support on the RISC-V architecture. The team has successfully ported mainstream environment-management toolchains — including Conda and Mamba — to the RISC-V platform, and has built an experimental linux-riscv64 native package channel based on conda-forge.

This milestone significantly enhances RISC-V’s capabilities in scientific computing and package management, laying a solid foundation for running complex AI models, embodied-intelligence (ROS) applications, bioinformatics (Bioconda) workloads, and data-science pipelines on the architecture. It represents another step forward in accelerating the maturation of the global RISC-V ecosystem.

The Main Bottleneck in the RISC-V AI Ecosystem

Although RISC-V hardware platforms have advanced rapidly in recent years, their software stacks for scientific computing and AI have faced persistent challenges. Developers have long lacked a unified, powerful, and user-friendly package manager capable of handling complex software dependencies.

Because the RISC-V architecture historically lacked support for the Conda ecosystem, deploying AI frameworks, ROS applications, or bioinformatics workflows on RISC-V has been difficult, time-consuming, and error-prone — severely limiting the architecture’s adoption in cutting-edge fields.

Key Technical Breakthrough

Conda and Mamba are indispensable tools for environment and package management across data science, machine learning, embodied intelligence (ROS), bioinformatics (Bioconda), and general software development. The openKylin RISC-V SIG’s porting effort enables RISC-V developers, for the first time, to use these powerful tools natively to manage complex Python dependencies and project environments.

Even more importantly, the experimental linux-riscv64 channel they have built is already taking shape. Based on conda-forge, the channel currently includes:

●    sysroot-riscv64: A foundational runtime environment for both cross-platform and native builds.

●    Core compiler toolchain: GCC 14.3 / GCC 15.2 for compiling and packaging native software on RISC-V.

●    Multiple native Python versions: Successfully built Python 3.10, 3.11, 3.12, 3.13, and 3.14 — covering the full range of versions required by mainstream AI frameworks, ROS distributions, and bioinformatics tools.

Looking Ahead: Collaborating with conda-forge and Advancing RVA23 Upstream Support

According to the openKylin RISC-V SIG, the current experimental channel is only the beginning. Building on this work, the team is now collaborating closely with the conda-forge open-source community to push native RISC-V packages into official upstream repositories.

The RISC-V SIG plans to deliver full upstream support for software packages targeting the RVA23 specification within the next 3–6 months. Meanwhile, to ensure a smooth transition for existing users, the openKylin community will continue to maintain the current RV64GC channel to ensure ongoing usability.

Ultimately, the goal is to enable RISC-V developers worldwide to access tens of thousands of precompiled packages — just as seamlessly and effortlessly as on x86 and ARM — simply by running the conda install command. This includes mainstream AI frameworks such as TensorFlow and PyTorch, as well as core tools from the ROS and Bioconda ecosystems.

Try It Now: Miniforge Installation Guide

The openKylin RISC-V SIG encourages developers and enthusiasts in the RISC-V community to try out this new capability and share feedback. To make onboarding easier, the team provides a Miniforge installation script and a concise tutorial.

Step 0: Install prerequisites

(Supported only on openKylin 2.0 RISC-V. Users of other systems should install dependencies manually.)

curl -LO

"https://conda.openkylin.top/miniforge/preinstall.sh.tar"

tar -xf preinstall.sh.tar

bash preinstall.sh

Step 1: Download installer and checksum

curl -LO

"https://conda.openkylin.top/miniforge/Miniforge3-25.9.1-0-Linux-riscv64.sh.tar"

curl -LO

"https://conda.openkylin.top/miniforge/Miniforge3-25.9.1-0-Linux-riscv64.sh.tar.sha256"

Step 2: Verify checksum

sha256sum -c "Miniforge3-25.9.1-0-Linux-riscv64.sh.tar.sha256"

# Expected output should contain "OK"

Step 3: Extract and install

tar -xf Miniforge3-25.9.1-0-Linux-riscv64.sh.tar

bash Miniforge3-25.9.1-0-Linux-riscv64.sh

# Follow on-screen instructions during installation

Step 4: Initialize your shell

"$HOME/miniforge3/bin/conda" init

# After completing the initialization, follow the on-screen instructions to restart your shell

After restarting your shell, you can begin using conda, mamba, and other commands to manage your RISC-V development environment.

About Conda, conda-forge, and Mamba

Conda

An open-source package and environment manager widely used in data science, machine learning, and scientific computing. It lets users create, switch, and manage isolated environments to avoid dependency conflicts across projects.

conda-forge

A community-driven channel providing packages for the Conda ecosystem. It is the largest and most active Conda repository, offering tens of thousands of prebuilt packages. Bringing conda-forge support to RISC-V effectively brings this vast ecosystem to the new architecture.

Mamba

A high-performance, C++ re-implementation of Conda. Designed to eliminate Conda’s dependency-resolution bottlenecks, it provides extremely fast package installation and environment solving, while remaining nearly fully compatible with Conda commands. Many developers consider it the next generation of Conda tooling.

About the openKylin RISC-V SIG

The openKylin RISC-V SIG is responsible for maintaining open-source software packages for the RISC-V architecture within the community, releasing the RISC-V version of openKylin, and performing package and system builds.

SIG homepage:

https://gitee.com/openkylin/community/tree/master/sig/RISC-V

For more information about openKylin, visit our DistroWatch page:

https://distrowatch.com/table.php?distribution=openkylin