Dora AI

Dataflow middleware for AI robotics applications, built in Rust

License

Apache-2.0

Overview

Dataflow-Oriented Robotic Architecture (dora-rs) is middleware designed to streamline and simplify the creation of AI-based robotic applications. Built with Rust for performance and reliability, Dora provides zero-overhead message transfer and scalability across machines and robots.

Dora (Dataflow-Oriented Robotic Architecture) is a modern robotic framework that brings zero-overhead message transfer and scalability to robotic applications. Built 100% in Rust, Dora is designed to be 10-17x faster than ROS2 while supporting multi-AI and multi-hardware applications.

It uses Apache Arrow for zero-copy communication, enabling efficient data sharing between nodes without serialization overhead. This architecture allows for CUDA zero-copy integration with PyTorch, making it ideal for AI-powered robotics applications.

Dora supports fast prototyping with YAML-based declarative dataflow definitions and hot reloading for live development. The framework is fully observable with logs, traces, and metrics via OpenTelemetry integration, making it easy to debug and monitor robotic applications in production.

The framework is cross-platform, running on Linux, MacOS, Windows, Docker, and Nix, with multi-language support for Rust, Python, C, and C++. This makes Dora accessible to robotics developers, AI practitioners, and hobbyists alike.

Why it matters

Dora is a modern alternative to ROS that prioritizes performance and simplicity. Built in Rust, it provides the dataflow architecture that robotics needs without the complexity overhead of ROS — making it attractive for teams building production AI robots.

Problem it solves

ROS was designed for traditional robotics and carries complexity that AI-first robots don't need. Dora provides a streamlined middleware focused on the data-intensive communication patterns that modern AI robots require.

Key takeaways

• Dataflow-oriented robotics middleware built in Rust for performance
• Simpler architecture than ROS for AI-centric robot applications
• Zero-copy data sharing between nodes for high throughput
• Python API for easy integration with ML frameworks
• Growing adoption in AI robotics and embodied intelligence

Who it is for

• Teams building AI-first robot systems
• Developers who find ROS overly complex for their needs
• Researchers combining ML pipelines with robot hardware
• Companies building production robot software stacks

Limitations

• Smaller ecosystem than ROS — fewer pre-built packages
• Rust core may be unfamiliar to robotics developers
• Less mature than ROS for traditional robotics applications
• Community and documentation still growing