Website | Python API | Rust API | Guide | Discord
- π dora-rs is a framework to run realtime multi-AI and multi-hardware applications.
- π¦ dora-rs internals are 100% Rust making it extremely fast compared to alternative such as being β‘οΈ 10-17x faster than
ros2
. - βοΈ Includes a large set of pre-packaged nodes for fast prototyping which simplifies integration of hardware, algorithms, and AI models.
Latency benchmark with Python API for both framework, sending 40M of random bytes.
2025
- [03/05] dora-rs has been accepted to GSoC 2025 π, with the following idea list.
- [03/04] Add support for Zenoh for distributed dataflow.
- [03/04] Add support for Meta SAM2, Kokoro(TTS), Improved Qwen2.5 Performance using
llama.cpp
. - [02/25] Add support for Qwen2.5(LLM), Qwen2.5-VL(VLM), outetts(TTS)
dora-rs | |
---|---|
APIs | Python >= 3.7 β
Rust β C/C++ π ROS2 >= Foxy π |
OS | Linux: Arm 32 β
Arm 64 β
x64_86 β
MacOS: Arm 64 β x64_86 β Windows: x64_86 π Android: π οΈ (Blocked by: elast0ny/shared_memory#32) IOS: π οΈ |
Message Format | Arrow β
Standard Specification π οΈ |
Local Communication | Shared Memory β
Cuda IPC π |
Remote Communication | Zenoh π |
Metrics, Tracing, and Logging | Opentelemetry π |
Configuration | YAML β |
Package Manager | pip: Python Node β
Rust Node β
C/C++ Node π οΈ cargo: Rust Node β |
- β = First Class Support
- π = Best Effort Support
- π = Experimental and looking for contributions
- π οΈ = Unsupported but hoped for through contributions
Everything is open for contributions π
Feel free to modify this README with your own nodes so that it benefits the community.
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
PyOrbbeckSDK | π | Image and depth from Orbbeck Camera | |||
PyRealsense | Linuxπ Macπ οΈ |
Image and depth from Realsense | |||
Video Capture | β | Image stream from Camera |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Keyboard | β | Keyboard char listener | |||
Microphone | β | Audio from microphone | |||
PyAudio(Speaker) | β | Output audio from speaker |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Feetech | π | Feetech Client | |||
Dynamixel | π | Dynamixel Client |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Agilex - UGV | π | Robomaster Client | |||
DJI - Robomaster S1 | π | Robomaster Client | |||
Dora Kit Car | π | Open Source Chassis |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Alex Koch - Low Cost Robot | π | Alex Koch - Low Cost Robot Client | |||
Lebai - LM3 | π | Lebai client | |||
Agilex - Piper | π | Agilex arm client |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Pollen - Reachy 1 | π | Reachy 1 Client | |||
Pollen - Reachy 2 | π | Reachy 2 client | |||
Trossen - Aloha | π | Aloha client |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Silero VAD | β | Silero Voice activity detection |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Whisper | β | Transcribe audio to text |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Yolov8 | β | Object detection |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
SAM2 | Cudaβ
Metalπ οΈ |
Segment Anything |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Qwen2.5 | β | Large Language Model using Qwen |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Qwen2.5-vl | β | Vision Language Model using Qwen2.5 VL | |||
InternVL | π | InternVL is a vision language model |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
RDT-1B | π | Infer policy using Robotic Diffusion Transformer |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
ArgosTranslate | π | Open Source translation engine | |||
Opus MT | π | Translate text between language |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Kokoro TTS | β | Efficient Text to Speech |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Llama Factory Recorder | π | Record data to train LLM and VLM | |||
LeRobot Recorder | π | LeRobot Recorder helper |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Plot | β | Simple OpenCV plot visualization | |||
Rerun | β | Visualization tool |
Title | Support | Description | Downloads | License | Release |
---|---|---|---|---|---|
Mujoco | π | Mujoco Simulator | |||
Carla | π | Carla Simulator | |||
Gymnasium | π | Experimental OpenAI Gymnasium bridge |
pip install dora-rs-cli
Additional installation methods
Install dora with our standalone installers, or from crates.io:
cargo install dora-cli
curl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/dora-rs/dora/main/install.sh | bash
powershell -c "irm https://raw.githubusercontent.com/dora-rs/dora/main/install.ps1 | iex"
git clone https://github.com/dora-rs/dora.git
cd dora
cargo build --release -p dora-cli
PATH=$PATH:$(pwd)/target/release
- Run the yolo python example:
## Create a virtual environment
uv venv --seed -p 3.11
## Install nodes dependencies of a remote graph
dora build https://raw.githubusercontent.com/dora-rs/dora/refs/heads/main/examples/object-detection/yolo.yml --uv
## Run yolo graph
dora run yolo.yml --uv
Make sure to have a webcam
To stop your dataflow, you can use ctrl+c
- To understand what is happening, you can look at the dataflow with:
cat yolo.yml
- Resulting in:
nodes:
- id: camera
build: pip install opencv-video-capture
path: opencv-video-capture
inputs:
tick: dora/timer/millis/20
outputs:
- image
env:
CAPTURE_PATH: 0
IMAGE_WIDTH: 640
IMAGE_HEIGHT: 480
- id: object-detection
build: pip install dora-yolo
path: dora-yolo
inputs:
image: camera/image
outputs:
- bbox
- id: plot
build: pip install dora-rerun
path: dora-rerun
inputs:
image: camera/image
boxes2d: object-detection/bbox
- In the above example, we can understand that the camera is sending image to both the rerun viewer as well as a yolo model that generates bounding box that is visualized within rerun.
The full documentation is available on our website. A lot of guides are available on this section of our website.
Dataflow-Oriented Robotic Architecture (dora-rs
) is a framework that makes creation of robotic applications fast and simple.
dora-rs
implements a declarative dataflow paradigm where tasks are split between nodes isolated as individual processes.
The dataflow paradigm has the advantage of creating an abstraction layer that makes robotic applications modular and easily configurable.
Communication between nodes is handled with shared memory on a same machine and TCP on distributed machines. Our shared memory implementation tracks messages across processes and discards them when obsolete. Shared memory slots are cached to avoid new memory allocation.
Nodes communicate with Apache Arrow Data Format.
Apache Arrow is a universal memory format for flat and hierarchical data. The Arrow memory format supports zero-copy reads for lightning-fast data access without serialization overhead. It defines a C data interface without any build-time or link-time dependency requirement, that means that dora-rs
has no compilation step beyond the native compiler of your favourite language.
dora-rs uses Opentelemetry to record all your logs, metrics and traces. This means that the data and telemetry can be linked using a shared abstraction.
Opentelemetry is an open source observability standard that makes dora-rs telemetry collectable by most backends such as elasticsearch, prometheus, Datadog...
Opentelemetry is language independent, backend agnostic, and easily collect distributed data, making it perfect for dora-rs applications.
Note: this feature is marked as unstable.
- Compilation Free Message passing to ROS 2
- Automatic conversion ROS 2 Message <-> Arrow Array
import pyarrow as pa
# Configuration Boilerplate...
turtle_twist_writer = ...
## Arrow Based ROS2 Twist Message
## which does not require ROS2 import
message = pa.array([{
"linear": {
"x": 1,
},
"angular": {
"z": 1
},
}])
turtle_twist_writer.publish(message)
You might want to use ChatGPT to write the Arrow Formatting: https://chat.openai.com/share/4eec1c6d-dbd2-46dc-b6cd-310d2895ba15
We are passionate about supporting contributors of all levels of experience and would love to see you get involved in the project. See the contributing guide to get started.
Our main communication channels are:
Feel free to reach out on any topic, issues or ideas.
We also have a contributing guide.
This project is licensed under Apache-2.0. Check out NOTICE.md for more information.