nitros 0.1.0

Communication made insanely easy - A simple, fast, and reliable communication library for robotics

NitROS - Communication made insanely easy

Nitro + ROS - A simple, fast, and reliable communication library for robotics.

🎯 Why NitROS?

Before (ROS)

1. Create msg file
2. Edit CMakeLists.txt
3. catkin_make
4. Wait 5 minutes
5. Fix build errors
6. Write code

After (NitROS)

pip install nitros

Done in 3 lines of code! ✨

🚀 Quick Start

Basic Usage

from nitros import Publisher, Subscriber

# Publisher
pub = Publisher("topic_name")
pub.send({"x": 1, "y": 2.5})

# Subscriber
def callback(msg):
    print(msg)

sub = Subscriber("topic_name", callback)

Camera Streaming Example

Publisher:

import cv2
from nitros import Publisher

pub = Publisher("camera", compression="image")
cap = cv2.VideoCapture(0)

while True:
    ret, frame = cap.read()
    pub.send(frame)  # Done!

Subscriber:

import cv2
from nitros import Subscriber

def show(frame):
    cv2.imshow('Camera', frame)
    cv2.waitKey(1)

sub = Subscriber("camera", show)  # Done!

📦 Features

• Zero configuration - Automatic peer discovery with mDNS
• Type flexibility - Send dicts, lists, numpy arrays, PyTorch tensors
• Built-in compression - JPEG for images (~10x), quantization+LZ4 for point clouds (~4-5x)
• Reliable - TCP-based, guaranteed delivery
• Simple API - Just Publisher and Subscriber
• Fire and forget - send() never blocks, even without subscribers

📚 Data Types

Send anything

# Dict - just send it
pub.send({"x": 1, "y": 2.5})

# Numpy array - auto-detected
import numpy as np
pub.send(np.array([1, 2, 3]))

# PyTorch tensor - auto-detected
import torch
pub.send(torch.tensor([1, 2]))

Compression for large data

# Images - automatic JPEG compression (~10x reduction)
pub = Publisher("camera", compression="image")
pub.send(camera_frame)  # numpy array

# Point clouds - quantization + LZ4 (~4-5x reduction)
pub = Publisher("lidar", compression="pointcloud")
pub.send(point_cloud)  # numpy array

🔧 Installation

Basic (MessagePack only):

pip install nitros

Full installation (with all features):

pip install nitros[full]

Optional dependencies:

pip install nitros[compression]  # Image/pointcloud compression
pip install nitros[discovery]    # mDNS discovery
pip install nitros[numpy]        # Numpy support

🎨 Design Philosophy

1. "Just works" - Minimal configuration, automatic discovery
2. "Do one thing well" - Perfect communication only
3. "Reliable by default" - TCP first, no packet loss
4. "Progressive enhancement" - Start simple, advanced features when needed

📊 Comparison

vs ROS

✅ Learning curve: 5 minutes vs 5 hours ✅ No need to build msg files ✅ Minimal dependencies ✅ High reliability (TCP) ✅ Low resource usage

vs ZeroMQ

✅ Built-in compression (images, point clouds) ✅ Automatic discovery (no configuration needed) ✅ Automatic type detection ✅ Robotics-specialized

🛠️ Requirements

• Python 3.7+
• msgpack (required)
• numpy (optional, for array support)
• torch (optional, for tensor support)
• opencv-python (optional, for image compression)
• lz4 (optional, for pointcloud compression)
• zeroconf (optional, for mDNS discovery)

📝 License

MIT License

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.