RtBot

A low latency, high efficiency, incremental analytical engine.

• site: https://rtbot.dev
• repo: https://github.com/rtbot-dev/rtbot

What is?

RtBot is a framework, written in c++, to process streams of numerical data in real time and to derive analytics from it with the lowest latency possible. You can use it to detect anomalies, local peaks or to simply continuously transform your data in real time, efficiently.

RtBot have been designed as an extensible dsp framework, where the building blocks are called operators and data flows through them according to the defined connections. Programs are written declaratively in plain json (or yaml), which are essentially definitions of a graph of operators with special input and output nodes.

Use cases

If you need a digital signal processing (dsp) multiplatform framework which can trigger signals on certain events happening on the input numerical stream then RtBot is for you.

Some use cases would be:

• Digital signal processing
• Low latency event detection
• Algorithmic trading

How to effectively use it

RtBot has been designed to work effectively accross different phases of software development, from the exploration over historical data to the production deployment on a web app or into a micro-controller. As such, it can drastically reduce the cost of producing software essentially due to the lack of translation phase. This is perhaps one of the main reasons to adopt it.

Develop an RtBot program means finding the right operators, their configuration and how they are connected, such that they produce the result you want as the data comes in. A common workflow would be to use the python library to develop an RtBot program using historical data, and then, once the program is ready, use the JavaScript version to deploy it into a web app or the c++ code directly to deploy it into any platform, depending on the needs.

Wrappers

JavaScript

RtBot code have been compiled into wasm and can be run both in node and the browser. The published npm package has TypeScript support.

To install:

# npm
npm install --save-dev @rtbot-dev/rtbot
# yarn
yarn add @rtbot-dev/rtbot
# pnpm
pnpm add @rtbot-dev/rtbot

A simple example about how to use it in a node program can be found at rtbot-dev/rtbot-example-ts.

A more interesting example can be found at rtbot-dev/rtbot-example-websocket-ts.

Python

The python wrapper have been designed with a data science public in mind.

Install

The best way to install the python library is to clone the repo and build the wheel. You will need bazel installed first, please follow the instructions for your os in the bazel page.

After cloning the repository build the wheel with:

bazel build //libs/wrappers/python:rtbot_wheel

Once the build finishes, you can install the produced wheel as usual:

# the wheel name will vary according to your os
pip install dist/bin/libs/wrappers/python/rtbot-_VERSION_-py3-none-manylinux2014_x86_64.whl

Another option is to download a pre-compiled version of the wheel from our GitHub releases page. Notice though that we currently support a limited set of os and python version combinations.

Example notebooks

Some example notebooks can be found at the examples/notebook directory.