MsgFlo setup for programming the c-base space station
C++ Python HTML CoffeeScript Shell JavaScript Other
Fetching latest commit…
Cannot retrieve the latest commit at this time.

c-flo Build Status Greenkeeper badge

MsgFlo setup for rewiring the c-base space station.

Hack weekend: Reprogramming the hackerspace hack weekend was held at c-base, July 7-9 2017. Read about it.


c-base is a crashed space station under Berlin that also happens to be one of the oldest hackerspaces. The station has lots of information systems and devices connected using the c-beam MQTT system. c-flo uses MsgFlo and Flowhub to make those devices and the connections between them visible. It also allows new connections between devices to be made, either directly, or using simple Python components to convert data in-between.

The aim of c-flo is to make all systems at c-base visible in a single graph, and to foster creativity in how these systems and devices interact.

Parts of the system

  • c-beam information transmission system over MQTT
  • c-flo, the MsgFlo coordinator setup for c-base
  • Raspberry Pi 3 running c-flo
  • Flowhub IDE for viewing and reprogramming the c-flo network


The c-flo system is available in c-base crew network. You can open the graph with the following URL:

This will download the Flowhub app to your browser and connect directly with the c-flo coordinator.

Connecting Systems

Each system or device at c-base can be programmed to appear as a node in the c-flo graph by making it a MsgFlo participant. The design principles and how to make systems visible to c-flo are documented below.

Design Principles

All devices and systems at c-base should interface with the c-beam MQTT network. Optimally this means receiving their input from a MQTT topic, and sending their status or output to another MQTT topic.

To maximize interoperability, systems should use their own MQTT topics and not be "hardcoded" to speak with a particular other system. artifact-name/functionality is a good naming scheme. For example display/open_url for a topic where a screen receives URLs to show, and display/opened for a topic where it sends what it is currently displaying.

The MQTT topic names should be made so that there can be multiple instances of same system running. So you might have display1/open and display_downstairs/open.

Connections between systems should be made visually using c-flo.

When needed, Python components can be written as "glue" to convert data from one system to the format wanted by another.

Modifications made to the c-flo network should be committed to this repository.

Existing Participants

When the c-flo effort started, c-base already had several artifacts communicating over the MQTT network. These are represented in MsgFlo-land as "foreign participants" loaded from YAML definitions in the components/ folder.

To document any missing artifacts, please amend the definitions there. See participant discovery for format documentation.

For example, here is how the Echelon network monitoring system was made to appear in c-flo:

component: c-flo/echelon
label: station network traffic monitoring
icon: wifi
inports: {}
    queue: system/echelon/traffic
    type: object

The foreign participant mechanism is fine for making existing systems appear in c-flo. However, any new artifacts should be made to announce themselves.

Self-announcing Participants

If you're building a new system that interfaces over c-beam, it is quite easy to make it self-announce itself on c-flo. This is done by periodically sending a MsgFlo discovery message on the fbp MQTT topic.

The discovery messages are formatted as JSON and contain the following information:

  • Component name (typically the GitHub project name, like c-base/ingress-table). There can be multiple devices running the same component
  • Role, the name of the node in the c-flo graph. This is used to distinguish different instances of same system, so siri and he1 can run same software component but are different devices
  • Inports, listing the topics the system listens to
  • Outports, listing the topics the system writes to

For example, here is the discovery message sent by farbgeber:

  "command": "participant",
    "protocol": "discovery",
    "payload": {
      "component": "c-base/farbgeber",
      "inports": [
        "queue": "farbgeber.IN",
        "type": "bang",
        "id": "in"
      "label": "Produce pleasing color palettes",
      "outports": [
        "queue": "farbgeber.PALETTE",
        "type": "object",
        "id": "palette"
      "role": "farbgeber",
      "id": "farbgeber92490",
      "icon": "tint"

These discovery messages should be sent when the system starts up and connects to the c-base bot network. In addition it should be re-sent roughly once per minute.

There are MsgFlo libraries available for various programming languages to handle the discovery flow automatically.

Some examples of self-announcing participants:

Dynamic Participants

Dynamic participants are started by c-flo itself on-demand. These are typically msgflo-python components used as glue to convert data between other systems or to add dynamic logic to the network.

Some examples of dynamic participants:

  • NetworkBars converts current c-base network traffic to a DMX light visualization
  • DetectABBA tells whether a currently playing song is by ABBA
  • VisualPaging shows current spoken announcements as web pages on connected displays

Testing Participants

Dynamic participants included in this repository can be included in our test automation setup. Tests are written in fbp-spec format and stored as .yaml files in the spec/ folder. For example, here is how we can test the ABBA detector:

name: 'Detecting ABBA'
topic: c-flo/DetectABBA
  type: 'fbp'
  data: |
  name: 'currently playing AC/DC'
  assertion: 'should return false'
      artist: 'AC/DC'
        equals: false
  name: 'currently playing ABBA'
  assertion: 'should return true'
      artist: 'ABBA'
        equals: true

Running locally

  • Install and start a Mosquitto message broker
  • Install the Node.js dependencies of this project with npm install
  • Install the Python dependencies of this project with pip install -r requirements.pip
  • Start the MsgFlo broker with MSGFLO_BROKER=mqtt://localhost npm start

Running with Docker

  • Ensure you have a running Docker daemon
  • Start the project with docker-compose up (use docker-compose-raspberrypi3.yml file if you want to run on RPi3)

Note: by default the MsgFlo coordinator and MQTT ports are only available on localhost. Edit the ports declarations in docker-compose.yml if you want to open them to the outside.

create Markup for WIKI

Command grunt createMarkup will generate a table from the participants folder that can be copied and inserted into the mqtt wiki page (

  • TODO: automate the process, so the wiki is always up to date
  • TODO: add support for FontAwesome to wiki.