Prototyping with a Node-RED container


If the preceding posts on Node.js and MongoDB have been too complex for you, let me introduce you to a simpler drag-and-drop “flow-editor,” Node-RED.. It was developed and open-sourced by IBM, and is now hosted by the OpenJS Foundation (the same foundation that hosts projects like jQuery, and Dojo). I’ve used Node-RED for quick prototypes and demos, as it’s easy to explain each step of the flow. It saves a lot of coding effort and reduces errors buy providing built-in and pre-built third-party “nodes”.

Starting Node-RED

As usual, we will create a working folder, and from that folder, start up the Node-RED container, obtained from Docker Hub.

docker run --name mynodered -it -p 1880:1880 -v $PWD:/data nodered/node-red-docker

BTW, one could skip mounting the local folder -v altogether, as Node-RED provides a way of importing and exporting flows created via the clipboard. But if you do, then any installed nodes and deployed flows will be saved in your folder. The Node-RED web-based User Interface is exposed on port 1880.

When it’s up and running, you’ll see [info] Server now running at, so go ahead and open your browser to that URL.

You can create a flow to quickly test Node-RED. From the palette on the left, drag the following nodes to the canvas, and connected them up as shown:

  • input > http node - leave the Method as GET and set the URL (route) to /test.
  • function > template node - leave the Property as msg.payload and Format asmustache template (more on that later), and set the Template to Query string = {{}}. Leave the Output a Plain Text.
  • output > http response node - just leave everything as-is.

Hit the red Deploy button, and you’ll get a Successfully deployed toast notification. Now, open another browser tab to http://localhost:1880/test?name=james+bond, and you’ll see the ouput Hello james bond!

Node-RED simple http flow

Visit the Node-RED Cookbook for more samples.


  • To stop the container that is interactive, just hit Ctrl-C
  • To re-start the container, type docker run mynodered
  • To list all containers (running or stopped) in case you forget their names, docker ps -all
  • To delete the (stopped) container, docker rm mynodered
  • To delete the downloaded image, docker rmi nodered/node-red-docker

Adding Capabilities via Nodes

There’s lots more you can do without writing any code! From the menu, go to Manage Palette > Install. For example, one could install:

  • node-red-node-watson - to get access to IBM Watson Assistant to build a chat bot.
  • node-red-contrib-chatbot - to allow Slack, Telegram, etc. to be the front-end to the Watson chat bot.

Note: many nodes are created by third parties, user beware.

Anyway, with just a few nodes, you’ll have a running chatbot! You can log the inputs or outputs of each node by attaching the green debug nodes, e.g.

Node-RED chatbot using IBM Watson Assistant and Telegram

I won’t go into details here though, since Aiman has a tutorial, titled How to create a Watson Chatbot on Node-RED, which integrates IBM Watson Assistant and Language Translator with Telegram and Spotify... without coding!

JSON handing in Node-RED

Actually, I lie. The aforementioned tutorial does use some JavaScript code, but there are other in-built ways to avoid coding in Node-RED altogether.

First, know that Node-RED passes data between nodes in JSON (JavaScript Object Notation) format - simply, it’s a collection of name-value pairs and/or ordered lists (i.e. arrays). See the W3Schools JSON - Introduction.

  • Name-value pairs: { "key": "value" }
  • Arrays: ["a", "b"]
  • All nested together:
    "name": {
        "first": "James",
        "last": "Bond"
    "code": "007",
    "age": 30,
    "cars": [
            "make": "Lotus",
            "model": "Esprit"
            "make": "Aston Martin",
            "model": "DB5"
            "make": "BMW",
            "model": "Z8"

Next, manipulating JSON data is done using the change node. Apart from statically assigning values, you can perform simple transforms on the JSON data using JSONata (another open-sourced IBM innovation). Try out the JSONata Exerciser:

  • Math operators e.g. age++ = 31
  • String manipulation and functions e.g name.first & name.last gives JamesBond and $join([ "a", "b", "c" ]) to produce "abc"
  • Array manipulation[2].make returns BMW and $append([1,2,3], [4,5,6]) creates an array with [1,2,3,4,5,6]
  • Functions e.g. $count(cars) is 3, but there are many functions like $now(), $max(…), $sum(…), $random(), etc.
  • Sorting e.g. cars^(<make) sorts a list in ascending order (i.e. Aston Martin first), or inversely, cars^(>make).
  • Regular expressions to search or filter data e.g. cars[make ~> /(Lotus|BMW)/].model to return only the Lotus and BMW models.

From the Node-REDchange node, select the option expression (the icon looks like J:) and press ... to build and test expressions, e.g.

Node-RED JSONata expression editor

Finally, converting JSON data into text or HTML output is done using the template node, which uses Mustache templates as mentioned above. Mustache is a template-system that replaces placeholders (within {{ and }} tags) with JSON data. You can try the playground at the Mustache home page.

  • Mustache Agent {{code}} is {{name.last}}, <b>{{name.first}}</b> {{name.last}}.
  • JSON { "name": { "first": "James", "last": "Bond" }, "code": "007" }
  • Output in HTML is “Agent 007 is Bond, James Bond.”


No code, simple “install” prototyping. Have fun!