and I’m really enjoying how visual and flexible it is. Recently, I started getting interested in machine learning, and I’m wondering—has anyone here used Node-RED in combination with ML models?
I’ve seen some nodes related to TensorFlow and Python integration, but I’m not sure how practical or stable those are for simple use cases like basic prediction or classification tasks.
I’m also preparing for tech interviews, and a lot of Machine Learning Interview Questions are coming up in my studies. It would be great if I could try building a small project in Node-RED to help me better understand ML concepts while keeping the setup beginner-friendly.
If anyone has tried integrating ML into their Node-RED flows—or has good examples or tutorials—I’d really appreciate the help. Even ideas for simple demo projects would be great.
Most ML programming is happening in Python and so Node.js features are somewhat scarcer. I know there are folk in the forum using image classification - mostly for security cameras but those, I think, are all pre-built classifiers.
In addition, a lot of ML requires good co-possessors. Either quality graphics cards or specialist ML devices. Since a lot of Node-RED is run on low-code, low-power devices, there are naturally a smaller number of people using Node-RED for ML.
Personally, I'd love to see more examples using Node-RED. Sadly I don't have the time to dedicate, I am merely a consumer of ML - mostly using GitHub Copilot to help with more traditional coding tasks.
By the way, you might be interested in:
Danfo - powerful javascript data analysis toolkit. Similar API to Python Pandas. Rich HTML output and Plotting (via integrated plotly).
It can convert data direct to tensors for using in Tensor.js for ML.
I'm just playing with it as a front-end library. Partly because I want to learn more about data analytics. Partly because I need some tooling for a new Todo/Actions/Projects tool I'm working on.
For the node-red/uibuilder driven todo tool, I want to be able to quickly manipulate, group and filter arrays of objects which is, of course, exactly what a dataframe is.
Arquero and possibly D3 are also similar types of column-oriented data tools but I like that Danfo stays close to Pandas.
As @TotallyInformation writes, most commonly you would run ML in a really fast and capable machine using the GPU power of such a thing. I just got my hands on the NVIDIA Jetson Orin Nano Super Developer Kit. It is such a test bench you would normally need for advanced AI and ML tasks. I think (it is really powerful)
Running it side-by-side with Node-RED (eventually in a RPi) and then integrating them to provide the results to Node-RED is a simple task that off-loads Node-RED and opens up a huge amount of other useful possibilities
Maybe this link is of interest. It is very interesting what Ultralytics provide:
