The flows are stored as JSON and there is a setting to save them as prettified to make them more readable (`flowFilePretty: true` in the settings.js which is on by default in recent release).

Also the flow storage is fully plugable (https://nodered.org/docs/api/storage/) so nothing to stop you choosing/implementing a different storage format e.g. YAML https://github.com/natcl/node-red-contrib-yaml-storage

But can you actually hand-write it? When I last looked IDs were UUIDs or similar, so you'd have to create a node in the GUI for a start, and even then the structure of it just clearly wasn't designed for manual editing or diffing. 'Storage' sort of sums it up really, doesn't it?

Given Node-RED is intended as a low-code programming tool, hand-writing the flow JSON has never really been a requirement - certainly not one I can recall being brought up in the community. It's certainly possible to generate flow JSON via other means as the format is not hugely complicated once you know the format.

It's an array of objects, each representing a 'thing' in the flow. Each object has a unqiue id (which can be any randomly generated value), and other meta data that defines what the object is and what its configuration is.

Given most nodes have been implemented with a graphical UI in mind for configuring the nodes, you'd have to refer to individual nodes to understand what node-specific properties they require.

> certainly not one I can recall being brought up in the community

Maybe I'm not/HN isn't 'the community', but when I mentioned it 6 months ago [0] you said it was 'an area [you'd] like to improve'. ;)

The relative popularity of my comment above suggests others would be interested too. I understand it's not the focus/target audience necessarily though, I was clearer about that in [0], just a shame (for me) that the low/no-code isn't built on top of code(ish)-allowed.

It could have been an integral part (and really useful!) of ETL pipelines at work, but not without a git-sane, text-editor-editable file format. (JSON ok - not the best perhaps but ok - I mean more about requirements of IDs and general structure of the document.)

[0] - https://news.ycombinator.com/item?id=27896478

Thank you for creating this tool, I use it on multiple machines and enjoy when i find another new use for it.

Are you aware of the projects feature and how it lets you maintain all your flows in a git repo, complete with "visual" diffing?

Yes, but I can also put MS Word .docx files 'in a git repo', and diff them with MS Word proprietary version control. That's not good enough for use at work, in a software project among software engineers who don't need or necessarily want no/low-code; and it's not what I want in personal 'home automation' use either, which is where I gather it's perhaps most popular and how I discovered it - and probably most of that audience does want no/low-code and doesn't know 'what a git is', but, you know, I'm a SE, I know what git is, I like it, can't bare the idea of having anything like HA outside of its control.

Well, the node representation relies on internal identifiers and is not really human-friendly, so I think you’re out of luck there.

(I do run my home automation and quite a few dashboard and minimal API projects on Node-Red and version then that way, but if I want something with an actual lifecycle, I’ll reach for Python or Go…)

Yes, that's my point! I wish that it weren't.

The on-disk text is still crazy but you can turn on version tracking in node-red that lets you track any changes made.