Https://tikoci.github.io/restraml/ to https://mt.lv/nightly-build is to costly?

RouterOS
1

@Amm0
Would be hard or costly to include in RouterOS API Schema & CLI Tools — restraml the "ab" releases the appears in https://mt.lv/nightly-build ?

Whether out of fear, laziness, or sheer negligence, MikroTik doesn't include any information or details regarding the "ab" versions of what they call "nightly builds."

Personally, I trust your tool more than their release notes to know about what changed between releases.

Would it be possible for you to include the versions that appear in the nightly build link in your mapper / comparison tool?

What would you need to do that?

2

what notes?
https://www.youtube.com/watch?v=lOjdy0K3X7s&t=8s

:rofl:

3

Do you know if the mt.lv/nightly-build is actually updated "nightly"? Does it always have all the extra-packages? I just haven't regularly followed it... but I recall not being nightly nor having all packages

It's a bit trickier since there are already a lot of beta/rc. And, it uses different URL schemes since URL goes via redirect to box.com so that's a bit more fragile. Also, it's only .npk files, not CHR disk images. So it not quite "play and play" into the CI. One solution may be to bring up stable CHR, then upgrade to nightly, then collect the schema files to the inspect.json/deep-inspect.json files that underpin everything.

Perhaps there could be one "nightly" target that one slot that gets replaced with latest nightly, since that likely the simplest scheme. Thoughts?

@fischerdouglas is talking about his last feature request for a better diff view in my restraml schema/webtools:

So we're on to the new one here :wink:

Side note, the manual.mikrotik.com already does seem to pickup the main branch, since that already predicted xfrm | RouterOS Manual

And I know since my Rosetta MCP for AI sweeps the new docs, so the tests flagged it since it cross-checks the inspect.json used by restraml.

4

I put the robots on it. We'll see what we can do.

...
Got the thread — fischerdouglas wants "ab" nightly builds from mt.lv/nightly-build crawled. Now checking how those version numbers and hosting actually work.

:record_button: Bash(curl -sL -I "https://mt.lv/nightly-build" 2>&1 | head -n 50
echo "---GET---"
curl -sL "https://mt.lv/nightly-build" 2>&1 | head -n 300)
HTTP/2 302
server: nginx/1.20.1
date: Thu, 13 Aug 2026 13:39:52 GMT
content-type: text/html; charset=utf-8
location: https://box.mikrotik.com/d/5ce46054d5d2487e8755/
...

I've been experimenting with different models. So I had Muse Spark 1.2 (Contributor) and running in Claude Code harness turned this into a GitHub issue:

5

And the clunkers churned out the nightly "channel" into restraml's inspect.json/OpenAPI files, so they now render in the restraml. "nightly" is part of the incude testing toggle

The diff.html view is here RouterOS Schema — Side-by-Side Diff
Screenshot 2026-08-13 at 3.05.27 PM

Or if you want to see the REST schema for POST /interface/xfrm/find from nightly: RouterOS REST API — Interactive Explorer

For those curious at the workflow, the "Nightly Build" runs independently from the others since it has to go via box.com and replaces the latest, so there is only one nightly. So the actual build that generate the schemes that the HTML just renders is here for Build #1 — which "just work" the first time:

Review Nightly / ab builds from mt.lv/nightly-build — single overwritten slot (strawman + viability notes) · Issue #90 · tikoci/restraml · GitHub

Built First Try Nightly Build · tikoci/restraml@c2eec4b · GitHub

If curious on AI (and costs), here I used mainly Meta Muse so I can see the price since it's not a subscription, billing looked to be ~$1.50 for half dozen sessions. Muse Spark 1.2 "Contributor" is heavily discounted and somewhere just below Opus and GPT 5.6-Sol but above Sonnet/GPT-5.5, which actually cost 10x more than Meta's discounted prices for letting them see your chats. Since tikoci/restraml are all open source, there is no privacy issue for this use case. And likely still have another $0.50 to polish the UI. e.g. Opus review really thinks users "deserve honesty" to "claim" the version number rather just the generic label "nightly".

UPDATE

Hmm. https://mt.lv/nightly-build now requires a password, at least for me and GitHub.

image
image1700×1332 115 KB

I get not publishing release notes for automated nightly build, but now MikroTik restricted anyone from download them, including restraml. Since I doubt box.com changes its settings on its own, we can eliminate "laziness" since the timing is curious.

But the bait-and-switch is annoying, since the timing is curious here. We got two builds from it... now my scripts are failing... so no new nightly build will be available until MikroTik remove the password needed to download them.

So @fischerdouglas you can ask MikroTik why they added a password...

6

Hey @Amm0
You're a total badass!

7

As I see it, I believe it has to do with dignity and courage—or the lack thereof.

8

I guess Mikrotik found this topic and saw the massive traffic hitting their box.mikrotik.com. Shutting it down.

9

And I'll also add that letting the forum known than from Night Build that some a$$le decided that WinBox 3 should no longer work in 7.25,
was the straw that broke the camel's back...

In short, talking DEMAND too much about the Night Builds was clearly a mistake.
(not a joke for the linked topic)

10

From VTI I suspect. Poor restraml did about a few dozen at most. And it be two a day (x86 and arm64), and if there was actually a new _ab* build. Don't blame me here.

11

@Amm0
It's not the download itself, but all this chatter...
I am convinced that you have nothing directly to do with it.

12

I was wondering what determined MikroTik to put a password on that link.
Congrats to all that contributed to this.
Scrapers, scrapers everywhere.
Meh.

13

It is all here: Https://tikoci.github.io/restraml/ to https://mt.lv/nightly-build is to costly? - #4 by Amm0

It is probably not about the amount of requests. But the fact someone scrapes the nightly builds box automatically and generating docs, extracting infos from it. Seems like Mikrotik does not like that.

14

I'd call it automation and orchestration, and no different than anyone else downloading "by hand" and testing.

As noted in Is https://mt.lv/nightly-build now password protected? - @normis published the link:

So calling my work here as "scraping" - that you're free to ignore - is a bit unfair and pejorative.

15

Tomato, tomahto. A script that programmatically hits a web endpoint to parse and extract data/files without an official API is literally the definition of a scraper.
If it walks like a scraper and quacks like a scraper, then it most likely is a scraper.

16

Ah, using curl to a website unauthenticated URL mentioned by an employee on a public forum is scrapping in your world too. Guilty.

But @Znevna you're kinda implying I'm unethical or wrong is my problem. So I'll note trials are allowed by https://mikrotik.com/software/legal (but also with no warranty or support - so left to just complain openly). The link was published by @normis on a public forum, so I don't know why folks thinks its secret, and the site has no disclaimer or warning suggesting any difference from other packages. In the US, the DMCA would allow even reverse engineering, but I doubt making [100K+] REST API calls to /console/inspect on a containerized CHR qualifies as "reverse engineer". And, on the output, MikroTik cannot copyright the command tree under Google v. Oracle so publishing the results cannot be covered by copyright.

You're entitled your opinion. If MikroTik does not want to publish nightly/_ab builds, or require authentication for them, that their choice. But there is nothing wrong/immoral for using a unauthenticated publicly available link to download a different version of packages, which I had a license to use.

17

Anyway... the password may be a temporary thing. Restraml checks daily at 7AM UTC to see if an unauthenticated download of "nightly-build" is allowed, and it will publish a new nightly if/when the situation changes.

Otherwise only 7.25_ab434 will remain the "nightly" in the restraml tools.

18

I didn't say anything about ethics or legality, I just called a spade a spade.
You got defensive because I called your process a scraper instead of a 'bespoke artisanal data aggregation orchestration thingie.'
You can quote the DMCA and Google v. Oracle all you want, but it doesn't change the fact that it's just a fancy scraper.
Drop the drama.

19

I have another Italian proverb, just extracted from the Giusti-Capponi 1871 book:
[Italian]
Chi ha il mestolo in mano, fa la minestra a modo suo
[/Italian]
that roughly translates to:
Who has the scoop in hand, they make the soup their way.

20

Where are the good old days when Oracle used to sue people for publishing benchmarks of their database...