NightScribe 0.1.4

Fixed

• The "Saved" line no longer moves the buttons. It appeared on the first
  save and stayed, pushing the Analyse button 77 px down under the
  observer's own cursor, once, for good. It keeps its room now and goes away by
  itself after a couple of seconds, which is what a confirmation is for.

• The plate's own ceiling gates the measurement when the plate is not from
  your camera. The camera profile's linearity limit is written in your ADU,
  and ADU are a property of one instrument: a Las Cumbres plate (a 248 800 ADU
  ceiling) was refused frame after frame as "nonlinear" and the chart came out
  empty with no explanation. The full well in electrons is what tells the two
  detectors apart, and the panel now says which of the two cases it is.

• The plate's filter becomes a band only when it really is one. A plate
  whose FILTER card says R against a sequence that carries Rc used to
  fall to the first band of the list, which is a silent wrong calibration (the
  zero point then hangs from another band's magnitudes). A Cousins R is the
  catalogue's Rc; a Pan-STARRS rp is not, and pretending otherwise
  would be inventing one.

• A series that measured nothing says why. No more empty chart with a clean
  list of frames next to it: the panel counts the flags, names the one that
  dominated and says what to do about it.

• A series that finishes no longer closes the workbench. Saving the
  measurement refreshed the project by RE-OPENING it, and opening a project
  switches the screen to it: the editor the series had just been measured in
  closed itself, and the page it rebuilt took the visit window down with it.
  The project's own blocks (the visits list with its counts, the curve, the
  cadence line, the runs and the header) refresh where they are now, and the
  visit window comes back filled when the observer leaves the workbench.

• "Measure the series…" answers the first click. The workbench is not
  instant to open, and on a real 379-frame visit it took 2.5 seconds during
  which the window did not repaint: with the visit window still on top, the
  pause read as a dead button, so the observer pressed again and the second
  press got the credit for opening it. The visit window steps aside BEFORE that
  work, and the whole open now wears the wait notice, so the press is answered
  on the spot. And the open got 1.8 seconds shorter: arming the Track &
  Stack tab's context reads the header of every frame of the visit (1.85 s for
  379 of them), and that was being paid even when the editor was opened for the
  series. A tab now arms itself when it takes the stage, which is when it is
  looked at.

• The airmass and the site are taken from the plate, not from your
  Settings. The engine computed the target's altitude from the site
  configured in Settings, which is right for your own observatory and wrong
  for anybody else's plate: on a Las Cumbres plate (a 1 m at McDonald) reduced
  from a site in Spain the airmass came out rising to 5.2 through the night
  where the plate's own header said 1.16 to 1.03. That number feeds the
  scintillation term of the error, so the error grew from 0.05 to 0.39 mag
  over the night, which the chart reads as green, then orange, then red. With
  the plate's own cards the same errors are flat at 0.03. The same rule the
  plate scale already followed applies now to the time: what the plate says
  about itself wins over the observer's Settings, and the Settings are the
  fallback for a rig that declares nothing. The detrend benefits too, since the
  airmass is its regressor: re-measure a series to get the corrected errors
  and the corrected trend (the points already stored keep the old numbers, they
  are the record of what was measured).

• The telescope's size is read from the plate as well. It is the input of
  that same scintillation term, and a Las Cumbres 1 m measured with the
  observer's 10" got it 2.5 times too big. Only a plate that really states the
  size is trusted, because that is the only way to read it without inventing it:
  Las Cumbres writes it into the telescope's own id (1m0a, 0m4c, 2m0a),
  while a name like Cecilia or CDK17 says nothing about the size, and there
  the observer's Settings stay: for their own telescope that is the right answer
  anyway. With both cards read, the errors of the series that was reported went
  from 0.05 to 0.39 mag across the night (the green-orange-red ramp) to a flat
  0.016.

Changed

• The detrend starts from "auto" for every kind of project except a transit.
  The extinction is a systematic that is always there, and a rotation, an
  outburst or a pulsation do not share its timescale, which is what makes the
  trend separable. It is the self-limiting policy: it always applies the
  airmass and only adds the extra terms when the scatter really drops. A
  transit keeps the raw curve on purpose, because its trend is solved together
  with the model (EXOTIC's own airmass term).

---

Downloads

• NightScribeSetup-0.1.4.exe: Windows installer (per-user, no admin rights needed).
• NightScribe-0.1.4-windows-x64.zip: Windows portable build; unzip and run nightscribe.exe gui.
• nightscribe_0.1.4_amd64.deb: Debian 13+ / Ubuntu 25.10+ (system Python; includes Find_Orb).
• nightscribe-*.x86_64.rpm: Fedora 42+ / openSUSE Tumbleweed (system Python; includes Find_Orb).

Windows can stop the installer in two different ways, and only one has a way through. SmartScreen shows Windows protected your PC and offers More info -> Run anyway. Smart App Control (Windows 11, on by default on new installations) blocks unsigned code outright, with no such option: either turn it off under Windows Security > App & browser control > Smart App Control, or run NightScribe on another machine. The binaries are unsigned because NightScribe is free software and has no code signing certificate; the details are in docs/user/01-getting-started. Other distros (Ubuntu 24.04 LTS, ...) install manually: see docs/ASTROMETRY.

On Linux, installing the .deb from your Downloads folder can print an apt note about the sandbox user (permission denied, read as root instead): it is harmless and the install finishes.

The NINA/PINS integration is in test: it is there to be tried and to report what is missing, and the rest is planned for 0.2. CCDciel is the complete observatory backend today.

Report issues at https://github.com/irydeo/nightscribe/issues