Changelog 0.2.2

Added

Recipes: copy, paste or save all plots, signal locations and operations from one data source. (previously known as ‘configurations’).
Chatbot-Integration: copy the ‘AI Recipe Guide’ to a LLM and ask it to analyze your measurement data – it will create a recipe that you can simply import into ScryLab.
New operations: Create Mask and Apply Mask. Create Mask builds a 0/1 mask signal from a threshold comparison (e.g. 1 where speed ≥ 50). Apply Mask then keeps only the samples of any signal where the mask is 1.
Split by Segments: split a measurement into one new data source per segment of a segment signal – e.g. one data source per lap. Segments are runs of equal value in a staircase signal (a lap counter channel works directly; the new ‘Counter from Trigger’ operation turns a lap beacon pulse into one, and ‘Segments from GPS’ detects laps from longitude/latitude channels via start-point re-entry within a radius).
Apply Mask also works on a whole data source: one mask filters every channel at the same time regions, regardless of each channel’s sample rate – e.g. keep only the samples where a condition channel is in range, across the entire measurement file.

Operations are now launched uniformly via the operation palette – no matter if they are appended to a signal or create a new signal.
3D-Stagger view operates now on plot-level (previously on row-level)
The Mask operation is now called Crop (same behavior, clearer next to the new mask operations). Projects saved with the old name are not migrated.
Remove Gaps now also removes NaN runs (drops the samples, then closes the resulting time jumps) – so it composes with Apply Mask.
Operations are now NaN-honest: NaN gaps (e.g. blanked by Apply Mask) stay NaN in the output instead of being filled with interpolated bridge values. Derivative and Integral compute segment-wise (a gap contributes zero area to the integral), the Butterworth filters and Moving Average re-blank the gaps, and Curve Fit / Windowed Linear Regression fit on valid samples only.

Improved performance for multi-row plots
The Data row in the signal properties now says what a signal measures – ‘Y over Time’, ‘Y over Frequency’, ‘Y over X (parametric)’, ‘Z over Time × Frequency’ (spectrogram) – for channels as well as workspace signals. Signals not on the shared time base are skipped by source operations (see tooltip).
Data sources with mixed content are now handled correctly by source operations: source operations act on the source’s SHARED x-base – for a purely frequency-domain source (only saved spectra) Crop/Shift work in Hz as expected, while in a mixed source signals whose x-axis is not the shared time base (a saved FFT, an XY signal) pass through untouched instead of being cut in the wrong domain, spectrograms are cropped/masked along their time columns (previously crashed), a saved spectrum no longer distorts the seamless offsets of Concatenate, and concatenating a non-time signal reports a clear error.