The dataset map got honest. SepiaLog draws a map of your project: files, the notes about them, and the words that connect them. The problem: automatic notes (the ones the app writes for you when a dataset is branched) were generating connections out of boilerplate.
Now the connecting phrases come only from wording you actually wrote. Derivations still show up, but through their own parent-to-child links, where they belong.
Hashtags stay hashtags. Writing `#quality #cleaning` used to create the two keyword tags and a phantom "quality cleaning" phrase connecting things that had nothing to do with each other. Fixed.
No more ghost datasets. If you renamed or deleted a file, its old notes hung around on the map as an orphan. Those notes are still in Timeline and Search, nothing is lost, but the map now shows the project as it is today.
A demo you can poke at. The built-in Stock prices demo project now has a proper set of linked notes so you can see keyword, phrase and dataset filtering do something real before you point the app at your own work.
Free, local, on your own machine: www.sepialog.com