Apprise 1.13.0 was announced as the last 1.x release, but 1.13.1 followed on August 31. The important transition remains the move toward breaking changes in 2.x; the practical response is a tested dependency policy rather than freezing on the first announcement.
The maintenance release changed the immediate choice
The 1.13.0 notes introduce notification services and improvements including Bark encryption and Telegram rich messages. The later 1.13.1 release fixes memory-backed attachments affecting Gentoo, allows periods in Pingram API keys and adds Signalgrid and Lauther.
The maintainer repeats the advice to keep applications on the 1.x range until ready for 2.x. Security support for a year or two is stated as an intention, not a dated contractual support guarantee. Calling 1.13.0 the final possible maintenance package is now demonstrably outdated.
A range and a lock solve different problems
The suggested requirement, apprise greater than or equal to 1.0.0 and less than 2.0.0, prevents the resolver from selecting 2.x. It does not identify a single tested build. Two fresh installations performed on different days can select different permitted versions.
For an original example, installation A resolves 1.13.0 before August 31 and installation B resolves 1.13.1 afterward. Both satisfy the major-version bound. A lock file or an exact resolved-version record is what lets a team reproduce A or B, including the relevant dependency set. The range controls allowed upgrades; the lock records a specific selection.
If Apprise is embedded in another application or supplied by a distribution, make the change through that application's supported dependency mechanism. An unrelated global installation does not necessarily update the copy that sends alerts.
Check delivery, not just package installation
A useful proposed rehearsal sends a uniquely identified test message through each configured channel, including the attachment type the application really uses. Record the sender result, the provider's acceptance when observable, and receipt at the intended destination. A successful package import cannot establish any of those downstream outcomes.
For instance, a text-only check does not exercise the memory-backed attachment path repaired in 1.13.1. A short message does not exercise splitting for a long Matrix notification. Select tests from actual usage rather than assuming one green message validates every format.
Keep credentials out of the recorded evidence, and perform the rehearsal with test recipients. PeopleAreGeek has not sent notifications from readers' configurations; this is a migration worksheet grounded in the release's specific changes.
Update the final-1.x claim with actual 1.13.1 maintenance release; separate a version range from a reproducible lock and propose end-to-end notification validation.