Signet 1.0 makes the stable release a manual upgrade
Signet, a system for sharing memory between AI agents, has moved its first 1.0 release into the stable channel and paused nightly builds. Existing users must initiate the major upgrade themselves. The transition changes the workspace layout, while older installations without a Dreaming setting need that memory-maintenance system enabled before they can write new memories.
Artificial Intelligence··Morning
Stable releases replace the nightly rhythm
The shared-memory tool Signet, which lets AI agents use context across different working environments, placed version 1.0 in its stable channel on 8 November. The project is pausing nightly distributions and intends to issue stable updates regularly. Much of the accumulated work had already circulated in over 500 nightlies after the preceding stable version, 0.157.3. The announcement therefore marks a change of release channel and upgrade policy for existing users.[1]
A major upgrade also moves the workspace
Users have to start the upgrade themselves because automatic updating excludes a new major version. The background service then moves the workspace into the Workspace v2 arrangement on its first start. If an obstruction prevents that move, the software reverses it and uses the previous arrangement. The project warns against opening an already migrated workspace with an older Signet version. Secrets transferred to the operating system’s keyring also require that keyring for recovery; backing up the workspace by itself is insufficient.[1]
Older workspaces need Dreaming enabled to add memories
Dreaming is Signet’s memory-maintenance system. For an existing workspace with no setting for it, the upgrade leaves it disabled. The former extraction system has stopped operating, so adding memories requires turning Dreaming on and connecting a functioning inference provider. Fresh installations enable it in setup. The project says its users are now mainly individual developers and people using several AI agents, although Signet began as a memory server intended for organisations.[1]
