pramana प्रमाण

Changelog

Every change to pramana, newest first. Releases are git tags (all tags, GitHub releases); the source of this page is CHANGELOG.md.

v0.2.12026-10-28release ↗

  • lib/remote.py in step with anumati 1.4.0's copy: parse_answer and Remote.start take an optional root (the requests that need the password; anumati's commands that run as you don't). pramana doesn't pass it, so every answer still needs the password: no change here.

v0.2.02026-10-26release ↗

  • Answer from your phone: when ekatva runs, each password window also hands its request (polkit's message, the action, the program) to ekatva, which shows it on your paired phone. Your password typed there fills the field here and goes to polkit the same way; denying there cancels. Wrong passwords are reported back; the window sends nothing secret out and finds ekatva by uid. Without ekatva nothing changes. lib/remote.py (the same module as anumati's), ADR 4.

v0.1.02026-10-24release ↗

  • First release: a polkit authentication agent for desktops that don't bring one (sway, Hyprland, river…). When an app asks for admin rights the standard way (mounting a disk, a settings change, pkexec), a window shows what's asking (the message, the polkit action, the program) and takes your password; it goes to polkit's own helper, never anywhere else.
  • Vyom's look in your varna theme, through the same readability checks as anumati's window: a theme can change tints, never hide anything.
  • Choose who to authenticate as when polkit offers several admins (you first). A wrong password keeps the window open to try again; Esc cancels and the app is told no.
  • Runs as the pramana user service in the graphical session, restarting on failure, and steps aside (exit 3, no restart) when another agent already serves the session. Logs to the journal (journalctl --user -u pramana).
  • install.sh installs it root-owned into /usr/local and enables the service.
  • It started inside anumati and became its own tool before release: anumati stays the approval window for AI agents' root commands.