2026-04-18 02:54 UTC
Thanks for the detailed review, really appreciate it — this is exactly the kind of feedback the project needed before a public release.
You were right on two of the three points, and I've just pushed fixes to Codeberg (commit 93b8985):
**GPG:** you nailed it. `config.py` was silently reusing the user's personal `~/.gnupg` keyring if a key was present, which is obviously wrong for an identity-isolated app. Fixed: PeerBox now uses a dedicated keyring at `system/gpg/` with `--homedir` enforced on every gpg call. Existing users get their key auto-migrated on next `config.py` run so contacts don't break.
**Dependencies:** `install_pysocks()` was calling `pip install --break-system-packages` at runtime, which is indefensible. Removed entirely — it was actually dead code (never called) but still a trap. Documented `PySocks` in the README as a required pip dep alongside cryptography and pywebview. The broader venv/lockfile refactor is a bigger chantier, planned for phase 6.
**SSH keys:** here I have to push back a bit — `ssh_setup.py` does generate a dedicated `id_rsa_peerbox` key, never reuses `id_rsa` or `id_ed25519`. BUT you indirectly uncovered a real bug: `request_creator.py` had a silent fallback reading `id_rsa.pub` / `id_ed25519.pub` if `id_rsa_peerbox.pub` was missing. That's gone now too.
**Quill 1.3.7 → 2.0.3:** fair point, on the roadmap for this weekend. Non-minified too for auditability.
Seriously, thanks for taking the time to read the code and call out what needed calling out. This is what keeps small projects honest.
Replies (4)
-
@randamumaki@lemmy.blahaj.zone 2026-04-18 05:32
Obvious LLM reply is obvious.
-
@ken@discuss.tchncs.de 2026-04-18 04:10
> Thanks for the detailed review I wouldn't call that detailed. It was what glared at me from skimming for a couple of minutes. > was calling pip install --break-system-packages at runtime, which is indefensible "You" still have three instances left of runtime `pip install --break-system-packages` without user interaction, one of which dead code. I think the follow-up beautifully clarified the "is this vibecoded?" question too. Labeling it `1.0.0` seems premature to say the least.
-
@FalschgeldFurkan@lemmy.world 2026-04-18 13:21
Aaaand there goes your credibility. >This is what keeps small projects honest. Oh, the irony!
-
@quack@lemmy.zip 2026-04-24 12:39
Yeah this response is not reassuring. Death to AI slop.