Replace the single VPN container (veth pair, host-side auth scripts) with a
multi-container setup on a shared bridge network:
- Dedicated VPN container handles all PIA auth, WireGuard config, NAT, and
optional port forwarding DNAT
- Service containers default-route through VPN container (leak-proof by topology)
- Host runs tinyproxy on bridge for PIA API bootstrap before WG is up
- WG interface is still created in host netns and moved into VPN container
namespace
- Monthly renewal to ensure that connection stays up (PIA allows connections to
last up to 2 months)
- Drop OpenVPN support entirely
Gatus is configured entirely via YAML (mapped from Nix attrsets),
making nix-config the single source of truth for all monitoring
config instead of Uptime Kuma's web UI/SQLite database.
Update the attic cache URL from s0.koi-bebop.ts.net to s0.neet.dev
and configure sandboxed workspaces to inherit the host's binary cache
settings (substituters, trusted keys, netrc auth via agenix).
Add a zfs-alerts module that runs a daily health check on ZFS machines,
sending detailed ntfy notifications for degraded pools, data errors, or
drive errors. Introduce an "ntfy" system role to decouple ntfy alerting
from the server/personal roles, and assign it to all machines.
Replace nix-serve-only setup with Attic for managed binary caching with
upstream filtering and GC. Move gitea actions runner from host into an
isolated NixOS container with private networking. nix-serve kept alongside
Attic during migration.
Add nftables forward rules to prevent sandboxed workspaces from
reaching RFC1918 private addresses while allowing public internet
and the host gateway (for DNS/NAT).
Remove credential passing to sandboxes (didn't work well enough).
Move onboarding config init from host-side setup into base.nix so
each workspace initializes its own Claude config on first boot.
Wrap claude command in VM and Incus workspaces to always skip
permission prompts.
Creates filesystem snapshots before backup for point-in-time consistency.
Uses mount namespaces to bind mount snapshots over original paths, so
restic records correct paths while reading from frozen snapshot data.
- Auto-detects filesystem type via findmnt
- Deterministic snapshot names using path hash
- Graceful fallback for unsupported filesystems
- Add incus.nix module for fully declarative Incus/LXC containers
- Build NixOS LXC images using nixpkgs.lib.nixosSystem
- Ephemeral containers: recreated on each start, cleaned up on stop
- Use flock to serialize concurrent container operations
- Deterministic MAC addresses via lib.mkMac to prevent ARP cache issues
- Add veth* to NetworkManager unmanaged interfaces
- Update CLAUDE.md with coding conventions and shared lib docs
- Add credentials bind mount in container.nix
- Create claude-credentials-dir service to copy credentials for VMs
- Generate .claude.json with onboarding skipped and workspace trusted
- Add allowUnfree to container config
Provides isolated development environments using either VMs (microvm.nix)
or containers (systemd-nspawn) with a unified configuration interface.
Features:
- Unified options with required type field ("vm" or "container")
- Shared base configuration for networking, SSH, users, packages
- Automatic SSH host key generation and persistence
- Shell aliases for workspace management (start/stop/status/ssh)
- Automatic /etc/hosts entries for workspace hostnames
- restartIfChanged support for both VMs and containers
- Passwordless doas in workspaces
Container backend:
- Uses hostBridge for proper bridge networking with /24 subnet
- systemd-networkd for IP configuration
- systemd-resolved for DNS
VM backend:
- TAP interface with deterministic MAC addresses
- virtiofs shares for workspace directories
- vsock CID generation
zuckerberg