ops-jrz1/configuration.nix

{ config, pkgs, beads, opencode, ... }:

let
  # ==========================================================================
  # Watchdog Scripts - Referenced directly by systemd, NOT in PATH
  # ==========================================================================

  killswitch = pkgs.writeShellApplication {
    name = "killswitch";
    runtimeInputs = with pkgs; [ procps systemd util-linux coreutils ];
    text = builtins.readFile ./scripts/killswitch;
  };

  cpu-watchdog = pkgs.writeShellApplication {
    name = "cpu-watchdog";
    runtimeInputs = with pkgs; [ procps gawk coreutils util-linux killswitch ];
    text = builtins.replaceStrings
      [ "/usr/local/bin/killswitch" ]
      [ "killswitch" ]
      (builtins.readFile ./scripts/cpu-watchdog);
  };

  egress-watchdog = pkgs.writeShellApplication {
    name = "egress-watchdog";
    # Note: -e removed intentionally - grep returns 1 when no matches
    bashOptions = [ "nounset" "pipefail" ];
    runtimeInputs = with pkgs; [ systemd gnugrep coreutils util-linux killswitch gawk ];
    text = builtins.replaceStrings
      [ "/usr/local/bin/killswitch" ]
      [ "killswitch" ]
      (builtins.readFile ./scripts/egress-watchdog);
  };

  # ==========================================================================
  # Admin Scripts - Added to systemPackages for interactive use
  # ==========================================================================

  dev-add = pkgs.writeShellApplication {
    name = "dev-add";
    runtimeInputs = with pkgs; [ shadow coreutils iproute2 gnugrep gawk ];
    text = builtins.readFile ./scripts/dev-add.sh;
  };

  dev-remove = pkgs.writeShellApplication {
    name = "dev-remove";
    runtimeInputs = with pkgs; [ shadow coreutils gnutar procps findutils ];
    text = builtins.readFile ./scripts/dev-remove.sh;
  };

in
{
  # Main NixOS configuration for ops-jrz1 server
  # Imports host-specific configuration from hosts/ops-jrz1.nix

  # Boot loader configuration (Legacy BIOS for Vultr VPS)
  boot.loader.grub.enable = true;
  boot.loader.grub.device = "/dev/vda";  # Install to MBR

  # Network configuration
  networking.useDHCP = false;
  networking.interfaces.ens3.useDHCP = true;

  # Time zone
  time.timeZone = "UTC";

  # Internationalization
  i18n.defaultLocale = "en_US.UTF-8";

  # System packages
  environment.systemPackages = with pkgs; [
    vim
    git
    htop
    curl
    tmux
    # Dev environment
    python3
    uv
    direnv
    # AI coding tools (via flake inputs)
    beads      # Issue tracker (bd CLI)
    opencode   # AI coding agent (opencode CLI)
    # For npm-based AI tools (gemini-cli, codex): users run npm install
    nodejs_22
    # Admin scripts (declarative deployment)
    dev-add
    dev-remove
  ];

  # Add ~/.local/bin and /usr/local/bin to PATH for manually installed tools
  environment.localBinInPath = true;
  environment.shellInit = ''
    export PATH="/usr/local/bin:$PATH"
  '';

  # Enable Nix flakes
  nix.settings.experimental-features = [ "nix-command" "flakes" ];

  # SSH configuration
  services.openssh = {
    enable = true;
    settings = {
      PermitRootLogin = "prohibit-password";
      PasswordAuthentication = false;
    };
  };

  # SSH authorized keys for root
  users.users.root.openssh.authorizedKeys.keys = [
    "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIOqHsgAuD/8LL6HN3fo7X1ywryQG393pyQ19a154bO+h delpad-2025"
  ];

  # Firewall (will be configured for Matrix services)
  networking.firewall = {
    enable = true;
    allowedTCPPorts = [ 22 80 443 ];  # SSH, HTTP, HTTPS

    # Egress controls for regular users
    # UID range: 1000 (first regular user) to 65534 (nobody - excluded from controls)
    # This covers all dev accounts while excluding system services
    extraCommands = ''
      # Log all new outbound connections from regular users
      iptables -A OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -j LOG --log-prefix "EGRESS: " --log-level info

      # Rate limit new outbound connections (30/min sustained, burst 60)
      iptables -A OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -m limit --limit 30/min --limit-burst 60 -j ACCEPT
      iptables -A OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -j LOG --log-prefix "EGRESS-LIMIT: " --log-level warning
      iptables -A OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -j REJECT
    '';

    # Clean up on stop
    extraStopCommands = ''
      iptables -D OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -j LOG --log-prefix "EGRESS: " --log-level info 2>/dev/null || true
      iptables -D OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -m limit --limit 30/min --limit-burst 60 -j ACCEPT 2>/dev/null || true
      iptables -D OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -j LOG --log-prefix "EGRESS-LIMIT: " --log-level warning 2>/dev/null || true
      iptables -D OUTPUT -m state --state NEW -m owner --uid-owner 1000:65534 \
        -j REJECT 2>/dev/null || true
    '';
  };

  # ACME for Let's Encrypt certificates
  security.acme = {
    acceptTerms = true;
    defaults.email = "dlei@duck.com";
  };

  # Allow deprecated olm library for Matrix bridges
  # Note: olm is deprecated with known CVEs but required by mautrix bridges
  # This is necessary for Matrix bridge functionality until alternatives are available
  nixpkgs.config.permittedInsecurePackages = [
    "olm-3.2.16"
  ];

  # Resource limits for user slices (prevent one user from crashing server)
  systemd.slices."user".sliceConfig = {
    MemoryMax = "80%";    # Users collectively can't exceed 80% RAM
    TasksMax = 500;       # Max 500 total processes across all users
    CPUWeight = 100;      # Fair sharing when contended, bursts allowed
  };

  # Per-user limits - create drop-in on activation
  # (user-.slice is the template for user-XXXX.slice)
  system.activationScripts.userSliceLimits = ''
    mkdir -p /run/systemd/system/user-.slice.d
    cat > /run/systemd/system/user-.slice.d/50-limits.conf << 'EOF'
    [Slice]
    MemoryMax=50%
    TasksMax=200
    CPUQuota=200%
    EOF
  '';

  # CPU watchdog - detects sustained abuse, kills offending user
  systemd.services.cpu-watchdog = {
    description = "CPU abuse watchdog";
    serviceConfig = {
      Type = "oneshot";
      ExecStart = "${cpu-watchdog}/bin/cpu-watchdog";
    };
  };

  systemd.timers.cpu-watchdog = {
    description = "Run CPU watchdog every minute";
    wantedBy = [ "timers.target" ];
    timerConfig = {
      OnBootSec = "1min";
      OnUnitActiveSec = "1min";
    };
  };

  # Egress watchdog - detects users hitting rate limits, kills after 3 strikes
  systemd.services.egress-watchdog = {
    description = "Egress abuse watchdog";
    serviceConfig = {
      Type = "oneshot";
      ExecStart = "${egress-watchdog}/bin/egress-watchdog";
    };
  };

  systemd.timers.egress-watchdog = {
    description = "Run egress watchdog every minute";
    wantedBy = [ "timers.target" ];
    timerConfig = {
      OnBootSec = "1min";
      OnUnitActiveSec = "1min";
    };
  };

  # This value determines the NixOS release compatibility
  system.stateVersion = "24.05";
}