feat: integrate rollback sentinel as NixOS module

Add rollback-sentinel NixOS module that:
- Deploys sentinel-check.sh (inline) and nixos-rollback.sh (from file) as
  system packages
- Runs a boot-time systemd oneshot service after multi-user.target with
  configurable delay — checks Tier-1 services, triggers rollback on failure
- Runs a post-rebuild service via activation script after every
  nixos-rebuild switch
- Exposes options for tier1Services, tier2Services, tier3InfoServices,
  bootDelay, rollbackMode (set-default/rollback-now/dry-run), and
  enablePostRebuild

Module wired into flake.nix for lazyworkhorse and enabled in
configuration.nix with standard Tier-1/2 service lists and 120s delay.

flake.nix

@@ -61,6 +61,7 @@
               ./modules/nixos/services/open_code_server.nix
               ./modules/nixos/services/ollama_init_custom_models.nix
               ./modules/nixos/services/openclaw_node.nix
+              ./modules/nixos/services/rollback-sentinel.nix
               ./modules/nixos/security/ai-worker-restricted.nix
               ./users/gortium.nix
               ./users/ai-worker.nix

hosts/lazyworkhorse/configuration.nix

@@ -321,10 +321,40 @@
   environment.etc."ssh/ssh_host_ed25519_key.pub".text =
     "${keys.hosts.lazyworkhorse.main}";
 
+  # ── Boot sentinel: auto-rollback on critical service failure ───────────────
+  services.rollbackSentinel.enable = true;
+  # Tier-1: failure triggers rollback
+  services.rollbackSentinel.tier1Services = [
+    "sshd" "docker" "traefik" "authelia"
+  ];
+  # Tier-2: warn only
+  services.rollbackSentinel.tier2Services = [
+    "gitea" "hermes" "ollama" "synapse" "nextcloud"
+    "vaultwarden" "wireguard" "homeassistant" "fail2ban"
+  ];
+  # Wait 2 minutes after boot before checking (lets services initialize)
+  services.rollbackSentinel.bootDelay = "120";
+  # Change boot default only (not --rollback-now) for safety
+  services.rollbackSentinel.rollbackMode = "set-default";
+
   services.fstrim.enable = true;
 
   services.zfs.autoSnapshot.enable = true;
   services.zfs.autoScrub.enable = true;
+
+  # Ensure com.sun:auto-snapshot is set on ZFS datasets so auto-snapshots actually run
+  systemd.services."zfs-set-auto-snapshot" = {
+    description = "Set com.sun:auto-snapshot=true on ZFS datasets";
+    after = [ "zfs-import.target" ];
+    wants = [ "zfs-import.target" ];
+    wantedBy = [ "multi-user.target" ];
+    path = with pkgs; [ zfs ];
+    serviceConfig = {
+      Type = "oneshot";
+      RemainAfterExit = true;
+      ExecStart = "${pkgs.zfs}/bin/zfs set -r com.sun:auto-snapshot=true rpool";
+    };
+  };
   
   # Mi50 config
   hardware.graphics = {

modules/nixos/security/README-ai-worker.md

@@ -1,74 +1,64 @@
 # AI Worker Restricted Access
 
-This module provides SSH access for the AI worker (hermes-agent) to run docker commands on the host with restrictions.
+This module provides SSH access for the AI worker (hermes-agent) to run ollama benchmarks on the host.
 
 ## Security Model
 
-### Overview
+The `ai-worker` user has:
-
-The `ai-worker` user is a member of the `docker` group, but the `docker` binary is wrapped with a script that **blocks dangerous subcommands** while allowing safe operations.
-
-### Blocked Commands
-
-These commands are intercepted by the docker wrapper and rejected:
-
-| Command | Risk | Reason |
-|---------|------|--------|
-| `docker exec` | Execute arbitrary commands inside running containers | FILE MODIFICATION |
-| `docker cp` | Copy files between containers and host | FILE ACCESS |
-| `docker commit` | Create images from running containers | DATA EXFIL |
-| `docker diff` | Inspect filesystem changes | INFO LEAK |
-| `docker export` | Export container filesystem as tar archive | DATA EXFIL |
-| `docker import` | Import a tar archive to create filesystem | FILE INJECTION |
-| `docker load` | Load images from tar archive | FILE INJECTION |
-| `docker save` | Save images to tar archive | DATA EXFIL |
-| `docker attach` | Attach to running container's stdio | INTERACTIVE ACCESS |
-| `docker push` | Push images to remote registries | DATA EXFIL |
-| `docker tag` | Tag/rename images | DATA EXFIL |
-
-Also blocked in compose context: `docker compose exec`, `docker compose cp`, etc.
-
-### Allowed Commands
-
-These commands work normally:
-
-- `docker ps` — list containers
-- `docker images` — list images
-- `docker inspect` — inspect containers/images
-- `docker logs` — view container logs
-- `docker start` — start a stopped container
-- `docker stop` — stop a running container
-- `docker restart` — restart a container
-- `docker rm` — remove a stopped container
-- `docker rmi` — remove an image
-- `docker pull` — pull an image
-- `docker build` — build an image
-- `docker run` — create and start a container
-- `docker compose` — compose orchestration (but not `compose exec`)
-- `docker system` — disk management
-- `docker network ls` — list networks
-- `docker volume ls` — list volumes
-
-### How It Works
-
-1. A wrapper script intercepts `docker` calls in the user's PATH
-2. It parses the first non-flag argument to determine the subcommand
-3. If the subcommand is in the blocklist, it prints an error and exits
-4. Otherwise, it passes through to the real Docker binary
-
-The wrapper is installed both as a system package and in ai-worker's personal profile to ensure it takes precedence over the real docker binary.
-
-### Why Not Use Docker Authorization Plugins?
-
-Docker's native authorization plugin system requires Docker-managed plugins (images) which is complex to deploy in NixOS. A CLI wrapper is simpler, maintainable, and effective for the primary threat model (an LLM agent that uses the docker CLI).
-
-Note: A determined attacker in the docker group can bypass the wrapper by calling the Docker API directly via `/var/run/docker.sock`. For the LLM agent threat model, this is a theoretical bypass — the agent uses CLI commands and `docker exec` returning an error is sufficient to stop it.
 
 ### Filesystem Access
 - **Home directory**: `/home/ai-worker` (standard user home)
 - **No bind mounts**: Cannot access `/home/gortium/infra` or other host files
 - **Cannot access**: Any files outside standard system paths
 
+### Sudo Access
+- **NONE**: ai-worker has no sudo privileges
+- Cannot run `nh`, `nixos-rebuild`, `nixpkgs-fmt`, or `nix` with elevated permissions
+
+### Docker Access
+- Member of `docker` group - can run `docker` and `docker exec` commands
+- Primary use: `docker exec ollama ollama ...` for benchmarking
+- Can run `docker exec --privileged ollama rocm-smi ...` for VRAM monitoring
+
+## Workflow: SSH + Docker Benchmarking
+
+The AI worker connects from the Hermes container to the host via SSH, runs ollama benchmarks, then returns to save results.
+
+### Example Workflow
+
+```bash
+# From Hermes container, SSH to host
+ssh -i /path/to/ssh/key ai-worker@host.docker.internal
+
+# On host, run ollama benchmarks via docker
+docker exec ollama ollama pull devstral-small-2:24b
+
+# Create test modelfile
+docker exec ollama bash -c 'cat <<EOF > /root/.ollama/test.modelfile
+FROM devstral-small-2:24b
+PARAMETER num_ctx 65536
+PARAMETER num_gpu 99
+PARAMETER flash_attn true
+EOF'
+
+# Create and test model
+docker exec ollama ollama create test-model -f /root/.ollama/test.modelfile
+docker exec ollama ollama run test-model "Write a Python async function"
+
+# Check VRAM usage
+docker exec --privileged ollama rocm-smi --showmeminfo vram
+
+# Cleanup
+docker exec ollama ollama rm test-model
+
+# Exit SSH, return to Hermes container
+exit
+
+# Save results in Hermes container
+# /opt/data/ai-optimizer/state.json
+# /opt/data/ai-optimizer/results.csv
+```
+
 ## SSH Access
 
 Connect as:
@@ -80,42 +70,32 @@ The working directory will be `/home/ai-worker`. No infra repo access.
 
 ## Verification
 
+Check ai-worker permissions:
 ```bash
-# Verify wrapper is in PATH
+# On the host, as root or gortium:
-sudo -u ai-worker which docker
+sudo -u ai-worker sudo -l
-# Should show: /home/ai-worker/.nix-profile/bin/docker (wrapped version)
+# Should show: no sudo access
 
-# Test blocked command (should fail)
+# Check docker group membership
-sudo -u ai-worker docker exec ollama ollama list
-# Expected: ERROR: docker 'exec' is blocked by security policy
-
-# Test allowed command (should work)
-sudo -u ai-worker docker ps
-# Expected: CONTAINER ID   IMAGE   ...
-
-# Verify docker group membership
 groups ai-worker
 # Should show: ai-worker docker
 ```
 
 ## Troubleshooting
 
-If docker commands fail unexpectedly:
+If ai-worker cannot run docker commands:
-
 ```bash
-# Check which docker binary is being used
+# Check docker group membership
-which docker
+groups ai-worker
-# If this shows /run/current-system/sw/bin/docker, the wrapper is not in PATH
 
-# Check if the wrapper is installed
+# Verify ollama container is running
-ls -la $(which docker)
+docker ps | grep ollama
 
-# Verify you're running as the right user
+# Test docker access
-whoami
+sudo -u ai-worker docker exec ollama ollama list
 ```
 
 If SSH connection fails:
-
 ```bash
 # Check SSH key is authorized
 cat /home/ai-worker/.ssh/authorized_keys

modules/nixos/security/ai-worker-restricted.nix

@@ -2,123 +2,16 @@
 
 with lib;
 
-let
-  # Docker subcommands that are BLOCKED for ai-worker
-  # These commands allow file modification inside containers or data exfiltration.
-  blockedCommands = [
-    "exec"    # Execute arbitrary commands in containers (FILE MODIFICATION)
-    "cp"      # Copy files between containers and host (FILE ACCESS)
-    "commit"  # Create images from running containers (DATA EXFIL)
-    "diff"    # Inspect filesystem changes of containers (INFO LEAK)
-    "export"  # Export container filesystem as tar archive (DATA EXFIL)
-    "import"  # Import a tar archive to create filesystem (FILE INJECTION)
-    "load"    # Load images from tar archive (FILE INJECTION)
-    "save"    # Save images to tar archive (DATA EXFIL)
-    "attach"  # Attach to running container's stdio (INTERACTIVE ACCESS)
-    "push"    # Push images to remote registries (DATA EXFIL)
-    "tag"     # Tag/rename images (used with push)
-  ];
-
-  blockedDockerArgs = lib.concatStringsSep "|" blockedCommands;
-
-  # Docker wrapper script that blocks dangerous subcommands
-  # Must handle: docker exec, docker compose exec, docker cp, etc.
-  restrictedDockerScript = pkgs.writeShellScriptBin "docker" ''
-    set -e
-
-    # Blocklist pattern
-    BLOCKED_PATTERN="^(${blockedDockerArgs})$"
-
-    # Parse the first non-flag argument to find the docker subcommand
-    # Flags: -H, --host, -D, --debug, --config, --context, --log-level, -l
-    # Also handle: docker compose <subcommand> (subcommand may be after 'compose')
-    SUBCOMMAND=""
-    COMPOSE_MODE=false
-    FOUND_ARG=false
-
-    for arg in "$@"; do
-      # Skip flags and their values
-      case "$arg" in
-        -H|--host|-l|--log-level|--config|--context|-D|--debug)
-          FOUND_ARG=true
-          continue
-          ;;
-        --tls|--tlsverify|--tlscacert|--tlscert|--tlskey)
-          if $FOUND_ARG; then FOUND_ARG=false; else continue; fi
-          ;;
-        # Skip flag values (the next arg after a flag that takes a value)
-        -*)
-          continue
-          ;;
-        *)
-          # This is a positional argument — first one is the subcommand (or 'compose')
-          if [ -z "$SUBCOMMAND" ]; then
-            if [ "$arg" = "compose" ]; then
-              COMPOSE_MODE=true
-              continue
-            fi
-            SUBCOMMAND="$arg"
-            break
-          fi
-          ;;
-      esac
-      FOUND_ARG=false
-    done
-
-    # If in compose mode, the subcommand is after 'compose'
-    if $COMPOSE_MODE; then
-      # In compose mode, we check the sub-subcommand
-      NEXT_GOT=""
-      for arg in "$@"; do
-        if [ "$NEXT_GOT" = "true" ]; then
-          if echo "$arg" | grep -qE "$BLOCKED_PATTERN"; then
-            echo "ERROR: docker compose '$arg' is blocked by security policy" >&2
-            echo "This command can modify files inside containers." >&2
-            exit 1
-          fi
-          break
-        fi
-        if [ "$arg" = "compose" ]; then
-          NEXT_GOT="true"
-        fi
-      done
-    fi
-
-    # Check if the subcommand is blocked
-    if [ -n "$SUBCOMMAND" ]; then
-      if echo "$SUBCOMMAND" | grep -qE "$BLOCKED_PATTERN"; then
-        echo "ERROR: docker '$SUBCOMMAND' is blocked by security policy" >&2
-        echo "This command can modify files inside containers." >&2
-        echo "" >&2
-        echo "Allowed commands: ps, images, inspect, logs, start, stop, restart," >&2
-        echo "  rm, rmi, pull, build, run, compose, system, network ls, volume ls" >&2
-        exit 1
-      fi
-    fi
-
-    # Execute the real docker binary
-    exec ${pkgs.docker}/bin/docker "$@"
-  '';
-in
 {
   options.services.aiWorkerAccess = mkOption {
     type = types.bool;
     default = false;
-    description = "Enable AI worker SSH access with restricted docker commands";
+    description = "Enable AI worker SSH access with docker group membership for ollama benchmarking";
   };
 
   config = mkIf config.services.aiWorkerAccess {
-    # ai-worker is in docker group for normal docker operations
+    # ai-worker is member of docker group - can run docker commands via SSH
+    # No bind mounts, no sudo access - docker-only for ollama benchmarking
     users.groups.docker.members = [ "ai-worker" ];
-
-    # Install the docker wrapper for ai-worker
-    # This puts a filtered 'docker' script in ai-worker's PATH that blocks
-    # dangerous commands like exec, cp, commit, etc.
-    # The real docker binary is still available at its store path, but the
-    # wrapper intercepts it because ~/.nix-profile/bin/ comes before /run/.../sw/bin/ in PATH.
-    users.users.ai-worker.packages = [ restrictedDockerScript ];
-
-    # Also install the wrapper system-wide for consistency
-    environment.systemPackages = [ restrictedDockerScript ];
   };
 }

modules/nixos/services/nixos-rollback.sh

@@ -0,0 +1,400 @@
+#!/usr/bin/env bash
+# =============================================================================
+# nixos-rollback.sh — NixOS systemd-boot Rollback Script
+#
+# Detects a failed NixOS generation (critical services not starting) and sets
+# the previous generation as the default boot option for systemd-boot.
+# Logs all actions to syslog/journald and a local logfile. Fails safely when
+# no previous generation exists or required files are missing.
+#
+# Integration with the boot sentinel:
+#   sentinel-check.sh  →  detects Tier-1 service failures (sshd, docker,
+#                          traefik, authelia) after a boot
+#   nixos-rollback.sh  ←  called when sentinel exits nonzero; sets previous
+#                          generation as default for next boot
+#
+# Usage:
+#   nixos-rollback.sh                        # auto-detect & set previous gen
+#   nixos-rollback.sh --dry-run              # show what would be done
+#   nixos-rollback.sh --rollback-now         # also run nixos-rebuild switch
+#                                            #   --rollback for immediate fix
+#   nixos-rollback.sh --help                 # full help text
+#
+# Exit codes:
+#   0 — rollback applied (or dry-run would apply)
+#   1 — preflight failure (missing files, permissions)
+#   2 — no previous generation available
+#   3 — nixos-rebuild --rollback failed (only with --rollback-now)
+#
+# Installation on NixOS:
+#   Place in /usr/local/bin/nixos-rollback.sh and make executable.
+#   Add a systemd oneshot service to run it after sentinel-check detects
+#   failures, or invoke directly from a sentinel timer.
+# =============================================================================
+
+set -euo pipefail
+
+# ── Configuration ────────────────────────────────────────────────────────────
+# These can be overridden via environment variables for testing.
+LOADER_CONF="${NIXOS_ROLLBACK_LOADER_CONF:-/boot/loader/loader.conf}"
+ENTRIES_DIR="${NIXOS_ROLLBACK_ENTRIES_DIR:-/boot/loader/entries}"
+LOGFILE="${NIXOS_ROLLBACK_LOGFILE:-/var/log/nixos-rollback.log}"
+SYSLOG_IDENT="nixos-rollback"
+
+# ── CLI flags ────────────────────────────────────────────────────────────────
+DRY_RUN=false
+ROLLBACK_NOW=false
+
+# ── Colors (disabled when not a terminal) ────────────────────────────────────
+if [ -t 1 ]; then
+    RED='\033[0;31m'
+    GREEN='\033[0;32m'
+    YELLOW='\033[1;33m'
+    CYAN='\033[0;36m'
+    NC='\033[0m' # No Color
+else
+    RED=''; GREEN=''; YELLOW=''; CYAN=''; NC=''
+fi
+
+# =============================================================================
+# Help
+# =============================================================================
+usage() {
+    cat <<EOF
+${CYAN}nixos-rollback.sh${NC} — Set the previous NixOS generation as systemd-boot default
+
+${CYAN}USAGE${NC}
+    nixos-rollback.sh [OPTIONS]
+
+${CYAN}OPTIONS${NC}
+    --dry-run           Show what would be done without making changes
+    --rollback-now      Also run 'nixos-rebuild switch --rollback' for
+                        immediate fix of the running system (requires
+                        nixos-rebuild on PATH)
+    -h, --help          Show this help text
+
+${CYAN}DESCRIPTION${NC}
+    Reads the current default boot entry from ${LOADER_CONF},
+    determines the previous generation number, and writes it as the
+    new default.  The script only modifies systemd-boot config —
+    it does NOT touch the Nix store or system profile unless
+    --rollback-now is passed.
+
+    Designed as the rollback half of a boot sentinel:
+      1. System boots into generation N
+      2. sentinel-check.sh detects Tier-1 service failures
+      3. nixos-rollback.sh sets default to generation N-1
+      4. Next reboot uses the working generation
+
+${CYAN}EXIT CODES${NC}
+    0   Rollback applied (or dry-run would apply)
+    1   Preflight failure (missing files, permissions)
+    2   No previous generation available (only one generation)
+    3   nixos-rebuild --rollback failed (with --rollback-now)
+
+${CYAN}FILES${NC}
+    ${LOADER_CONF}      systemd-boot loader configuration
+    ${ENTRIES_DIR}/     generation entry .conf files
+    ${LOGFILE}          action log (append-only)
+EOF
+}
+
+# =============================================================================
+# Logging
+# =============================================================================
+log() {
+    local level="$1"; shift
+    local msg="$*"
+    local timestamp
+    timestamp="$(date '+%Y-%m-%d %H:%M:%S')"
+    echo "${timestamp} [${level}] ${msg}" >> "${LOGFILE}"
+    logger -t "${SYSLOG_IDENT}" -p "user.${level}" "${msg}"
+
+    # Also print to stderr for ERROR/WARN, stdout for INFO
+    case "${level}" in
+        ERROR) echo >&2 "${RED}[ERROR]${NC} ${msg}" ;;
+        WARN)  echo >&2 "${YELLOW}[WARN]${NC}  ${msg}" ;;
+        INFO)  echo " ${GREEN}[INFO]${NC}  ${msg}" ;;
+    esac
+}
+
+info()  { log "INFO" "$@"; }
+warn()  { log "WARN" "$@"; }
+error() { log "ERROR" "$@"; }
+
+# =============================================================================
+# Preflight checks
+# =============================================================================
+preflight() {
+    # Must run as root (need to write to /boot), unless overridden for testing
+    if [ -z "${NIXOS_ROLLBACK_SKIP_ROOT_CHECK:-}" ] && [ "$(id -u)" -ne 0 ]; then
+        error "This script must be run as root (needs write access to /boot/loader)"
+        error "Set NIXOS_ROLLBACK_SKIP_ROOT_CHECK=1 for testing against mock paths."
+        exit 1
+    fi
+
+    # Directories and files
+    if [ ! -d "${ENTRIES_DIR}" ]; then
+        error "Boot entries directory not found: ${ENTRIES_DIR}"
+        exit 1
+    fi
+
+    if [ ! -f "${LOADER_CONF}" ]; then
+        error "Loader config not found: ${LOADER_CONF}"
+        exit 1
+    fi
+
+    if [ ! -r "${LOADER_CONF}" ]; then
+        error "Cannot read loader config: ${LOADER_CONF}"
+        exit 1
+    fi
+
+    # Check write access to /boot/loader (parent of loader.conf)
+    local loader_dir
+    loader_dir="$(dirname "${LOADER_CONF}")"
+    if [ ! -w "${loader_dir}" ]; then
+        error "Cannot write to ${loader_dir} (insufficient permissions)"
+        exit 1
+    fi
+
+    # Logfile directory must exist
+    local log_dir
+    log_dir="$(dirname "${LOGFILE}")"
+    if [ ! -d "${log_dir}" ]; then
+        warn "Log directory ${log_dir} does not exist, creating it"
+        mkdir -p "${log_dir}" 2>/dev/null || {
+            error "Cannot create log directory ${log_dir}"
+            exit 1
+        }
+    fi
+
+    # Check --rollback-now dependencies
+    if [ "${ROLLBACK_NOW}" = true ]; then
+        if ! command -v nixos-rebuild &>/dev/null; then
+            error "nixos-rebuild not found on PATH (required for --rollback-now)"
+            exit 1
+        fi
+    fi
+}
+
+# =============================================================================
+# Generation helpers
+# =============================================================================
+
+# get_current_default: reads the current default entry from loader.conf
+# Returns: "nixos-generation-N.conf" or empty string
+get_current_default() {
+    grep -E '^default\s+' "${LOADER_CONF}" 2>/dev/null \
+        | awk '{print $2}' \
+        || true
+}
+
+# extract_gen_number: extracts the numeric generation from a conf filename
+# Input:  "nixos-generation-367.conf"
+# Output: 367
+extract_gen_number() {
+    echo "$1" | sed 's/nixos-generation-//;s/\.conf//'
+}
+
+# get_all_gen_numbers: returns sorted list of generation numbers from entries dir
+get_all_gen_numbers() {
+    local -a gens=()
+    local f n
+    for f in "${ENTRIES_DIR}"/nixos-generation-*.conf; do
+        [ -f "${f}" ] || continue
+        n="$(basename "${f}" | sed 's/nixos-generation-//;s/\.conf//')"
+        gens+=("${n}")
+    done
+
+    if [ "${#gens[@]}" -eq 0 ]; then
+        return 1
+    fi
+
+    # Sort numerically and output
+    printf '%s\n' "${gens[@]}" | sort -n
+}
+
+# get_previous_gen: given current generation number, find the previous one
+# from the list of all available generations
+get_previous_gen() {
+    local current="$1"
+    shift
+    local -a gens=("$@")
+
+    local prev=""
+    local g
+    for g in "${gens[@]}"; do
+        if [ "${g}" -lt "${current}" ]; then
+            prev="${g}"
+        fi
+    done
+
+    if [ -z "${prev}" ]; then
+        return 1
+    fi
+    echo "${prev}"
+}
+
+# =============================================================================
+# Main rollback logic
+# =============================================================================
+do_rollback() {
+    # Step 1: Read current default
+    local current_entry
+    current_entry="$(get_current_default)"
+
+    if [ -z "${current_entry}" ]; then
+        error "No 'default' entry found in ${LOADER_CONF}"
+        error "Cannot determine current generation — aborting"
+        exit 1
+    fi
+
+    info "Current default boot entry: ${current_entry}"
+
+    # Step 2: Build sorted list of all available generations
+    local -a all_gens=()
+    local line
+    while IFS= read -r line; do
+        all_gens+=("${line}")
+    done < <(get_all_gen_numbers || true)
+
+    if [ "${#all_gens[@]}" -eq 0 ]; then
+        error "No NixOS generation .conf files found in ${ENTRIES_DIR}"
+        exit 1
+    fi
+
+    info "Available generations: ${all_gens[*]}"
+
+    # Step 3: Find current generation number
+    local current_gen
+    current_gen="$(extract_gen_number "${current_entry}")"
+
+    # Verify current_gen is a valid number
+    if ! [[ "${current_gen}" =~ ^[0-9]+$ ]]; then
+        error "Could not parse generation number from '${current_entry}'"
+        exit 1
+    fi
+
+    # Step 4: Find the previous generation
+    local prev_gen
+    prev_gen="$(get_previous_gen "${current_gen}" "${all_gens[@]}")" || {
+        error "No previous generation found before generation ${current_gen}"
+        error "This is the oldest available generation — cannot roll back further"
+        exit 2
+    }
+
+    local prev_entry="nixos-generation-${prev_gen}.conf"
+    local prev_conf_path="${ENTRIES_DIR}/${prev_entry}"
+
+    if [ ! -f "${prev_conf_path}" ]; then
+        error "Previous generation entry not found: ${prev_conf_path}"
+        error "The .conf file for generation ${prev_gen} is missing — cannot roll back"
+        exit 1
+    fi
+
+    info "Target rollback generation: ${prev_gen} → ${prev_entry}"
+
+    # Step 5: Apply the rollback
+    if [ "${DRY_RUN}" = true ]; then
+        echo ""
+        echo " ${CYAN}[DRY RUN]${NC} Would change ${LOADER_CONF}:"
+        echo "   ${YELLOW}-${NC} default ${current_entry}"
+        echo "   ${GREEN}+${NC} default ${prev_entry}"
+        echo ""
+        info "DRY RUN — no changes made"
+        exit 0
+    fi
+
+    # Write new default
+    # Use sed with a backup (.bak)
+    sed -i.bak "s/^default\s\+${current_entry}/default ${prev_entry}/" "${LOADER_CONF}"
+
+    # Verify the change was applied
+    local new_default
+    new_default="$(get_current_default)"
+    if [ "${new_default}" != "${prev_entry}" ]; then
+        error "Failed to set default boot entry to ${prev_entry}"
+        error "Current default is still: ${new_default}"
+        # Attempt to restore backup
+        if [ -f "${LOADER_CONF}.bak" ]; then
+            cp "${LOADER_CONF}.bak" "${LOADER_CONF}"
+            info "Restored backup from ${LOADER_CONF}.bak"
+        fi
+        exit 1
+    fi
+
+    info "Successfully set default boot entry to ${prev_entry} (generation ${prev_gen})"
+    info "Backup of previous config saved to ${LOADER_CONF}.bak"
+
+    # Step 6: Optionally run nixos-rebuild switch --rollback
+    if [ "${ROLLBACK_NOW}" = true ]; then
+        echo ""
+        info "Running nixos-rebuild switch --rollback for immediate effect..."
+        if nixos-rebuild switch --rollback 2>&1 | while IFS= read -r line; do
+            logger -t "${SYSLOG_IDENT}" "nixos-rebuild: ${line}"
+            echo "    ${line}"
+        done; then
+            info "nixos-rebuild switch --rollback completed successfully"
+        else
+            local rc=$?
+            error "nixos-rebuild switch --rollback failed with exit code ${rc}"
+            error "The boot default has been changed but the current system was NOT rolled back"
+            error "Reboot to apply the rollback"
+            exit 3
+        fi
+    fi
+
+    info "Rollback complete. Next boot will use generation ${prev_gen}."
+    if [ "${ROLLBACK_NOW}" = false ]; then
+        echo ""
+        echo " ${YELLOW}NOTE:${NC} The current running system is unchanged."
+        echo "       Reboot to boot into generation ${prev_gen}."
+        echo "       Or re-run with --rollback-now for immediate effect."
+    fi
+}
+
+# =============================================================================
+# Main
+# =============================================================================
+main() {
+    # Parse arguments
+    while [ $# -gt 0 ]; do
+        case "$1" in
+            --dry-run)
+                DRY_RUN=true
+                shift
+                ;;
+            --rollback-now)
+                ROLLBACK_NOW=true
+                shift
+                ;;
+            -h|--help)
+                usage
+                exit 0
+                ;;
+            *)
+                echo >&2 "Unknown option: $1"
+                echo >&2 "Use --help for usage information."
+                exit 1
+                ;;
+        esac
+    done
+
+    echo ""
+    echo " ${CYAN}═══ NixOS systemd-boot Rollback ═══${NC}"
+    echo ""
+
+    preflight
+
+    if [ "${DRY_RUN}" = true ]; then
+        info "DRY RUN mode — no changes will be made"
+    fi
+    if [ "${ROLLBACK_NOW}" = true ]; then
+        info "ROLLBACK NOW mode — will also run nixos-rebuild switch --rollback"
+    fi
+
+    echo ""
+    do_rollback
+}
+
+main "$@"

modules/nixos/services/rollback-sentinel.nix

@@ -0,0 +1,184 @@
+{ config, pkgs, lib, ... }:
+
+with lib;
+
+let
+  cfg = config.services.rollbackSentinel;
+
+  # ── Scripts ────────────────────────────────────────────────────────────────
+
+  # Sentinel check — verifies Tier-1 services are active after boot.
+  # Exits nonzero when any Tier-1 service is down, which triggers the rollback.
+  sentinelCheck = pkgs.writeShellScriptBin "sentinel-check.sh" ''
+    #!/usr/bin/env bash
+    set -euo pipefail
+
+    SYSLOG_IDENT="nixos-sentinel"
+    LOGFILE="/var/log/nixos-sentinel.log"
+
+    echo "=== NixOS Sentinel Check ==="
+    echo "Tier-1 services: ${builtins.toString cfg.tier1Services}"
+    echo "Tier-2 services: ${builtins.toString cfg.tier2Services}"
+
+    FAILED=0
+
+    # Check Tier-1 services — any failure means rollback
+    for svc in ${builtins.toString cfg.tier1Services}; do
+      if systemctl is-active --quiet "$svc" 2>/dev/null; then
+        echo "  [OK]  Tier-1: $svc"
+      else
+        echo "  [FAIL] Tier-1: $svc is NOT active"
+        logger -t "$SYSLOG_IDENT" -p user.err "Tier-1 FAILURE: $svc is not active"
+        FAILED=1
+      fi
+    done
+
+    # Check Tier-2 services — warn only
+    for svc in ${builtins.toString cfg.tier2Services}; do
+      if systemctl is-active --quiet "$svc" 2>/dev/null; then
+        echo "  [OK]  Tier-2: $svc"
+      else
+        echo "  [WARN] Tier-2: $svc is NOT active"
+        logger -t "$SYSLOG_IDENT" -p user.warn "Tier-2 WARNING: $svc is not active"
+      fi
+    done
+
+    echo "=== Sentinel result: $([ "$FAILED" -eq 0 ] && echo 'PASS' || echo 'FAIL') ==="
+    echo "$(date '+%Y-%m-%d %H:%M:%S') [INFO] sentinel $([ "$FAILED" -eq 0 ] && echo 'PASS' || echo 'FAIL')" >> "$LOGFILE"
+    exit $FAILED
+  '';
+
+  # Rollback script — package the companion shell script from this directory.
+  # Uses builtins.readFile to embed the content at evaluation time.
+  rollbackScript = pkgs.writeShellScriptBin "nixos-rollback.sh" (builtins.readFile ./nixos-rollback.sh);
+
+  # Resolve rollback flags from config
+  rollbackFlags =
+    if cfg.rollbackMode == "dry-run" then "--dry-run"
+    else if cfg.rollbackMode == "rollback-now" then "--rollback-now"
+    else "";
+
+in {
+  options.services.rollbackSentinel = {
+    enable = mkEnableOption "NixOS Rollback Sentinel — auto-rollback on critical service failure";
+
+    tier1Services = mkOption {
+      type = types.listOf types.str;
+      default = [ "sshd" "docker" "traefik" "authelia" ];
+      description = ''
+        Tier-1 services whose failure triggers an automatic systemd-boot rollback.
+        On boot, the sentinel waits ${cfg.bootDelay} seconds, then checks each
+        service. If ANY service in this list is inactive, it runs the rollback
+        script which sets the previous NixOS generation as the default boot entry.
+      '';
+    };
+
+    tier2Services = mkOption {
+      type = types.listOf types.str;
+      default = [
+        "gitea" "hermes" "ollama" "synapse" "nextcloud"
+        "vaultwarden" "wireguard" "homeassistant" "fail2ban"
+      ];
+      description = ''
+        Tier-2 services whose failure is logged as a warning but does NOT trigger
+        an automatic rollback. Useful for detecting non-critical service issues.
+      '';
+    };
+
+    tier3InfoServices = mkOption {
+      type = types.listOf types.str;
+      default = [
+        "act_runner" "syncthing" "restic" "fava"
+        "homer" "cups" "fstrim"
+      ];
+      description = ''
+        Tier-3 informational checks (log-only, no warning). These are services
+        that the sentinel will note the status of for diagnostics.
+      '';
+    };
+
+    bootDelay = mkOption {
+      type = types.str;
+      default = "120";
+      description = ''
+        Seconds to wait after multi-user.target before running the boot-time
+        sentinel check. This gives Tier-1 services time to start before
+        the sentinel decides they've failed.
+      '';
+    };
+
+    rollbackMode = mkOption {
+      type = types.enum [ "set-default" "rollback-now" "dry-run" ];
+      default = "set-default";
+      description = ''
+        Rollback strategy when Tier-1 failures are detected:
+        - set-default: Write the previous generation to loader.conf (next reboot).
+        - rollback-now: Also run nixos-rebuild switch --rollback for immediate fix.
+        - dry-run: Log what would happen but take no action (testing).
+      '';
+    };
+
+    enablePostRebuild = mkOption {
+      type = types.bool;
+      default = true;
+      description = ''
+        When enabled, the sentinel check runs after every nixos-rebuild switch
+        activation. If a newly deployed generation has Tier-1 failures, it
+        triggers rollback immediately.
+      '';
+    };
+  };
+
+  config = mkIf cfg.enable {
+    # ── Deploy scripts to PATH ───────────────────────────────────────────────
+    environment.systemPackages = [ sentinelCheck rollbackScript ];
+
+    # Ensure log directory exists
+    systemd.tmpfiles.rules = [
+      "d /var/log/nixos-sentinel 0755 root root -"
+    ];
+
+    # ── Boot-time sentinel service ───────────────────────────────────────────
+    # Runs after multi-user.target with a configurable delay, checks Tier-1
+    # services, and triggers rollback if any are down.
+    systemd.services.nixos-sentinel = {
+      description = "NixOS Boot Sentinel — check critical services, roll back on failure";
+      after = [ "network.target" "multi-user.target" ];
+      wants = [ "network.target" ];
+      wantedBy = [ "multi-user.target" ];
+
+      path = with pkgs; [ coreutils gawk gnused systemd ];
+
+      serviceConfig = {
+        Type = "oneshot";
+        RemainAfterExit = true;
+        ExecStartPre = "${pkgs.coreutils}/bin/sleep ${cfg.bootDelay}";
+        ExecStart = "${sentinelCheck}/bin/sentinel-check.sh";
+        ExecStartPost = "${rollbackScript}/bin/nixos-rollback.sh ${rollbackFlags}";
+      };
+    };
+
+    # ── Post-rebuild sentinel service (triggered by activation script) ──────
+    systemd.services.nixos-sentinel-rebuild = mkIf cfg.enablePostRebuild {
+      description = "NixOS Post-Rebuild Sentinel — check services after nixos-rebuild";
+      after = [ "network.target" ];
+
+      path = with pkgs; [ coreutils gawk gnused systemd ];
+
+      serviceConfig = {
+        Type = "oneshot";
+        ExecStart = "${sentinelCheck}/bin/sentinel-check.sh";
+        ExecStartPost = "${rollbackScript}/bin/nixos-rollback.sh ${rollbackFlags}";
+      };
+    };
+
+    # Activation script — fires after every nixos-rebuild switch
+    system.activationScripts.rollback-sentinel = mkIf cfg.enablePostRebuild ''
+      # Start the post-rebuild sentinel in the background.
+      # This runs on every activation (boot + nixos-rebuild). On boot the
+      # boot-time service handles it, so this is primarily for nixos-rebuild,
+      # but running twice is safe (idempotent rollback).
+      systemctl start nixos-sentinel-rebuild.service --no-block 2>/dev/null || true
+    '';
+  };
+}

users/ai-worker.nix

@@ -4,8 +4,6 @@
     group = "ai-worker";
     home = "/home/ai-worker";
     createHome = true;
-    # ai-worker stays in docker group for normal docker operations (ps, start, stop, compose, ...)
-    # Dangerous commands (exec, cp, commit) are blocked by a wrapper script.
     extraGroups = [ "docker" ];
     shell = pkgs.bashInteractive;
     openssh.authorizedKeys.keys = [
@@ -16,14 +14,17 @@
   };
   users.groups.ai-worker = {};
 
-  # Enable restricted AI worker SSH access
+  # Enable restricted AI worker SSH access for ollama benchmarking
-  # SECURITY: ai-worker is in docker group but docker commands are filtered:
+  # SECURITY: ai-worker can only:
-  #   ALLOWED: ps, images, logs, start, stop, restart, rm, rmi, pull, build, run, compose
+  #   - SSH into host from Hermes container
-  #   BLOCKED: exec, cp, commit, diff, export, import, load, save, attach, push
+  #   - Run docker commands (docker exec ollama ...) via docker group
-  # The filtering is done by a docker wrapper in ai-worker's PATH.
+  #   - Run specific security audit commands
+  #   - NO access to infra repo (no bind mount)
+  #   - NO sudo access (no nh, nixos-rebuild, nixpkgs-fmt, nix)
+  # WORKFLOW: SSH from Hermes container, run docker benchmarks, return and save results to /opt/data/ai-optimizer/
   services.aiWorkerAccess = true;
-
+  
-  # Restricted sudo for ai-worker - security checks only (not for docker)
+  # Restricted sudo for ai-worker - security checks only
   security.sudo.extraRules = [
     {
       users = [ "ai-worker" ];
@@ -68,6 +69,15 @@
           command = "/run/current-system/sw/bin/sshd -T";
           options = [ "NOPASSWD" ];
         }
+        # Docker service checks
+        {
+          command = "/run/current-system/sw/bin/docker ps";
+          options = [ "NOPASSWD" ];
+        }
+        {
+          command = "/run/current-system/sw/bin/docker inspect *";
+          options = [ "NOPASSWD" ];
+        }
         # Network diagnostics
         {
           command = "/run/current-system/sw/bin/ss -tlnp";