@delorenj/pjangler 1.3.7 → 1.4.3

package/templates/hermes-agent/template/.scripts/80-registry.sh

@@ -40,8 +40,10 @@ python3 - "$REGISTRY_FILE" "$AGENT_ID" "$REPO" "$ROLE" "$DISPLAY_NAME" \
   "$PLANE_WORKSPACE" "$PLANE_PROJECT_ID" "$(yaml_get plane.identifier)" \
   "$RUNTIME_REPO" "$HERMES_BIN" "$HERMES_AGENT_REPO" "$HERMES_RUNTIME_GIT_URL" \
   "$HERMES_RUNTIME_GIT_REF" "$HERMES_RUNTIME_GIT_SHA" "$FLEET_ENV" \
-  "hermes-${AGENT_ID}-gateway.service" "hermes-${AGENT_ID}-heartbeat.timer" <<'PYEOF'
+  "hermes-${AGENT_ID}-gateway.service" "hermes-${AGENT_ID}-heartbeat.timer" \
+  "$(yaml_get service_state.gateway)" "$(yaml_get service_state.heartbeat)" <<'PYEOF'
 import datetime
+import copy
 import errno
 import os
 import pathlib
@@ -56,11 +58,17 @@ except ImportError:
  slack_status, slack_team_id, slack_team_name, slack_user_id, slack_bot_id,
  slack_username, bloodbank_enabled, bloodbank_scope, bloodbank_target, plane_ws, plane_id,
  plane_ident, runtime_repo, hermes_bin, hermes_repo, hermes_git_url,
- hermes_git_ref, hermes_git_sha, fleet_env, gw, heartbeat) = sys.argv[1:33]
+ hermes_git_ref, hermes_git_sha, fleet_env, gw, heartbeat,
+ gateway_state, heartbeat_state) = sys.argv[1:35]
 p = pathlib.Path(path)
 if p.is_symlink():
     raise SystemExit(f"refusing to update registry symlink: {p}")
-data = yaml.safe_load(p.read_text()) or {"schema_version": 1, "agents": {}}
+data = yaml.safe_load(p.read_text(encoding="utf-8")) or {"schema_version": 1, "agents": {}}
+if not isinstance(data, dict):
+    raise SystemExit("fleet registry root must be a mapping")
+agents = data.setdefault("agents", {})
+if not isinstance(agents, dict):
+    raise SystemExit("fleet registry agents must be a mapping")
 if bloodbank_enabled == "":
     bloodbank_enabled_value = False
 elif bloodbank_enabled == "true":
@@ -69,7 +77,13 @@ elif bloodbank_enabled == "false":
     bloodbank_enabled_value = False
 else:
     raise SystemExit("bloodbank.enabled must be the strict YAML boolean true or false")
-data.setdefault("agents", {})[agent_id] = {
+existing = agents.get(agent_id, {})
+if not isinstance(existing, dict):
+    raise SystemExit(f"fleet registry entry for {agent_id} must be a mapping")
+provisioned_at = existing.get("provisioned_at")
+if not isinstance(provisioned_at, str) or not provisioned_at:
+    provisioned_at = datetime.datetime.now(datetime.timezone.utc).isoformat().replace("+00:00", "Z")
+managed = {
   "repo": repo, "role": role, "display_name": display,
   "project_path": project, "role_dir": role_dir,
   "profile_name": profile,
@@ -101,9 +115,31 @@ data.setdefault("agents", {})[agent_id] = {
     "git_sha": hermes_git_sha,
     "fleet_env": fleet_env,
   },
-  "systemd": {"gateway_unit": gw, "heartbeat_timer": heartbeat},
-  "provisioned_at": datetime.datetime.now(datetime.timezone.utc).isoformat().replace("+00:00", "Z"),
+  "systemd": {
+    "gateway_unit": gw,
+    "heartbeat_timer": heartbeat,
+    "gateway_state": gateway_state,
+    "heartbeat_state": heartbeat_state,
+  },
+  "provisioned_at": provisioned_at,
 }
+
+def merge_managed(current, update):
+    result = copy.deepcopy(current)
+    for key, value in update.items():
+        if isinstance(value, dict) and isinstance(result.get(key), dict):
+            result[key] = merge_managed(result[key], value)
+        else:
+            result[key] = copy.deepcopy(value)
+    return result
+
+entry = merge_managed(existing, managed)
+# This is retired managed schema, not extension metadata.  Keeping it would
+# falsely advertise a second per-agent Bloodbank execution path.
+systemd = entry.get("systemd")
+if isinstance(systemd, dict):
+    systemd.pop("consumer_unit", None)
+agents[agent_id] = entry
 rendered = yaml.safe_dump(data, sort_keys=False)
 p.parent.mkdir(parents=True, exist_ok=True)
 

package/templates/hermes-agent/template/.scripts/99-summary.sh

@@ -1,32 +1,85 @@
 #!/usr/bin/env bash
-# Print the final summary and next steps.
+# Print a reconciliation summary derived from durable role/service state.
 # shellcheck source=_lib.sh
 source "$(dirname "$0")/_lib.sh"
 load_role_env
 
 GW_UNIT="hermes-${AGENT_ID}-gateway.service"
 HB_TIMER="hermes-${AGENT_ID}-heartbeat.timer"
+GW_STATE="$(yaml_get service_state.gateway)"
+HB_STATE="$(yaml_get service_state.heartbeat)"
+TELEGRAM_STATE="$(yaml_get telegram.provisioning_status)"
+SLACK_STATE="$(yaml_get slack.provisioning_status)"
+RECONCILE_STATE="$(yaml_get reconcile.enabled)"
+RECONCILE_OPT_OUT="$(yaml_get reconcile.explicit_opt_out)"
+BOARD_ID="$(yaml_get ticket_provider.board_id)"
+
+# Reconcile durable claims with live unit truth at presentation time.  A stale
+# role.yaml must never turn an exited launcher or failed heartbeat into an
+# operational summary.
+if [[ "$HB_STATE" == "active" || "$GW_STATE" =~ ^(active|deferred)$ ]]; then
+  if systemd_user_available; then
+    if [[ "$HB_STATE" == "active" ]]; then
+      hb_live="$(systemd_timer_health_snapshot "$HB_TIMER" "hermes-${AGENT_ID}-heartbeat.service")"
+      [[ "$hb_live" == ok\|* ]] || HB_STATE="error (stale active claim)"
+    fi
+    if [[ "$GW_STATE" == "active" ]]; then
+      gw_live="$(systemd_service_health_snapshot "$GW_UNIT" running)"
+      [[ "$gw_live" == ok\|* ]] || GW_STATE="error (stale active claim)"
+    elif [[ "$GW_STATE" == "deferred" ]]; then
+      gw_live="$(systemd_gateway_deferred_snapshot "$GW_UNIT")"
+      [[ "$gw_live" == "ok|deferred" ]] || GW_STATE="error (stale deferred claim)"
+    fi
+  else
+    [[ "$HB_STATE" != "active" ]] || HB_STATE="unverified (user manager unavailable)"
+    [[ "$GW_STATE" != "active" && "$GW_STATE" != "deferred" ]] \
+      || GW_STATE="unverified (user manager unavailable)"
+  fi
+fi
+
+case "$HB_STATE:$GW_STATE" in
+  active:active) MODE="OPERATIONAL" ;;
+  active:deferred) MODE="OPERATIONAL_WITH_GATEWAY_DEFERRED" ;;
+  installed:installed|installed:deferred) MODE="INSTALLED_NOT_ACTIVE" ;;
+  *) MODE="INCOMPLETE" ;;
+esac
 
 cat >&2 <<EOF
 
-╭─ Provisioned: $AGENT_ID ───────────────────────────────────────────────╮
+╭─ Hermes deployment: $AGENT_ID ───────────────────────────────────────╮
 │
-│  Talk:        @$BOT_HANDLE  (Telegram DM)
-│  Shell:       $ROLE_DIR/hermes chat "status"
-│  Board:       $PLANE_BASE/$PLANE_WORKSPACE/projects/$(cat "$ROLE_DIR/.scripts/.plane-project-id" 2>/dev/null || echo "(skipped)")
-│  Runtime:     $ROLE_DIR/runtime   (pure-local, ignored)
-│  Hermes bin:  $HERMES_BIN
-│  Fleet env:   $FLEET_ENV
+│  Mode:           $MODE
+│  Gateway:        $GW_STATE  ($GW_UNIT)
+│  Heartbeat:      $HB_STATE  ($HB_TIMER)
+│  Reconciliation: $RECONCILE_STATE  (explicit opt-out: ${RECONCILE_OPT_OUT:-false})
+│  Telegram:       $TELEGRAM_STATE
+│  Slack:          $SLACK_STATE
+│  Board:          ${BOARD_ID:-deferred}
+│  Runtime:        $ROLE_DIR/runtime   (pure-local, ignored)
+│  Hermes bin:     $HERMES_BIN
+│  Fleet env:      $FLEET_ENV
 │
-│  Start fleet daemons:
-│    systemctl --user start $GW_UNIT
-│    systemctl --user start $HB_TIMER
-│
-│  Tail logs:
-│    journalctl --user -fu $GW_UNIT
+EOF
+
+if [[ "$TELEGRAM_STATE" == "verified" ]]; then
+  printf '│  Talk:           @%s  (Telegram DM)\n' "$BOT_HANDLE" >&2
+fi
+if [[ "$SLACK_STATE" == "verified" ]]; then
+  printf '│  Slack bot:      %s\n' "$(yaml_get slack.bot_username)" >&2
+fi
+printf '│  Shell:          %s/hermes chat "status"\n' "$ROLE_DIR" >&2
+
+if [[ "$GW_STATE" == "deferred" ]]; then
+  cat >&2 <<EOF
 │
-│  Fleet status:
-│    python3 -c "import yaml,pathlib; print(yaml.safe_dump(yaml.safe_load(pathlib.Path('$REGISTRY_FILE').read_text())['agents']))"
+│  Gateway activation is deferred. Store and verify a dedicated Telegram or
+│  Slack credential through the provisioner; do not start $GW_UNIT manually.
+EOF
+elif [[ "$GW_STATE" == "active" ]]; then
+  printf '│  Gateway log:    journalctl --user -fu %s\n' "$GW_UNIT" >&2
+fi
+
+cat >&2 <<'EOF'
 │
 ╰────────────────────────────────────────────────────────────────────────╯