# Cockpit Current State Updated: 2026-05-04 14:50 Asia/Kuwait ## Operating mode Rocky/Hermes is the user-facing chief orchestrator. OpenClaw is now connected as an internal worker path through a bounded file-bridge loop. The loop is safe-by-default: dry-run/report-only, no secrets, no service changes, no deletion, no trading execution. ## Closed workload loop Active scheduled jobs: 1. Daily Chief Orchestrator Report - Job id: a325b28023e0 - Schedule: 07:30 daily - Delivery: Ali Telegram chat - Role: concise human-facing daily summary and approval queue. 2. Autonomous OpenClaw Cockpit Loop - Job id: f89a47fb2d24 - Schedule: every 4 hours - Delivery: local only - Script: ~/.hermes/scripts/openclaw_autonomous_loop.py - Role: collect health, attempt bounded OpenClaw worker job unless backoff is active, update state/reports, mirror non-secret cockpit files to ROG. ## OpenClaw worker status OpenClaw direct model-run is wired into the loop but currently unstable under long prompts / possible subscription or fallback limits. Latest classification: backoff-active after timeout/fallback issue. Observed failure detail: - Router internal timeout hit after 120s. - stderr included Ollama fallback unavailable at http://127.0.0.1:11434. - This may be caused by Codex/OpenAI subscription/rate limits, auth/session pressure, or OpenClaw fallback routing. Current mitigation: - Do not hammer OpenClaw repeatedly. - Backoff for 3 hours after timeout/rate/fallback failures. - Hermes/tool-first reporting remains reliable. - OpenClaw will be retried automatically by the 4-hour loop. ## Next build priority Visual cockpit dashboard MVP: - local first - self-updating from cockpit state/reports/bridge files - overview cards and drill-down links - approval queue - service pipeline status - hosted on ROG LAN/private access first - no public exposure/tunnel/port-forward without Ali approval ## Approval gates remain Requires Ali approval: - paid API/subscription changes - public exposure / DNS / tunnel / port forward - credential copying/rotation/exposure - destructive data operations - production service changes - live trading - XRP policy changes ## Hermes approval mode Updated: 2026-05-04 Hermes approval mode has been set to `smart` in `/Users/alharbi/.hermes/config.yaml`. Meaning: - low-risk commands may be auto-approved by Hermes auxiliary judgment - high-risk/destructive/public/paid/credential/production/live-trading/XRP actions remain gated by Ali approval policy - gradual movement toward YOLO/off will be considered only after the cockpit, backup, and approval/audit loop prove reliable ## 2026-05-04 continuation after storage recovery Ali freed Mac storage by safely removing downloaded update data and photos. Verified current Mac disk: `/` and Data have ~86GiB available; Data volume ~58% used. Dashboard build continued and is live on ROG with pages: - `/` overview - `/services.html` - `/agents.html` - `/approvals.html` - `/pipeline.html` - `/reports.html` - `/basic.html` fallback OpenClaw usage status: direct OpenClaw Codex route produced `OPENCLAW_OK` in stdout, but the CLI failed to exit cleanly before timeout and emitted Ollama fallback-unavailable warnings. Router was patched to treat useful stdout + timeout as a partial result instead of total failure while preserving stderr for review. Continue bounded OpenClaw use with backoff; do not hammer subscription. ## 2026-05-06 Symphony-style worker queue Rocky is now the default decider for the Symphony-style cockpit worker queue. Ali should not need to manage ticket churn. Internal tasks can be created, merged, prioritized, routed, reviewed, or marked stale by Rocky without asking unless a hard approval gate is crossed. Added cockpit artifacts: - `SYMPHONY_QUEUE_POLICY.md` - `tasks/QUEUE.md` - `state/symphony-queue.json` - `scripts/symphony-decider.py` - dashboard page `/queue.html` Anti-churn defaults: max 6 worker turns, max 2 review/fix cycles, same-group concurrency 1, then Rocky must accept, narrow, or block with a short reason.
=== timestamp ===
2026-06-21 12:27:20 +03 +0300
=== host ===
Static hostname: rog
Icon name: computer-laptop
Chassis: laptop 💻
Machine ID: 1b06c820d83a4a95bbc9587bfb191699
Boot ID: a0b2f31bd1c34b4cbc19a01f8e5e2aab
Operating System: Ubuntu 24.04.4 LTS
Kernel: Linux 6.8.0-124-generic
Architecture: x86-64
Hardware Vendor: ASUSTeK Computer Inc.
Hardware Model: G74Sx
Firmware Version: G74Sx.202
Firmware Date: Mon 2011-08-15
Firmware Age: 14y 10month 6d
=== load/memory ===
12:27:21 up 8 days, 2:03, 0 user, load average: 1.27, 1.35, 1.67
total used free shared buff/cache available
Mem: 11Gi 7.2Gi 384Mi 118Mi 4.5Gi 4.5Gi
Swap: 4.0Gi 4.0Gi 248Ki
=== disks ===
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/ubuntu--vg-ubuntu--lv 914G 132G 744G 15% /
/dev/sdb1 916G 16G 855G 2% /data
=== docker ===
NAMES STATUS IMAGE
qbittorrent-mam Up 8 days 065792d2b11f
radarr Up 8 days lscr.io/linuxserver/radarr:latest
qbittorrent Up 8 days lscr.io/linuxserver/qbittorrent:latest
prowlarr Up 8 days lscr.io/linuxserver/prowlarr:latest
sonarr Up 8 days lscr.io/linuxserver/sonarr:latest
gluetun Up 8 days (healthy) qmcgaw/gluetun:latest
scrypted Up 8 days koush/scrypted:latest
go2rtc Up 8 days alexxit/go2rtc:latest
omada-controller Up 8 days (healthy) mbentley/omada-controller:6.2.10.17
honcho-api-1 Up 8 days (healthy) honcho-api
honcho-deriver-1 Up 8 days honcho-deriver
honcho-redis-1 Up 8 days (healthy) redis:8.2
honcho-database-1 Up 8 days (healthy) pgvector/pgvector:pg15
homeassistant Up 8 days ghcr.io/home-assistant/home-assistant:stable
moh-scheduler-preview Up 8 days (healthy) nginx:1.27-alpine
alarian-alsarraf-pilot Up 8 days node:22-alpine
open-webui Up 8 days (healthy) ghcr.io/open-webui/open-webui:main
adguardhome Up 8 days adguard/adguardhome:latest
glance Up 8 days glanceapp/glance:latest
paperless-ngx Up 8 days (healthy) ghcr.io/paperless-ngx/paperless-ngx:latest
paperless-db Up 8 days postgres:15-alpine
paperless-redis Up 8 days redis:7-alpine
paperless-gotenberg Up 8 days gotenberg/gotenberg:8
paperless-tika Up 8 days apache/tika:latest
scrutiny Up 8 days ghcr.io/analogj/scrutiny:master-omnibus
dozzle Up 8 days amir20/dozzle:latest
searxng Up 8 days searxng/searxng:latest
kiwix-serve Up 8 days ghcr.io/kiwix/kiwix-serve:latest
overseerr Up 8 days lscr.io/linuxserver/overseerr:latest
n8n Up 8 days n8nio/n8n:latest
homepage Up 8 days (healthy) ghcr.io/gethomepage/homepage:latest
calibre-web Up 8 days lscr.io/linuxserver/calibre-web:latest
calibre Up 8 days lscr.io/linuxserver/calibre:latest
mousesearch Up 8 days sevenlayercookie/mousesearch:latest
homebridge Up 8 days homebridge/homebridge:latest
flaresolverr Up 8 days ghcr.io/flaresolverr/flaresolverr:latest
=== binaries ===
Hermes Agent v0.12.0 (2026.4.30)
Project: /home/rog/.hermes/hermes-agent
Python: 3.11.15
OpenAI SDK: 2.32.0
Up to date
/usr/bin/openclaw
v22.22.0
Python 3.12.3
=== user services matching hermes/openclaw ===
openclaw-node.service loaded active running OpenClaw Node Host (v2026.4.14)
=== openclaw node status ===
● openclaw-node.service - OpenClaw Node Host (v2026.4.14)
Loaded: loaded (/home/rog/.config/systemd/user/openclaw-node.service; enabled; preset: enabled)
Active: active (running) since Sun 2026-06-21 12:25:12 +03; 2min 11s ago
Main PID: 38176 (openclaw-node)
Tasks: 11 (limit: 14223)
Memory: 229.7M (peak: 346.4M)
CPU: 10.293s
CGroup: /user.slice/user-1000.slice/[email protected]/app.slice/openclaw-node.service
└─38176 openclaw-node
Jun 21 12:25:12 rog systemd[1611]: openclaw-node.service: Scheduled restart job, restart counter is at 5457.
Jun 21 12:25:12 rog systemd[1611]: Started openclaw-node.service - OpenClaw Node Host (v2026.4.14).