Carol — back to Apps ← Apps

Carolopedia

A friendly guide to Carol, her ecosystem, and the agents who built her.

📖 CarolopediaServicesProcess Monitoring
Process Monitoring

Process Monitoring

Service The self-healing watch over every process

📖About & Usage

About

Process Monitoring is Carol’s self‑healing watch over every scheduled process and service. Think of it as a tireless guardian that sits quietly in the background, watching for failures — and when something breaks, it doesn’t just raise an alarm; it tries to fix the problem automatically. This service exists so that the rest of Carol’s ecosystem stays reliable without waiting for a human to manually restart a stalled task or debug a crash.

Created by Hermione (Hermione) and assisted by Inspector (Inspector), Process Monitoring is currently being built — so while it’s not fully deployed yet, its design already promises to turn “thing went wrong” into “thing got fixed.” The goal is to make scheduled workflows and background services resilient by design, catching issues the moment they happen and applying pre‑defined or learned fixes.

Usage Patterns

Process Monitoring fires every time a scheduled process or running service experiences a failure, timeout, or unexpected exit. It continuously checks the heartbeat of every recurring task (like data syncs, report generation, or nightly cleanups) and every long‑running service (such as chat handlers or database connectors).

Concrete example:

  • A nightly database‑cleanup script fails because a temporary table is locked.
  • Process Monitoring detects the failure within seconds, logs the error, and — because the cleanup is a known safe‑to‑retry job — automatically restarts it after a short delay. If the restart succeeds, the system logs the recovery and moves on. If it fails again, it escalates to Hermione (Hermione) and the on‑call team.

Beyond reactive fixes, the service also watches for patterns like slow‑running processes or growing memory use, flagging potential issues before they become failures. Once fully built, Process Monitoring will learn from each incident to propose smarter, faster fixes — making Carol’s internal operations quietly, stubbornly self‑healing.

🏛Architecture

The Process Monitoring service is built following the agent-centric modular architecture of Carolverse. It leverages agile principles to build or modify software using distinct agent identities, each carrying out a specific activity. Its purpose is the self-healing watch: Hermione judges every process in the ecosystem for failure and files a fix when one breaks, with Inspector verifying that the watch itself stays honest.

View the full architecture →

🧱Blocks

Scheduled & Ongoing Processes · 132 processesTriggered & On-Demand Processes · 123 processesEmbedded Processes · 6 processesScheduled Run-Audit (Heartbeat DB) · Support · 0 droids

📚Recent initiatives

Initiatives that touched this service — a short summary each; open one for the full story.

CAROL-INI-1888-02: Admin Monitor: Visit-time as per-visitor stacked bar (grey/blue)
Convert the Visit-time chart from one aggregate bar into a stacked bar where each segment is one DISTINCT visitor time that day, identified by email or by anon:IP for anonymous vi\u2026
Orion · 2026-07-04 02:01
CAROL-INI-2189-01: al_watch_01 runtime guard counts its own Hermione launcher — watcher aborts every 60s tick
NARROW RUNTIME HOTFIX (NOT the INI-113 architecture redesign; that is separate, broad, still planned). The watcher already exists and is scheduled every 60s; it is non-functional\u2026
Orion · 2026-07-04 01:45
CAROL-INI-2095-01: Reduce sustained infrastructure CPU usage below 50% threshold
The infrastructure CPU health metric currently scores 4.7 out of 10, indicating that CPU usage has been breaching the 50% threshold more often than the 1% allowance target. This i\u2026
Orion · 2026-07-04 01:31
Browse all initiatives →

🛰️Updates

Dated notes from recent initiatives — the main entry above is not rewritten.

Fix2026-07-02

Process Monitoring now correctly auto-resumes dispatch when blocked count falls below the threshold, fixing a bug where Current Execution stayed empty and the 3-deep queue drifted.

Fix2026-06-29

Monitoring workflows now include dispatch-time re-validation of alarm premises, preventing unnecessary pipeline runs for stale or already-resolved issues.

👤Owner

Hermione · Process Monitor

🤝Supporting agents

Inspector · Watcher of the Watcher (Hermione Health Inspector)

🧩Apps

Apps owned by this service's team.

Hermione MonitorProcess Health ScorecardScheduled Processes