Carolopedia
A friendly guide to Carol, her ecosystem, and the agents who built her.
📖About & Usage
About
Quality Management is Carolverse’s continuous-improvement engine, owned by Prometheus. While Hermione the Process Monitor keeps day-to-day operations humming by healing issues as they appear, Quality Management takes a longer view: it sifts through closed Carol Initiatives, failed reviews from Test Runner, and past incident records to spot repeating problems. It then scores each service’s quality maturity and files concrete improvement tasks that the build pipeline can act on—closing the loop through the RSI Dashboard Dashboard (where metrics feed scoreboards that spark new improvements).
Think of it as a coach who watches the game tape so every service keeps getting better. Quality here is treated as overhead a service pays for: every service gets one quality metric free of charge. If a service owner wants extra metrics, they must request them. Until owners are fully equipped to make those calls, Orion (with support from Ninad) backfills that decision-making. This service is currently being built.
Usage Patterns
Quality Management fires automatically after significant events that signal room for learning. For example, imagine a customer-facing microservice that fails three deployment reviews in a single month because of a recurring configuration drift. The Quality Management service would mine those closed review failures, detect the pattern, and lower that service’s quality maturity score for deployment reliability. It would then file an improvement initiative in Carol Initiatives targeting the drift, which enters the build pipeline—teams can see the initiative, link it to a metric, and track its progress on the Quality Scorecard.
In practice, the service’s owner would notice the dip on the scorecard and might decide to request an additional quality metric specifically for configuration hygiene. Until they’re ready to make that choice, Orion and Ninad temporarily shepherd the decision. Over time, this loop—incident → pattern discovery → initiative → metric—helps every Carolverse service become more resilient without anyone having to dig through logs manually.
🏛Architecture
Architecture not yet documented for this service.
Define architecture →🧱Blocks by trackwhat’s a track? →
📓The words this service uses (13)
Each is defined once in the dictionary and explained on its own page — this service does not restate them.
📚Recent initiatives
Initiatives that touched this service — a short summary each; open one for the full story.
🛰️Updates
Dated notes from recent initiatives — the main entry above is not rewritten.
On 2026-05-09, all ~470 automated checks pass again after stale assertions were updated to match the current estate. This resolves the recurring 69–73 failures in Quality Management.
Regression tests are being added for the three Albus bypass lane fixes (CAROL-INI-3402, CAROL-INI-3406, CAROL-INI-3408) to ensure they remain effective. This improves coverage for Quality Management.
A ruling on 2026-07-23 mandates that when Albus's diagnosis finds non-compliance with the cookbook, it must file a redirect to Orion instead of attempting autonomous self-heal, affecting the non-compliance handling process of Quality Management.
As of 2026-07-23, per Ninad ruling, the Escalation Queue doctrine is corrected: it now contains only operator-attention items (rsi-diagnosis-abandoned or escalated). Workable blocked items that RSI will handle appear in the Dispatch Queue as next-in-line, amending cookbook 916. Quality Management
The Dispatch Queue card on the Monitor now correctly shows the next blocked initiative eligible for diagnosis in RSI mode, instead of showing raw sprint-lane items. Quality Management
Improved error handling in the troubleshooter avoids immediate failure on transient LLM errors, enhancing the reliability of Quality Management.
The self-heal loop now enforces a hard cap of three diagnosis attempts per blocked initiative, preventing excessive spend on dead-end diagnoses. This change reduces waste and aligns with doctrine, affecting Quality Management.
Fixed a bug where Albus's diagnosis was reading poisoned evidence due to loop-spam, causing false 'no root cause' verdicts. This improves diagnosis reliability. Quality Management
The self-heal loop no longer stops after a fixed cap of 3 diagnoses; it now continues diagnosing while learning and stops only on a plateau with a hard backstop. This prevents blind repetition into the same wall.
Quality Management is no longer a standalone service; it is now a block under Governance, owned by Clara, after Orion's ownership was revoked by Ninad.
Fixed the RSI diagnosis cap logic so families exceeding the completed-diagnoses limit are correctly tagged `rsi-diagnosis-abandoned` instead of being routed to the forbidden `parked` status, which was causing an infinite loop and wasting DeepSeek calls every minute.
Fixed a for-X title template bug that caused duplicate diagnosis filings for targets 999901314, 999901475, and 999900505. The sprint backlog now correctly deduplicates these entries Quality Management.
A bug was fixed where the self-heal system Quality Management would retrigger into the same failure after receiving a generic remedy, because it ignored the concreteness tag; now it escalates empty advice instead.
Effective 2026-07-04, the RSI loop picker now routes the oldest parked family into retrospective meta-diagnosis instead of indefinite parking, resolving the issue where the 3-strikes meta-diagnosis had never fired due to parking interception. This affects 9 currently parked families in Quality Management.
The enforcement mechanism for the fix-the-pipeline objective has been reworked: blocked work will not be retriggered until a verified pipeline fix ships, and the success measure is now durable-pipeline-fix. This supersedes prior rules from CAROL-INI-2275 and 2311. See Quality Management for cookbook 425 rewrite.
Quality Management now has a core quality RSI loop with a per-service Process Health metric that scores scheduled droid run-audit success. Historical snapshots are stored in [[data/quality_rsi.db]].
👤Owner
Prometheus · Head of Quality Management🧩Apps
Apps owned by this service's team.
Quality ScorecardRSI DashboardSentinel Monitor