The Platform
18 services across 5 intelligence layers and 4 hardening domains.
The canonical REV3 architecture organizes the factory into 5 explicit layers. Advisory intelligence is always subordinate to the authoritative control plane. No advisory module may replace governance.
Frozen baseline, ADRs, diagrams, repository conventions
Authoritative state, policy, task contracts, evidence, release authority
Recommendations, safety, planning, economics, learning — all subordinate to control plane
Control Tower, Dashboard, bounded execution workers
Observability, scenario testing, deployment, verification
The foundational intelligence layer. Every advisory service depends on the cognition fabric for knowledge retrieval, inference tracking, and retrieval trace auditing.
CognitionContextStore
12 knowledge source kinds — specs, decisions, risks, postmortems, policies, tasks, evidence, architecture, release, governance, learning, economic
ArtifactIndexRegistry
19 discovery rules scanning docs, operations, readiness, policy-packs, and test artifacts with freshness tracking and stale detection
InferenceLedger
SHA-256 input/output hashing, token counting, latency tracking, model lineage refs for every AI invocation
RetrievalTraceLedger
Session context, rerank basis, strategy tracking (semantic/keyword/hybrid), result persistence
CognitionEngine
Embed → rerank → search pipeline with pluggable mock/live mode and RTX 3090 integration
Planning Memory
Analog retrieval from past project outcomes with confidence scoring and source refs
Scenario Planning
3 strategies generated per scope — aggressive, balanced, conservative — with risk calibration
Plan Comparison
7-dimension component diff: sprints, effort, risk, capacity, critical path, dependencies, tasks
Effort Estimation
Complexity-based estimation from lines changed, files touched, tests, schema changes
Plan Calibration
30% weighted prior adjustment from estimated-vs-actual outcome deltas
Sprint Composition
Priority-sorted capacity fitting with utilization tracking and risk posture
Cost Attribution
Per-task and per-program cost tracking across 6 categories: execution, verification, coordination, rework, infrastructure, governance
Cost-of-Delay
Urgency scoring with delay impact classification (LOW/MEDIUM/HIGH/CRITICAL) and portfolio ranking effects
Program ROI
Risk-adjusted return, marginal utility scoring, payback sprint estimation
Economic Forecast
Forecast-vs-actual variance tracking with OVER/UNDER/MATCH direction detection
Exception Handling
Auto-escalation for BLOCK severity, resolution workflow with actor/note tracking
Capability Graph
Actor profiles (human/agent/team/service) with skill taxonomy and availability tracking
Skill Matching
Coverage scoring and gap detection — NONE/PARTIAL/MISSING severity per required skill
Bottleneck Prediction
Queue saturation detection from review load — NONE/EMERGING/ACTIVE/CRITICAL severity
Load Balancing
Overload/underload detection with transfer suggestions across actors
Coordination Risk
Cross-project shared actor/resource conflict detection with auto-escalation for CRITICAL
Learning Corpus
7 source kinds ingested: postmortem, closeout, evidence, alert, incident, decision, forecast error
Pattern Mining
Tag-based recurring failure detection with support count and LOW/MEDIUM/HIGH confidence
Policy Recommendation
Advisory policy deltas with evidence lineage — PROPOSED/ACCEPTED/REJECTED/EXPIRED lifecycle
Estimation Calibration
Prior adjustment from real outcomes with sample size tracking
Lesson Promotion
CANDIDATE → PROMOTED → ACTIVE → RETIRED lifecycle with max-age retention policy
Execution
Incident records, lineage manifests, override receipts (RECOVERY/BREAK_GLASS/ADMIN_CORRECTION), dispatch replay
Governance
Exception records (TEMPORARY/EMERGENCY/TRANSITIONAL/WAIVER), policy bundles, completeness scoring, debt tracking
Release
Strategy advisory, cohort analytics, runbook quality scoring with drift detection
Compliance
Coverage mapping, waiver risk analysis, continuous posture scoring, evidence lineage with tamper detection
See it in action
18 services. 5 intelligence layers. 10/10 maturity.