NOTQIN — OT Track

NOTQIN Core

5 min read

NOTQIN / Smartex

Cross-project propagation

NOTQIN Replay Fixture Discipline Propagation Spec 2026-06-30 adapts SijiLab replay-fixture discipline to one demo-critical telemetry path.

Deployment propagation

NOTQIN Demo Deployment Gate Propagation Spec 2026-06-30 adapts Wellys-PV deployment-gate findings to a lightweight NOTQIN demo/pilot gate.

Frontend safety propagation

NOTQIN Frontend Demo Safety Gate Propagation Spec 2026-06-30 adapts Wellys-PV frontend safety lessons to the selected NOTQIN demo path.

Current execution spec

NOTQIN Execution Spec 2026-06-29 drives the July-August repo work: event-derived prospect focus, first outreach assets, demo readiness, and the narrow-wedge boundary.

⚠️ 2026-06-28 audit: Repo has been consolidated into C:\Users\21260\IdeaProjects\smartex monorepo. Old path C:\Users\21260\IdeaProjects\notqin is abandoned.

1. Project Summary

A full-stack Industrial Intelligence platform — 18 apps + 10 services + 13 packages in a monorepo. Real-time IoT telemetry (36 machine types, 6 sectors), MQTT, Apache Flink stream processing, InfluxDB 3 Core time-series, 13 Grafana dashboards, 48-tool Claude AI analyst, CBAM carbon engine, factory MES (NestJS), DataOps orchestrator, edge pilot kit, and Kubernetes deployment with 53 manifests.

2. Architecture

Monorepo Structure

C:\Users\21260\IdeaProjects\smartex/
├── apps/                      18 applications
│   ├── analyst-service/       Python FastAPI — 48-tool NL Q&A, SSE, multi-sector
│   ├── anomaly-service/       Python FastAPI — 96 rules, MQTT consumer, Redis cache
│   ├── state-service/         Python FastAPI — digital twin, sub-ms reads
│   ├── factory-api/           NestJS + Prisma — MES/MOM (batches, materials, quality, planning)
│   ├── factory-app/           React + Vite — customer dashboard (i18n FR/EN/AR, AI Assistant)
│   ├── forge-api/             NestJS — provisioner management, plugin registry
│   ├── forge-app/             React — internal ops console
│   ├── mcp-server/            TypeScript — MCP proxy (120+ factory tools)
│   ├── orchestrate-service/   Python — DataOps engine (40+ connectors, 29 pipeline configs)
│   ├── ingestion-service/     Python — pluggable OT adapters (OPC UA, Modbus, S7, Sparkplug)
│   ├── provisioner-service/   TypeScript/Python — auto-provision droplet in ~5 min
│   ├── site-simulator/        Python — per-customer site simulation
│   └── ...
├── services/                  10 notqin-* services
│   ├── notqin-mcp-server/     Python MCP (stdio transport done, SSE deferred)
│   ├── notqin-spb-validator/  Rust — Sparkplug B validator + SIEM
│   │   └── ... (context-graph, cra-attest, edge-updater, euro-map-bridge, etc.)
├── packages/                  13 shared packages
│   ├── cbam-engine/           CBAM Phases 1–7, EF resolver, CN-code resolver
│   ├── agent-core/            AI agent infrastructure
│   ├── sector-packs/          ~28 sector packs (aeronautics, automotive, cement, steel, textile, agri, pharma, mining)
│   ├── knowledge-spine/       Per-customer pgvector RAG
│   └── ...
└── k8s/                       53 Kubernetes manifests (Kustomize, RBAC, Loki, AlertManager, cert-manager)

Data Pipeline (Main)

ESP32 Sensors / S7 PLC / OPC UA / Modbus TCP
  ↓ MQTT (Mosquitto TLS 8883 or EMQX HA cluster)
Apache NiFi 2.8 (route/filter/transform)
  ↓
Apache Flink 2.2.1 (BearingAnomalyJob, 10s tumbling window, EWMA)
  ↓
InfluxDB 3 Core (multi-DB: telemetry / telemetry_1m / telemetry_1h, SQL queries, Python downsampling jobs)
  ↓
Grafana 12.3 (13 dashboards + Unified Alerting rules)
  ↓
Claude AI Analyst (48 tools: telemetry trends, incident investigation, evidence graph, AAS, energy)

Supporting Services

ServiceStackPurpose
Anomaly ServicePython FastAPI96 rules across 23 machine types, MQTT consumer, 60s cooldown
State ServicePython FastAPI + Redis 8.8Digital twin, sub-ms fleet status
Analyst ServicePython FastAPI + Claude API48 tools, SSE streaming, incident investigation
ML ShadowPython Random ForestShadow inference pipeline, MQTT → InfluxDB
Telegraf 1.33GoMQTT → InfluxDB + Kafka bridge
InfluxDB JobsPython 3.12Downsampling (1m/1h), OEE computation

3. Current Status

Last commit: Jun 20, 2026 (8 days ago) — fix(deploy): complete authenticated evidence bridge Active branch: master Total commits: 436

Scale

MetricValue
Apps18
Services10
Packages13
Machine types (code)36 across 6 sectors
Machine types (deployed in compose)23 across 4 sectors
Sectors in compose4 (aeronautics, textile, agri-food, cement)
Sectors in code6 (+ steel, automotive)
Anomaly rules96
Analyst tools48
Grafana dashboards13 + alerting rules
Kubernetes manifests53
Docker Compose files14
Pipeline configs29
Test files268

Version Changes (since Apr 2026 vault note)

ComponentOldNew
InfluxDB2.73 Core (major upgrade — Flux replaced by SQL + Python jobs)
Grafana11.312.3
Redis78.8
Flink—2.2.1
NiFi—2.8.0
Prometheus—3.9.0
Telegraf—1.33

4. What Is Built

Core Infrastructure

  • ✅ Docker Compose full stack (22 services, health checks)
  • ✅ MQTT pub/sub with TLS (Mosquitto 8883 + WS 8084)
  • ✅ EMQX HA overlay (3-node cluster + HAProxy)
  • ✅ Kafka 4.3 KRaft overlay (dual-write Telegraf, Kafka UI)
  • ✅ InfluxDB 3 Core (3 databases, Python downsampling jobs, OEE)
  • ✅ Grafana provisioning (13 dashboards, alerting rules, datasources)
  • ✅ Prometheus scraping (Telegraf, Flink, anomaly, state, analyst)
  • ✅ Kubernetes manifests (53 files, Kustomize, HPA, Loki, AlertManager)
  • ✅ NiFi canvas (MQTT → route → filter → publish)

Application Services

  • ✅ Analyst Service: 48 tools, Dockerfile deployed, SSE streaming
  • ✅ Anomaly Service: 96 rules, MQTT subscriber, Prometheus metrics
  • ✅ State Service: Redis digital twin, FastAPI, sub-ms lookups
  • ✅ ML Shadow: Random Forest inference, MQTT → InfluxDB
  • ✅ Flow-server: Directed graph execution engine
  • ✅ Provisioner Service: auto-deploy droplet in ~4 min
  • ✅ Factory API: NestJS MES/MOM (42 Prisma models)
  • ✅ Factory App: React customer dashboard (i18n FR/EN/AR)
  • ✅ MCP Server: 120+ factory tools
  • ✅ Ingestion Service: 5 adapters with tests (OPC UA, Modbus TCP, Sparkplug B, CSV/DB, retrofit)

Simulator

  • ✅ 6 sectors, 36 machine types in code
  • ✅ 4 sectors deployed in main compose
  • ✅ 1Hz realistic telemetry per machine
  • ✅ Steel + Automotive sectors: code complete, not yet in main compose

Edge & Field

  • ✅ notqin-edge-pilot: Siemens S7 + Raspberry Pi edge kit
  • ✅ notqin-workbench: Team-facing web app (React/FastAPI/Keycloak)
  • ✅ smartex-grid: Smart Grid energy forecasting
  • ✅ temasa-twin: TEMASA digital twin plant simulator

5. Known Gaps (2026-06-28)

GapStatus
InfluxDB backup/DR strategy✅ Documented — docs/influxdb-backup-dr.md (rclone, RPO/RTO, recovery scenarios). Implementation pending.
10,000 msg/sec load test❌ Not run
MQTT TLS for internal traffic (simulator ↔ broker still plain)❌ Still a gap
OT write-back (INV-01)❌ Intentionally fenced (NASSAA research)
MCP server SSE transport🟡 stdio done, SSE deferred
Steel + Automotive sectors in main compose🟡 Code exists, not wired
Some notqin-* services (cra-attest, edge-updater, euromap-bridge)🟡 Partial scaffolds

6. Next Actions

  • Run and document load test at 10,000 msg/sec
  • Document InfluxDB backup + disaster recovery strategy
  • Fix MQTT TLS for internal traffic (simulator → broker)
  • Wire Steel + Automotive sectors into main compose
  • Complete MCP server SSE transport
  • Close remaining notqin-* service scaffolds

Graph View

Backlinks