Skip to content

Flexware-Innovation/Flexware_ProveIT_2026_Community

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
dev
dev
 
 
.gitignore
.gitignore
 
 
LICENSE
LICENSE
 
 
README.md
README.md
 
 

Repository files navigation

Please use this repository to uplift your colleagues and line side personnel. Do not use this repository to consolidate your own power. Use this repo to give yourself more autonomy to do good. If you make a fork I encourage you to MIT license it.

Unified Namespace (UNS) Industrial IoT Platform

A Kubernetes-based industrial data platform that combines real-time MQTT telemetry, graph-based knowledge storage, LLM-powered analytics, and OEE tracking into a unified namespace architecture for smart manufacturing.

Architecture Overview

graph TB
    subgraph "10 - Networking"
        EDGE[RPi Edge Gateway<br/>Traefik Reverse Proxy]
    end

    subgraph "09 - Frontends"
        FE[MQTT Topic Tree Explorer<br/>Next.js Dashboard]
    end

    subgraph "08 - Backend APIs"
        MW[Middleware]
        MCP[MCP Server]
        CE[Chart Engine]
        ML[ML Predictor]
        MS[Machine Simulator]
        SA[Schema Advisor]
        HITL[HITL API]
        CR[Curated Republisher]
        UI[Uncurated Ingestor]
        CI[Curated Ingestor]
        VFP[Virtual Factory Publisher]
    end

    subgraph "07 - Ignition"
        IGN[Ignition Gateway<br/>SCADA / HMI / SparkMES OEE]
    end

    subgraph "06 - Data Sources"
        MDP[Manufacturing Data Publisher<br/>ISA-95 Simulation]
        OPC[OPC-UA Simulator]
    end

    subgraph "05 - Message Brokers"
        EMQX_U[EMQX Uncurated Broker<br/>Raw Telemetry]
        EMQX_C[EMQX Curated Broker<br/>Clean Data]
        EMQX_T[EMQX Test Broker]
    end

    subgraph "03/04 - Databases"
        PG[PostgreSQL]
        MSSQL[SQL Server]
        NEO[Neo4j Graph Database<br/>Vector Embeddings]
    end

    MDP -->|MQTT publish| EMQX_U
    MDP -->|MQTT publish| EMQX_C
    MS -->|MQTT publish| EMQX_C
    VFP -->|MQTT publish| EMQX_C

    EMQX_U -->|subscribe| UI
    EMQX_U -->|subscribe| CR
    EMQX_U -->|subscribe| MW
    EMQX_C -->|subscribe| CI
    CR -->|republish| EMQX_C

    UI -->|ingest| NEO
    CI -->|ingest| NEO
    CR -->|lineage| NEO
    MW -->|mappings| PG
    MCP -->|queries| PG

    CE -->|RAG queries| NEO
    ML -->|predictions| NEO
    SA -->|schema| NEO
    HITL -->|feedback| NEO
    MS -->|definitions| NEO

    FE -->|SSE| MW
    FE -->|WebSocket| EMQX_U
    FE -->|REST| HITL
    FE -->|REST| CE

    EDGE -->|route| FE

    IGN -->|SparkDB| MSSQL
Loading

Directory Structure

kubernetes/dev/
├── 01-namespaces/              # Kubernetes namespace definitions
├── 02-storage/                 # Persistent volume configurations
├── 03-databases/               # Relational databases
│   ├── postgres/               #   PostgreSQL - topic mappings, middleware state
│   ├── mssql/                  #   SQL Server - SparkMES OEE, production data
│   └── timescaledb/            #   TimescaleDB (testing)
├── 04-graph-database/          # Knowledge graph
│   └── neo4j/                  #   Neo4j Enterprise - topic hierarchy, embeddings, lineage
├── 05-message-brokers/         # MQTT message infrastructure
│   ├── emqx-curated/           #   EMQX broker for validated/curated data
│   ├── emqx-uncurated/         #   EMQX broker for raw telemetry
│   └── emqx-test/              #   EMQX broker for development/testing
├── 06-data-sources/            # Data generation and simulation
│   ├── manufacturing-data-publisher/  # ISA-95 hierarchy simulator (100 machines, 50 systems)
│   └── opc-ua-simulator/       #   OPC-UA industrial equipment simulator
├── 07-ignition/                # Industrial automation platform
│   └── gateway-primary/        #   Ignition SCADA gateway with SparkMES OEE scripts
├── 08-backend-apis/            # Backend microservices
│   ├── middleware/              #   MQTT topic mapping and transformation (FastAPI)
│   ├── mcp-server/             #   Model Context Protocol server for AI tooling
│   ├── chart-engine/           #   LLM-powered chart generation with RAG
│   ├── ml-predictor/           #   Time-series forecasting and anomaly detection
│   ├── machine-simulator/      #   Simulated machine management and MQTT publishing
│   ├── schema-advisor/         #   LLM-powered schema recommendations
│   ├── hitl-api/               #   Human-in-the-loop feedback API
│   ├── curated-republisher/    #   Uncurated → curated broker bridge
│   ├── curated-ingestor/       #   Curated MQTT → Neo4j ingestion
│   ├── uncurated-ingestor/     #   Raw MQTT → Neo4j ingestion with embeddings
│   └── virtual-factory-publisher/  # Kubernetes cluster metrics → MQTT
├── 09-frontends/               # Web applications
│   └── mqtt-topic-tree/        #   Real-time MQTT topic explorer (Next.js)
├── 10-networking/              # Edge networking
│   └── rpi-gateway/            #   Raspberry Pi Traefik gateway config
└── 11-tools/                   # Specialized tooling
    └── plcopen-editor/         #   IEC 61131-3 PLC programming tools
        ├── openplc-runtime/    #     OpenPLC execution runtime
        └── plcopen-api/        #     REST API for PLCOpen programs

Data Flow

flowchart LR
    subgraph Sources["Data Sources"]
        MDP["Manufacturing<br/>Data Publisher"]
        OPC["OPC-UA<br/>Simulator"]
        VFP["Virtual Factory<br/>Publisher"]
        MSIM["Machine<br/>Simulator"]
    end

    subgraph Brokers["MQTT Brokers"]
        UC["EMQX<br/>Uncurated"]
        CU["EMQX<br/>Curated"]
    end

    subgraph Ingestion["Ingestion Layer"]
        UI["Uncurated<br/>Ingestor"]
        CI["Curated<br/>Ingestor"]
        CR["Curated<br/>Republisher"]
        MW["Middleware"]
    end

    subgraph Storage["Persistence"]
        NEO[("Neo4j<br/>Graph + Vectors")]
        PG[("PostgreSQL")]
        MS[("SQL Server")]
    end

    subgraph Analytics["Analytics & AI"]
        CE["Chart Engine<br/>LLM + RAG"]
        ML["ML Predictor<br/>AutoGluon"]
        SA["Schema Advisor<br/>LLM"]
        MCP["MCP Server"]
    end

    subgraph Presentation["Presentation"]
        FE["MQTT Topic Tree<br/>Next.js"]
        IGN["Ignition<br/>SCADA/HMI"]
    end

    MDP -->|raw| UC
    MDP -->|clean| CU
    VFP --> CU
    MSIM --> CU

    UC --> UI
    UC --> CR
    UC --> MW
    CR -->|bridge| CU
    CU --> CI

    UI -->|"graph + embeddings"| NEO
    CI -->|"graph + lineage"| NEO
    CR -->|lineage| NEO
    MW -->|mappings| PG

    NEO --> CE
    NEO --> ML
    NEO --> SA
    MCP --> MW
    MCP --> PG

    CE --> FE
    MW -->|SSE| FE
    UC -->|WebSocket| FE
    IGN --> MS

    style Sources fill:#e1f5fe
    style Brokers fill:#fff3e0
    style Ingestion fill:#f3e5f5
    style Storage fill:#e8f5e9
    style Analytics fill:#fce4ec
    style Presentation fill:#f1f8e9
Loading

Component Descriptions

Databases (03-04)

Component Purpose
PostgreSQL Stores MQTT topic mappings, middleware state, and MCP server data
SQL Server SparkMES production data, OEE calculations, product orders, and location hierarchy
Neo4j Graph database storing the MQTT topic tree hierarchy, message history, schema mappings, and 384-dimension vector embeddings for semantic search

Message Brokers (05)

The platform uses a dual-broker architecture with EMQX:

Broker Purpose
Uncurated Receives raw, high-frequency telemetry from all data sources. Ingestors and the republisher subscribe here.
Curated Contains validated, transformed data. The republisher bridges selected topics from uncurated to curated.
Test Isolated broker for development and testing with no ACL restrictions.

All brokers use a rule engine that republishes incoming messages to an $ingest/ prefix topic, enabling ingestors to process data without affecting the main topic tree.

Data Sources (06)

Component Purpose
Manufacturing Data Publisher Simulates 100 industrial machines (CNC, robots, conveyors, welders, presses, AGVs) and 50 enterprise systems (ERP, MES, SCADA, DCS, WMS) following ISA-95 hierarchy
OPC-UA Simulator Simulates OPC-UA industrial equipment endpoints

Ignition Gateway (07)

The Ignition SCADA gateway provides traditional industrial monitoring and integrates with SparkMES for OEE (Overall Equipment Effectiveness) tracking. The gateway tag change script automatically:

  • Detects new machines from MQTT PartCode tags
  • Registers them in the SparkDB location hierarchy
  • Creates CounterOEE UDT instances for automated OEE calculation
  • Maps machine states (Running, Paused, E-Stop, etc.) to OEE event classifications
  • Syncs per-machine run rates and starts product orders

Backend APIs (08)

Service Purpose
Middleware Real-time MQTT topic mapping and transformation. Provides SSE streams to the frontend.
MCP Server Model Context Protocol server exposing middleware and ingestor capabilities as AI tools (for Claude, etc.)
Chart Engine Generates data visualizations from natural language queries using RAG over the Neo4j knowledge graph
ML Predictor Time-series forecasting and anomaly detection using AutoGluon. Retrains models periodically.
Machine Simulator Creates and manages simulated machines. Generates machine descriptions and ladder logic via LLM.
Schema Advisor LLM-powered recommendations for MQTT topic schema design and data structure improvements
HITL API Human-in-the-loop API for user feedback, data annotation, and topic mapping approvals
Curated Republisher Bridges messages from the uncurated broker to the curated broker with transformation and lineage tracking
Uncurated Ingestor Consumes raw MQTT messages, generates vector embeddings (MiniLM-L6-v2), and persists to Neo4j
Curated Ingestor Consumes curated MQTT messages and persists to Neo4j with lineage metadata
Virtual Factory Publisher Reads Kubernetes cluster metrics and publishes them as MQTT topics under a VirtualFactory namespace

Frontends (09)

Component Purpose
MQTT Topic Tree Next.js real-time dashboard for exploring the MQTT topic hierarchy, inspecting messages, visualizing the knowledge graph, and generating charts via natural language

Networking (10)

Component Purpose
RPi Gateway Traefik-based reverse proxy for edge deployment on Raspberry Pi. Implements a restricted routing policy — only whitelisted API routes are exposed.

Tools (11)

Component Purpose
OpenPLC Runtime IEC 61131-3 PLC program execution runtime with Modbus TCP support
PLCOpen API REST API for PLCOpen/IEC 61131-3 XML schema validation and PLC program management

Service Interactions

graph TB
    subgraph Frontend
        TREE["MQTT Topic Tree<br/>(Next.js)"]
    end

    subgraph "AI / Analytics Services"
        CE["Chart Engine"]
        ML["ML Predictor"]
        SA["Schema Advisor"]
        MCP["MCP Server"]
    end

    subgraph "Data Processing"
        MW["Middleware"]
        CR["Curated Republisher"]
        UI["Uncurated Ingestor"]
        CI["Curated Ingestor"]
        HITL["HITL API"]
        MSIM["Machine Simulator"]
        VFP["Virtual Factory Publisher"]
    end

    subgraph "Message Brokers"
        EU["EMQX Uncurated"]
        EC["EMQX Curated"]
    end

    subgraph "Databases"
        NEO[("Neo4j")]
        PG[("PostgreSQL")]
    end

    subgraph "External"
        LLM["OpenAI API"]
        K8S["Kubernetes API"]
    end

    TREE -->|"SSE"| MW
    TREE -->|"WebSocket (MQTT)"| EU
    TREE -->|"REST"| HITL
    TREE -->|"REST"| CE

    CE -->|"Bolt"| NEO
    CE -->|"REST"| UI
    CE -->|"REST"| LLM

    ML -->|"Bolt"| NEO
    ML -->|"REST"| MSIM
    ML -->|"REST"| LLM

    SA -->|"Bolt"| NEO
    SA -->|"REST"| MCP
    SA -->|"REST"| LLM

    MCP -->|"REST"| MW
    MCP -->|"REST"| UI
    MCP -->|"SQL"| PG

    HITL -->|"Bolt"| NEO
    HITL -->|"REST"| SA

    MW -->|"MQTT"| EU
    MW -->|"MQTT"| EC
    MW -->|"SQL"| PG

    CR -->|"MQTT sub"| EU
    CR -->|"MQTT pub"| EC
    CR -->|"Bolt"| NEO

    UI -->|"MQTT sub"| EU
    UI -->|"Bolt"| NEO

    CI -->|"MQTT sub"| EC
    CI -->|"Bolt"| NEO

    MSIM -->|"MQTT pub"| EC
    MSIM -->|"Bolt"| NEO
    MSIM -->|"REST"| LLM

    VFP -->|"MQTT pub"| EC
    VFP -->|"Bolt"| NEO
    VFP -->|"REST"| K8S
Loading

Key Design Patterns

Dual MQTT Broker (Uncurated / Curated)

Raw telemetry flows into the uncurated broker without filtering. The curated republisher selectively bridges validated messages to the curated broker after transformation. This separation keeps noisy raw data isolated from clean, production-quality data streams.

ISA-95 Topic Hierarchy

MQTT topics follow the ISA-95 standard: enterprise/site/area/line/cell/asset/metric. The Manufacturing Data Publisher simulates this hierarchy, and the Neo4j graph database models it as a navigable tree structure.

Message-to-Graph Pipeline

Every MQTT message is ingested into Neo4j as a graph node with relationships to its topic path segments. The uncurated ingestor generates 384-dimensional vector embeddings (MiniLM-L6-v2) for each topic, enabling semantic similarity search across the topic namespace.

LLM-Native Analytics

Multiple services integrate with LLMs for intelligent data processing:

  • Chart Engine uses RAG (Retrieval-Augmented Generation) to find relevant topics in Neo4j before generating visualizations
  • ML Predictor uses AutoGluon for time-series forecasting with LLM-assisted interpretation
  • Schema Advisor provides data structure recommendations through multi-turn conversations
  • Machine Simulator generates machine descriptions and ladder logic

Model Context Protocol (MCP)

The MCP Server exposes middleware and ingestor capabilities as standardized AI tools, enabling AI assistants (like Claude) to interact with the platform's data through tool use.

Edge Gateway Security

The Raspberry Pi gateway uses Traefik with an explicit allowlist — only specific API routes are exposed to the edge network. Administrative and write-heavy endpoints are blocked by default.

Technology Stack

Layer Technologies
Orchestration Kubernetes (k3s)
Message Broker EMQX (MQTT, WebSocket, TLS)
Graph Database Neo4j Enterprise (Bolt, APOC, Vector Indexes)
Relational DB PostgreSQL, Microsoft SQL Server
Backend Python, FastAPI, aiohttp, asyncio
ML/AI AutoGluon, sentence-transformers, scikit-learn, OpenAI API
Frontend Next.js, React, TypeScript, Tailwind CSS, Chart.js
SCADA/HMI Inductive Automation Ignition, SparkMES
PLC OpenPLC Runtime, IEC 61131-3 / PLCOpen XML
Edge Networking Traefik, Tailscale
Protocols MQTT, OPC-UA, Modbus TCP, SSE, WebSocket, MCP
Containerization Docker, local registry

Namespace Organization

graph LR
    subgraph "Data Stores"
        NS_PG["postgres"]
        NS_MS["mssql"]
        NS_NEO["neo4j"]
    end

    subgraph "Message Brokers"
        NS_EC["emqx-curated"]
        NS_EU["emqx-uncurated"]
        NS_ET["emqx-test"]
    end

    subgraph "Applications"
        NS_DS["data-sources<br/><i>MFG Publisher, OPC-UA Sim</i>"]
        NS_IGN["ignition<br/><i>SCADA Gateway</i>"]
        NS_SVC["services<br/><i>Ingestors, Republisher,<br/>Simulator, Schema Advisor,<br/>HITL API, Virtual Factory</i>"]
        NS_API["backend-apis<br/><i>Middleware, MCP Server,<br/>Chart Engine, ML Predictor,<br/>PLCOpen API</i>"]
        NS_FE["frontends<br/><i>MQTT Topic Tree</i>"]
    end

    subgraph "Tools"
        NS_PLC["openplc<br/><i>OpenPLC Runtime</i>"]
    end

    NS_DS -->|MQTT| NS_EC
    NS_DS -->|MQTT| NS_EU
    NS_SVC -->|ingest| NS_NEO
    NS_API -->|queries| NS_PG
    NS_API -->|queries| NS_NEO
    NS_FE -->|WebSocket| NS_EU
    NS_FE -->|REST| NS_SVC
    NS_FE -->|REST| NS_API
    NS_IGN -->|SQL| NS_MS
Loading

Getting Started

Prerequisites

  • Kubernetes cluster (k3s recommended)
  • Local Docker registry
  • kubectl and docker CLI tools

Build and Deploy Workflow

Frontend applications (Next.js):

cd dev/09-frontends/<app-name>
npm run build
docker build -t <app-name>:latest .
docker tag <app-name>:latest localhost:5000/<app-name>:latest
docker push localhost:5000/<app-name>:latest
kubectl rollout restart deployment/<app-name> -n frontends

Backend APIs (Python):

cd dev/08-backend-apis/<app-name>
docker build -t <app-name>:latest .
docker tag <app-name>:latest localhost:5000/<app-name>:latest
docker push localhost:5000/<app-name>:latest
kubectl rollout restart deployment/<app-name> -n backend-apis

Useful Commands

# Check deployment status
kubectl get pods -n <namespace>

# View logs
kubectl logs -f deployment/<app-name> -n <namespace>

# Check all deployments
kubectl get deployments --all-namespaces

License

This project is licensed under the MIT License -- see the LICENSE file for details.

About

Open Source Demo from Flexware Innovation ProveIT 2026

Resources

Readme

License

MIT license
Activity
Custom properties

Stars

16 stars

Watchers

1 watching

Forks

8 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages