# ZeroGuess AI — Manufacturing AI Education Platform (Full Documentation)

## Overview

ZeroGuess AI is a manufacturing AI education platform created by Tobias Lante. It provides hands-on learning experiences for engineers, developers, and students who want to understand how AI works in manufacturing environments.

The platform centers around OpenShopFloor — a free Manufacturing AI Playground that connects to a live factory simulation running 24/7. Users can chat with AI agents that have access to real-time factory data, build visual AI workflows, and complete gamified manufacturing challenges.

## Mission

Manufacturing AI is hard to learn because you need a factory to practice on. ZeroGuess AI solves this by providing a live factory simulation with 111 AI tools, so anyone can learn manufacturing AI without needing factory access, NDAs, or expensive infrastructure.

## OpenShopFloor — The Playground

### What It Is
OpenShopFloor is a free web-based Manufacturing AI Playground. It connects to a live factory simulation and provides AI agents, visual flow editors, and gamified challenges for learning manufacturing AI.

### URL
https://openshopfloor.zeroguess.ai

### Key Features

#### 1. Live Factory Simulation
A full manufacturing simulation runs 24/7 with:
- 8 CNC machines (9001-9018)
- 5 injection molding stations (SGM-001 to SGM-005)
- Assembly lines
- Quality inspection stations
- Tool management
- Real-time OEE calculation
- Order tracking and scheduling
- Stock/inventory management
- Energy monitoring

#### 2. 111 MCP Tools
The factory data is exposed through 6 MCP (Model Context Protocol) servers:
- **ERP Server** (mcp-erp): Orders, materials, scheduling, customer data
- **Manufacturing Server** (mcp-fertigung): Machine status, production history, capacity
- **QMS Server** (mcp-qms): Quality inspections, SPC data, Cpk values, calibration
- **WMS Server** (mcp-wms): Stock levels, warehouse movements, purchase orders
- **TMS Server**: Tool inventory, tool life, maintenance schedules
- **Knowledge Graph Server**: Dependency analysis, impact analysis

Each MCP server provides 15-20 tools that AI agents can call to query factory data.

#### 3. AI Chat Interface
Users can chat with an AI agent that has access to all 111 MCP tools. The agent can:
- Query real-time factory data (OEE, stock, orders, quality)
- Analyze production patterns
- Identify bottlenecks and risks
- Suggest optimizations
- Execute multi-step analyses

#### 4. Visual Flow Editor
A Node-RED-based drag-and-drop editor for building AI workflows:
- Connect MCP tools, LLMs, and decision logic visually
- Run flows against live factory data
- Inspect every step's output
- No coding required

#### 5. AI Agents Marketplace
Pre-built agents for common manufacturing tasks:
- OEE Monitor & Diagnostics
- Material & Capacity Planning
- Delivery Risk & Quality Guard
- Energy Optimizer
- Strategic Planner

#### 6. Gamified Challenges
6 challenges from beginner to expert:

**OEE Champion (Beginner, 1 hour)**
Goal: Achieve and maintain OEE above 85% across all machines for 24 simulated hours.
Tools: factory_get_latest_oee, factory_get_machine_oee, factory_get_production_history, factory_get_capacity_overview

**Zero Delay (Intermediate, 2 hours)**
Goal: Ensure zero late deliveries over 48 simulated hours by managing priorities and capacity.
Tools: factory_get_orders_at_risk, factory_get_customer_otd, factory_check_material_readiness, factory_get_va05_summary

**Lean Inventory (Intermediate, 2 hours)**
Goal: Reduce inventory holding costs by 20% without causing any material stockouts.
Tools: factory_get_low_stock_items, factory_get_stock_item, factory_get_pending_purchases, factory_get_md04_overview

**Quality First (Advanced, 1 hour)**
Goal: Achieve Cpk > 1.33 on all quality characteristics and resolve all SPC alarms.
Tools: factory_get_spc_alarms, factory_get_cpk_overview, factory_get_calibration_due, factory_get_quality_notifications

**Energy Saver (Beginner, 1 hour)**
Goal: Reduce energy consumption per part by 15% over 24 simulated hours.
Tools: factory_get_energy_overview, factory_get_energy_per_part, factory_get_base_load, factory_get_energy_per_machine

**Full Auto (Expert, 4 hours)**
Goal: Keep the factory running autonomously for 72 simulated hours with no manual intervention.
Tools: factory_get_latest_oee, factory_get_capacity_overview, factory_get_orders_at_risk, factory_get_low_stock_items, factory_get_spc_alarms, factory_get_energy_overview, factory_get_kpi_dashboard

### Architecture
- Frontend: Next.js on Cloudflare Pages
- Backend: Express.js Gateway on Kubernetes
- Factory Simulation: Docker containers (ERP, WMS, Manufacturing, Assembly)
- Database: PostgreSQL
- Node-RED: Per-user pods on Kubernetes
- LLM: Platform-provided or bring your own

### Pricing
- Free: Full access to all features, 100K tokens/month for platform LLM
- Bring Your Own Key: Connect OpenAI, Anthropic, or any OpenAI-compatible endpoint for unlimited usage

## About the Creator

**Tobias Lante** is a manufacturing technology specialist who built ZeroGuess AI as an independent education project. After years in manufacturing consulting, he was frustrated by the barriers to learning manufacturing AI — NDAs, IT delays, lack of real data. He built a complete factory simulation so anyone can learn manufacturing AI by doing.

- LinkedIn: https://www.linkedin.com/in/tobias-lante-99477a239/
- GitHub: https://github.com/TobiasLante

## Links

- ZeroGuess AI: https://www.zeroguess.ai
- OpenShopFloor Playground: https://openshopfloor.zeroguess.ai
- Live Factory Dashboard: https://osf-factory.zeroguess.ai
- GitHub (Frontend): https://github.com/TobiasLante/openshopfloor
- Impressum & Privacy: https://www.zeroguess.ai/impressum.html