Product Architecture
1. Overall architecture
Prodia uses an AI + BI dual-core architecture so that conversation, analytics, diagnosis, and action guidance form one coherent product experience.
| Architecture layer | Composition | Responsibility |
|---|---|---|
| Interaction layer | Web workspace, dashboards, conversational entry points | Accept questions, present results, support multi-turn interaction |
| Intelligence layer | LLM, Agent, knowledge capabilities | Understand intent, invoke tools, organize analytical paths |
| Analytics layer | Metric engine, scenario analysis tools, diagnosis models | Run output, OEE, quality, takt, fault, and SPC analysis |
| Data layer | AI SCADA / MES / databases / data platforms | Provide equipment, process, output, quality, and event data |
2. Architecture direction: UNS + MCP + Agent
Today Prodia delivers value through the AI + BI combination. Over time it evolves along a coordinated architecture of UNS + MCP + Agent.
| Component | Architectural role | Value |
|---|---|---|
| UNS | Unify factory objects, states, and semantics | Let the system first understand the factory |
| MCP | Expose analysis, diagnosis, and collaboration capabilities to AI | Make industrial capabilities callable in a standard way |
| Agent | Organize task decomposition, capability orchestration, and result structuring | Push analysis toward business closed loops |
See UNS, MCP and Agent Strategy for the strategic explanation, and Agent Technology Overview for the agent-side view.
3. Why AI and BI must work together
If you only have AI
- It may speak fluently but calculate unreliably
- It lacks consistent industrial semantics and metric definitions
- Results become hard to verify and trace
If you only have BI
- Learning cost remains high
- Users are trapped in fixed reports and rigid dashboards
- Diagnosis and next-step guidance are still weak
Prodia’s combination
- AI understands the question and organizes the analysis path
- BI provides stable metric definitions and trustworthy calculation
- Agent turns results into executable diagnosis and recommendations
4. How Prodia improves trustworthiness
| Mechanism | Description |
|---|---|
| Business semantic tools | Results come from governed business tools rather than ad hoc free-form generation |
| Unified metric definitions | OEE, yield, takt, fault, and other indicators follow consistent business logic |
| Clarification before analysis | The system asks follow-up questions when scope or grain is unclear |
| Knowledge enhancement | SOPs, cases, and historical experience are used to support recommendations |
| Explainable outputs | Results are presented with reasoning paths, focus points, or investigation directions whenever possible |
5. End-to-end data loop
- Data acquisition from equipment, AI SCADA, MES, or databases
- Data integration through governance, cleaning, modeling, and semantic alignment
- Analytical execution using the right domain tool and model for the question
- Interactive presentation through natural language, charts, and recommendations