AI Capabilities

AI Capabilities define how the IoTRoutes AI behaves when executing a task.

They represent the behavioral layer of the AI system, where each capability describes a specific function, role, or responsibility that an AI agent can perform inside the platform.

Instead of relying on a single generic AI assistant, IoTRoutes uses a capability-driven model, where each capability acts like a specialized AI agent with its own rules, instructions, and knowledge context.

What is an AI Capability?

An AI Capability is a configurable unit that defines:

• what the AI is allowed to do,
• how it should behave,
• what knowledge it can access,
• and how it should respond in a given context.

Each capability is designed for a specific use case such as:

• user assistance,
• data interpretation,
• IoT message transformation,
• predictive analysis,
• or operational support.

Capability Structure

Each AI Capability is composed of four main elements:

1. Agent Capability (Core Definition)

The Agent Capability defines the identity and purpose of the AI function.

It includes:

• Name
• Description

This layer answers the question:

“What is this AI responsible for?”

Examples:

• Chat Assistant
• IoT Message Converter
• Activity Summarizer
• Predictive Analyzer

2. Instructions (System Behavior)

Instructions define the behavioral rules of the capability.

They act as system-level guidance used by the AI model during execution.

Instructions typically include:

• how the AI should respond,
• tone and structure of responses,
• constraints or limitations,
• formatting rules,
• domain-specific logic.

This layer answers:

“How should the AI behave when performing this task?”

Example (conceptual):

• Always summarize IoT activities in a structured format
• Focus on technical clarity
• Avoid unnecessary explanations
• Prioritize accuracy over verbosity

3. Knowledge Associations

Each capability can be linked to one or more AI Knowledges.

These knowledge sources provide contextual intelligence to the AI.

A Knowledge can include:

• platform documentation,
• business rules,
• technical guides,
• API references,
• workflow explanations,
• troubleshooting procedures.

This layer answers:

“What information can the AI rely on to perform correctly?”

Only the knowledges assigned to a capability are loaded during execution, ensuring:

• contextual precision,
• reduced noise,
• improved performance,
• and controlled access to information.

4. Contextual AI Execution

When a capability is executed, IoTRoutes dynamically builds the AI context.

The execution process includes:

1. Selecting the capability
2. Loading its instructions
3. Loading associated knowledges
4. Sending the structured context to the assigned AI provider
5. Generating a response based on this controlled environment

This ensures that every AI response is:

• context-aware,
• capability-specific,
• and aligned with system rules.

How Capabilities Work in Practice

A capability is not just a configuration item — it behaves like a specialized AI agent profile.

For example:

Chat Capability

• Focus: user interaction
• Knowledge: platform usage, UI, FAQs
• Behavior: conversational, helpful, guided responses

IoTMessageToPMS Capability

• Focus: message transformation
• Knowledge: PMS structure, IoT schema mapping rules
• Behavior: strict formatting, deterministic output

PredictWithLLM Capability

• Focus: forecasting and analysis
• Knowledge: historical patterns, analytics rules
• Behavior: analytical, structured reasoning output

Each capability produces a different AI behavior even if the same model is used underneath.

Capability Isolation Principle

One of the core principles of IoTRoutes AI is:

A capability only uses the knowledge and rules assigned to it.

This guarantees:

• predictable behavior,
• reduced hallucination,
• secure data segmentation,
• and easier governance.

Role in the AI Architecture

AI Capabilities sit between:

• AI Service Providers (execution layer)
• AI Knowledges (context layer)

They act as the decision and behavior layer that controls:

• how AI thinks,
• what AI is allowed to do,
• and what context it can access.

Summary

AI Capabilities are the foundation of AI behavior in IoTRoutes.

They combine:

• Agent definition (what it does)
• Instructions (how it behaves)
• Knowledge associations (what it knows)