FAQ

This FAQ covers the core Data Science concepts, model governance approach, and when to use ML, GenAI, MCP, CEP, and visualizations

Frequently asked questions

What is the Data Science app responsible for?

Show answer

Data Science is where you configure and govern reusable AI, ML, and analytics assets.

It centralizes model settings, execution parameters, and related definitions so they can be invoked from APIs, batch jobs, or stream processing flows.

See Overview.

What are the main building blocks in Data Science?

Show answer

Think in five main groups:

• ML Models for traditional model configuration and execution settings

• GenAI Models for LLMs, agents, prompts, tools, and governance

• MCP Servers for exposing existing platform capabilities over MCP

• Complex Event Processing for real-time stream logic and derived signals

• Data Visualizations for dashboards and reporting views

See Overview.

How is Data Science different from Devops?

Show answer

Data Science defines governed AI, ML, and analytics assets.

Devops provides the runners, sagas, gateways, and deployments that invoke them.

A simple mental model is:

• Data Science defines the model or analytical asset

• Devops decides where and how that asset is executed

See Devops FAQ.

How is Data Science different from Configuration?

Show answer

Configuration stores reusable logic such as queries, rules, and dynamic handlers.

Data Science stores reusable analytical and AI assets such as models, agents, CEP flows, and dashboards.

They often work together, but they solve different layers of runtime behavior.

See Configuration FAQ.

What is an ML model in Rierino?

Show answer

An ML model definition stores the metadata and parameters needed to run or schedule traditional machine learning logic.

It can include versioning, assets, scheduling, parameters, and step-based pipelines.

This keeps inference and training settings consistent across environments.

See ML Models.

What is a GenAI model in Rierino?

Show answer

A GenAI model definition governs an LLM-backed capability.

It includes provider settings, prompts, memory, tool access, interfaces, and runner-level access control.

This is the main starting point for AI agents and governed LLM usage on the platform.

See GenAI Models.

When should I use ML Models versus GenAI Models?

Show answer

Use ML Models for predictive scoring, classification, regression, feature pipelines, and similar structured model execution.

Use GenAI Models for natural language interaction, content generation, tool-using agents, and LLM-driven reasoning.

If the output is primarily statistical prediction, start with ML. If it is conversational or prompt-driven, start with GenAI.

What is an AI agent in this model?

Show answer

An AI agent is a governed GenAI model with access to approved tools and context.

Those tools can include sagas, states, systems, prompts, and interfaces.

That lets the agent do more than generate text. It can perform controlled business actions.

How is tool access controlled for GenAI agents?

Show answer

Tool access is explicitly configured.

A model can be allowed to call selected sagas, read or write selected states, interact with selected systems, and use selected scripts or things.

Governance settings also limit repeated calls and excessive tool loops.

See GenAI Models.

Do GenAI capabilities have special runtime requirements?

Show answer

Yes.

GenAI-enabled runners usually need more memory and are best kept separate from lightweight general-purpose runners.

This helps isolate resource-heavy agent workloads and keeps standard services lean.

What is an MCP server in Rierino?

Show answer

An MCP server maps existing platform capabilities into the Model Context Protocol.

In practice, it exposes selected sagas, prompts, and resources so MCP clients can consume them as tools or context.

See MCP Servers.

When should I use an MCP server instead of a GenAI model?

Show answer

Use a GenAI model when you need an LLM-backed agent or prompt-driven capability.

Use an MCP server when you want to expose existing tools and resources in a standard protocol for external MCP clients.

They complement each other. One governs model behavior. The other exposes capabilities.

What is Complex Event Processing used for?

Show answer

Complex Event Processing is for real-time stream analysis.

Use it when you need windowed aggregations, event correlation, continuous enrichment, or derived signals based on incoming data streams.

It is a better fit than request-response APIs when logic depends on continuous event flow.

See Complex Event Processing.

How is CEP different from a normal saga?

Show answer

A saga is request or process orchestration.

CEP is continuous stream processing.

Use a saga for step-based business flows. Use CEP when data keeps arriving and the system must evaluate patterns or aggregates over time.

What are data visualizations for?

Show answer

Data visualizations are embedded dashboards and reporting views.

They let teams present metrics, trends, charts, and tables using configured layouts and linked data sources.

This is useful for operational visibility and business-facing reporting inside the platform.

See Data Visualizations.

Are Data Science assets only for batch workloads?

Show answer

No.

They can be used in real-time APIs, in background processes, or in stream-oriented pipelines.

A saga can call an ML or GenAI handler directly. A batch process can train or backfill. A CEP flow can compute live signals.

See Overview.

Why keep models and AI settings in one central app?

Show answer

It improves governance, reuse, and rollout safety.

Teams can version assets, review changes, control access, and update execution settings without scattering model logic across many services.

How does Data Science relate to Design?

Show answer

Design builds the admin UI and screens.

Data Science provides model-driven capabilities those screens may call or display.

For example, a UI can trigger an agent interaction, render an analytics view, or display outputs from an ML-backed endpoint.

See Design FAQ.

What should I understand first if I am new to Data Science?

Show answer

Start with these concepts:

• ML model for traditional predictive logic

• GenAI model for LLM and agent behavior

• MCP server for protocol-based capability exposure

• CEP flow for continuous event analysis

• Visualization for reporting and dashboards

Once these are clear, the rest of the section becomes much easier to navigate.

Where should I go next for deeper Data Science FAQs?

Show answer

Start with the subsection that matches your goal:

• Need governed predictive models → ML Models

• Need agents, prompts, or LLM governance → GenAI Models

• Need MCP exposure for tools and resources → MCP Servers

• Need streaming analytics → Complex Event Processing

• Need embedded reporting → Data Visualizations

Subsection-specific FAQs can then go deeper into model lifecycle, agent setup, prompts, tool governance, CEP patterns, and dashboard design.