Rulesets Explained

Suggest edits

= Rulesets in Gen AI Builder provide behavioral and policy guidance to AI Assistants and Griptape Structures. They define how your AI should behave — shaping tone, style, compliance, and persona.

Rulesets complement Knowledge Bases and Retrievers:

  • Knowledge Bases provide what the AI knows.
  • Rulesets define how the AI responds.

Rulesets are essential for building governed, brand-aligned, and compliant AI experiences in enterprise environments.

Before you start

Prerequisites for understanding Rulesets:

  • Basic understanding of Assistants and how they generate responses
  • Familiarity with Knowledge Bases and Retrievers in AI Factory
  • Awareness of Griptape Structures and how they control RAG pipelines

Suggested reading:

What is a Ruleset?

A Ruleset is a collection of natural language instructions that guide the behavior of an Assistant or Structure.

Rulesets:

  • Control tone and style
  • Enforce legal, ethical, or organizational policies
  • Define Assistant personas
  • Provide consistent behavior across applications
  • Support auditable AI governance

Without Rulesets: Assistants rely solely on base model behavior — which may not align with your enterprise standards.

How Rulesets work

At runtime:

→ Ruleset → Behavioral Instructions → LLM Prompt → AI Response
  • The Retriever gathers content from Knowledge Bases.
  • The Ruleset is injected into the Assistant’s system prompt.
  • The LLM uses:
  • The Ruleset to guide tone, style, and behavior.
  • The retrieved content to ground the response.
  • The user query as primary input.
  • The final response reflects both content grounding and Ruleset-driven behavior.

Rulesets are implemented as Griptape Structures and fully support modular, reusable patterns.

Why use Rulesets

  • Governed behavior — enforce compliance and legal requirements.
  • Brand alignment — ensure Assistants reflect your organizational voice.
  • Persona support — build distinct personas for different use cases.
  • Explainability — provide clear behavioral logic for audit and review.
  • Sovereign AI — define controlled behavior fully within your governed infrastructure.

Rulesets are critical for production-grade Assistants — especially in regulated industries.

Components of a Ruleset

A Ruleset includes:

  • Natural language Rules — clear instructions to the LLM (e.g., tone, forbidden phrases, required disclaimers).
  • Metadata — versioning, tracking, and audit support.
  • Aliases — to enable shared Rulesets across multiple Assistants.

Rules are expressed in natural language and injected at prompt time via Griptape.

Patterns of use

Single Ruleset → One Assistant

  • Simple pattern.
  • One Ruleset assigned to one Assistant.
  • Example: Customer Support Tone → Support Assistant.

Shared Ruleset → Multiple Assistants

  • A global Ruleset reused across many Assistants.
  • Example: Financial Advice Disclaimer → used by:
  • Wealth Management Copilot
  • Internal Policy Checker
  • Client-Facing Advisor Assistant

Layered Persona Rulesets

  • Compose an Assistant’s behavior from multiple Rulesets:
  • Brand Tone
  • Legal Compliance
  • Assistant Persona
  • Example:
  • Global Compliance + Banking Brand Tone + Wealth Advisor Persona

Dynamic Persona Switching

  • Advanced use case:
  • Assistants switch persona Rulesets at runtime based on user context.
  • Example:
  • One Assistant dynamically applies:
  • Retail Client Tone for B2C users.
  • Institutional Investor Tone for B2B users.

Best practices

  • Write clear, unambiguous Rules in plain natural language.
  • Test Rulesets in context — ensure they interact well with retrieved content.
  • Avoid overlapping or contradictory Rules.
  • Modularize Rulesets for reuse:
  • Global Compliance
  • Tone of Voice
  • Persona-specific behavior
  • Use versioned Rulesets and maintain an audit trail in regulated environments.
  • Review and refresh Rulesets regularly to align with evolving policies and brand guidance.

Sovereign AI alignment

Rulesets play a key role in Sovereign AI within EDB PG AI:

  • You control all Ruleset content — no external models define behavior.
  • Rulesets are fully auditable and can be versioned and reviewed.
  • Behavior governance is enforced inside your infrastructure — aligned with Knowledge Bases and Model Serving.
  • Combined with observability, Rulesets enable:
  • Transparent governance of AI behavior.
  • Alignment with legal and compliance frameworks.
  • Explainability for end users and auditors.

Summary

Rulesets are a foundational element of production AI pipelines:

  • Control behavior, tone, and compliance
  • Enable consistent brand voice
  • Support complex personas and layered behavior
  • Fully auditable and versionable
  • Critical for Sovereign AI governance in enterprise environments

Without Rulesets, Assistants lack necessary guardrails and cannot meet enterprise-grade governance requirements.

Next steps

Rulesets ensure your AI behaves the way you want — with full governance, explainability, and auditability — powering trusted, production-grade AI experiences with EDB PG AI.

Start defining Rulesets for your Assistants today.

← Prev

Retrievers explained

↑ Up

AI Factory Explained

Next →

Sovereign AI Explained

Could this page be better? Report a problem or suggest an addition!