Introduction to Guardrail

Build safer AI applications with policy-driven guardrail workflows

Overview

The Guardrail module helps you discover, configure, and deploy safety protections for model traffic in Bud AI Foundry. You can combine provider probes, tune severity, choose guard types, and deploy to projects or endpoints through a guided workflow.

What You Can Do

Discover provider probes Browse guardrails from Bud Sentinel and cloud providers with searchable metadata, tags, and status indicators.

Compose reusable profiles Create profiles with selected probes, guard types, and thresholds to standardize safety policy across teams.

Deploy with flexibility Deploy as a standalone Guardrail Endpoint or attach to an existing deployment.

Operate with visibility Review details through profile pages and project-level tables with status, timestamps, and deployment context.

Core Capabilities

Capability	Description	Why It Matters
Catalog & Discovery	Search, tags, and status-based filtering	Faster probe selection
Multi-step Setup	Guided provider → probe → deployment flow	Reduced setup errors
Guard Types	Configure input/output (and applicable retrieval/tools contexts)	Better risk control
Severity Tuning	Profile-level threshold with probe/rule refinements	Policy precision
Project Integration	Guardrail visibility inside project details	Easier governance

Typical Use Cases

Pre-Response Safety

Screen prompts and outputs before they reach users.

Compliance Enforcement

Apply consistent policies to regulated workloads.

Prompt Injection Defense

Block jailbreak and misuse patterns early in request flow.

PII Protection

Detect sensitive data exposure and reduce leakage risk.

Getting Started

Quick Start

Create and deploy your first guardrail quickly.

Guardrail Concepts

Learn profiles, probes, rules, and execution behavior.

Step-by-Step Tutorial

Build a production-ready guardrail profile.