Getting Started

Installation

npm install open-guardrail

Or install only the parts you need:

npm install open-guardrail-core open-guardrail-guards

Quick Start

import { defineGuardrail, promptInjection, pii, keyword } from 'open-guardrail';

const guard = defineGuardrail({
  guards: [
    promptInjection({ action: 'block' }),
    pii({ entities: ['email', 'phone'], action: 'mask' }),
    keyword({ denied: ['hack', 'exploit'], action: 'block' }),
  ],
});

const result = await guard('user input text here');
if (!result.passed) console.log('Blocked:', result.action);
// result.output contains masked text when PII is detected

Or use the pipe() shorthand:

const result = await pipe(
  promptInjection({ action: 'block' }),
  pii({ entities: ['email'], action: 'mask' }),
).run('user input');

How It Works

1. Create guards — each guard is a function that inspects text and returns allow, block, warn, or override
2. Build a pipeline — chain guards together with pipe() or createPipeline()
3. Run text — call .run(text) to get a PipelineResult with pass/fail, individual guard results, and optional transformed output

Guard Actions

Action	Meaning
allow	Text passed this guard
block	Text should be rejected
warn	Text passed but flagged
override	Text was modified (e.g., PII masking)

Pipeline Modes

• fail-fast (default) — stops at the first block result
• run-all — runs every guard, then picks the highest-priority action

Result Structure

interface PipelineResult {
  passed: boolean;           // true if not blocked
  action: GuardAction;       // highest-priority action
  results: GuardResult[];    // per-guard details
  input: string;             // original input
  output?: string;           // modified text (if any guard used override)
  totalLatencyMs: number;
  metadata: { ... };
}

Next Steps

• YAML Config — configure guards without code
• Guards Overview — browse all 33 built-in guards
• SDK Adapters — integrate with OpenAI, Vercel AI, LangChain