LLM Guardrail CLI

CLI commands for LLM-based guardrails in PraisonAI TypeScript

LLM Guardrail CLI Commands

The praisonai-ts CLI provides the guardrail command for content validation.

Basic Usage

# Check content against guardrails
praisonai-ts guardrail check "Your content here"

# Check with custom criteria
praisonai-ts guardrail check "Content to validate" --criteria "Must be professional"

# Get JSON output
praisonai-ts guardrail check "Hello world" --json

Example Output:

{
  "success": true,
  "data": {
    "status": "passed",
    "score": 0.95,
    "message": "Content passes all criteria",
    "reasoning": "The content is appropriate and safe"
  }
}

Status Values

Status	Description
passed	Content meets all criteria
failed	Content does not meet criteria
warning	Content partially meets criteria

SDK Usage

For programmatic guardrail usage:

import { LLMGuardrail } from 'praisonai';

const guard = new LLMGuardrail({
  name: 'safety',
  criteria: 'Content must be safe and appropriate',
  threshold: 0.8
});

const result = await guard.check('Hello world');
console.log(result.status); // 'passed', 'failed', or 'warning'
console.log(result.score);  // 0-1
console.log(result.reasoning);

For more details, see the LLM Guardrail SDK documentation.