Support for custom guardrail rules

[nagasatish007]

Which SDK is this for?

Both

Feature Type

New Guardrail

Problem Statement

The built-in guardrails (PII detection, prompt injection, content moderation) are great, but I need to add custom rules specific to my domain (e.g., blocking certain industry-specific terms).

Proposed Solution

Add a plugin system that allows developers to register custom guardrail functions that run alongside the built-in ones.

Alternatives Considered

No response

Use Case

In my healthcare application, I need to block certain medical terms and ensure HIPAA compliance beyond standard PII detection. I want to define custom rules without forking the SDK.

Example Code

import { TealOpenAI, CustomGuardrail } from 'tealtiger';

const medicalTermsGuardrail: CustomGuardrail = {
  name: 'medical-terms-blocker',
  check: async (input: string) => {
    const blockedTerms = ['diagnosis', 'prescription', 'treatment'];
    const found = blockedTerms.find(term => input.toLowerCase().includes(term));
    
    return {
      passed: !found,
      reason: found ? `Blocked medical term: ${found}` : undefined,
    };
  },
};

const client = new TealOpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  enableGuardrails: true,
  customGuardrails: [medicalTermsGuardrail],
});

Would you like to contribute this feature?

• I'm willing to submit a PR to implement this feature
• I need help implementing this feature

Additional Context

No response