Skip to content

Multi-Provider AI Guardrails #76

Description

@shaneutt

Extend the guardrails filter to support external content safety providers and response-side inspection.

If feasible make this generic, so that we can plug into a specific provider now (say, Nemo), but keep the schema generalized to support others easily later.

Scope:

  • External guardrail provider integration via webhook callout
  • Built-in PII detection (SSN, credit card, phone, email, URL patterns)
  • Response-side guardrails (inspect LLM output before returning to client)
  • Actions: allow, reject, mask (anonymize matched content)
  • Per-route guardrail policy configuration
  • Configurable fail-open/fail-closed on provider errors

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Type

Projects

Relationships

None yet

Development

No branches or pull requests

Issue actions