Guardrails

warning

TakeShape's Guardrails features are considered experimental. Use in production with the knowledge that these features may change.

Introduction

Guardrails provide safety mechanisms to prevent unintended or malicious behavior in your agent. Guardrails can be used to enforce topic constraints, block profanity, PII, and ensure brand safety.

Limitations & Billing

Guardrails are currently limited to one guard per project.

Guardrails are billed as a separate line item based on usage, and usage can be viewed on the billing page of the TakeShape dashboard.

Configuring Guardrails

Guardrails are a schema-only feature at this point in time. Check out the schema config.