Guardrails

enjo's guardrail ensure ai behaves responsibly by blocking sensitive, restricted, or malicious content in both user inputs and ai outputs they help prevent unintended responses that could expose confidential data or cause harmful interactions key features global & agent guardrails administrators define global rules; agent admins can add extra restrictions per agent content & word filters block offensive, sensitive, or restricted topics using predefined categories and regex how it works step 1 creating guardrails global guardrails log in to enjo dashboard go to settings → global guardrails enable global guardrails to enforce non removable restrictions across all agents agent guardrails l og in to enjo dashboard go to ai agent → settings → guardrails add or customize guardrails for that specific agent note global rules cannot be removed or modified agent guardrails only add more restrictions on top of the global ones use revert to global guardrail to reset back to global only defaults step 2 configuring filters content filters block input/output types such as hate speech or sensitive topics denied topics add restricted topics with a name, description, and sample phrases denied words/phrases block exact words or phrases using regex for precise filtering step 3 testing guardrails once you make changes in configuring filters , you’ll be prompted with an option to test your updated guardrails enter a sample message to send to the ai run the test to see both the original ai response and the final response after guardrails are applied click details to see exactly which guardrail blocked or modified the output step 4 saving guardrails after configuration or testing, click save to apply changes to global or agent guardrails