TextSafetyConfig

Options & Configuration ยท AiDotNet.Safety

Configuration for text safety modules.

For Beginners

These settings control how text content is checked for safety issues. Enable the checks you need โ€” toxicity catches offensive language, PII detection finds personal information, jailbreak detection catches prompt manipulation attempts.

How It Works

Controls which text safety checks are enabled and their sensitivity thresholds.

Constructors

Constructor Summary
TextSafetyConfig()

Properties

Property Summary
PiiType[] PIITypes { get; set; } Gets or sets the kinds of PII the PII detector looks for.
bool? BlockJailbreaks { get; set; } Gets or sets whether a detected jailbreak or prompt injection blocks the input.
bool? CopyrightDetection { get; set; } Gets or sets whether copyright/memorization detection is enabled.
bool? HallucinationDetection { get; set; } Gets or sets whether hallucination detection is enabled.
bool? JailbreakDetection { get; set; } Gets or sets whether jailbreak/prompt injection detection is enabled.
bool? PIIDetection { get; set; } Gets or sets whether PII (personally identifiable information) detection is enabled.
bool? ToxicityDetection { get; set; } Gets or sets whether toxicity detection is enabled.
double? HallucinationThreshold { get; set; } Gets or sets the contradiction score (0-1) at or above which a pair of sentences is reported as a self-contradiction (a hallucination signal).
double? JailbreakSensitivity { get; set; } Gets or sets the jailbreak detection sensitivity (0-1).
double? ToxicityThreshold { get; set; } Gets or sets the toxicity score threshold (0-1).
int? MaxInputLength { get; set; } Gets or sets the maximum input text length.
string[] Languages { get; set; } Gets or sets the languages to support for text safety checks.