Keyword Filter
Detects and blocks text containing specified banned keywords or phrases. Uses case-insensitive matching with word boundaries to identify forbidden terms and triggers if any configured keyword is found.
Configuration
{
"name": "Keyword Filter",
"config": {
"keywords": ["confidential", "secret", "internal only", "do not share"]
}
}
Parameters
keywords(required): List of banned keywords or phrases to detect. Trailing.,,,!,?,;, and:characters are removed before matching. Values containing only those characters are rejected during configuration validation.
What It Returns
Returns a GuardrailResult with the following info dictionary:
{
"guardrail_name": "Keyword Filter",
"matched": ["confidential", "secret"],
"checked": ["confidential", "secret", "internal only", "do not share"],
"sanitized_keywords": ["confidential", "secret", "internal only", "do not share"]
}
matched: List of keywords found in the textchecked: List of keywords that were configured for detectionsanitized_keywords: List of keywords used for matching after trailing punctuation is removed