List of Guardrail Checks
Each Guardrail Check has a specific purpose, it’s own parameters, supported hooks, and sources.
The logic for all of the Guardrail Checks (including Partner Guardrails) is open source.
Checks if the request or response text matches a regex pattern.
Parameters: rule:
Supported On:
Checks if the content contains a certain number of sentences. Ranges allowed.
Parameters: minSentences:
Supported On:
Checks if the content contains a certain number of words. Ranges allowed.
Parameters: minWords:
Supported On:
Checks if the content contains a certain number of characters. Ranges allowed.
Parameters: minCharacters:
Supported On:
Check if the response JSON matches a JSON schema.
Parameters: schema:
Supported On:
Check if the response JSON contains any, all or none of the mentioned keys.
Parameters: keys:
Supported On:
Checks if the content contains any, all or none of the words or phrases.
Parameters: words:
Supported On:
Checks if all the URLs mentioned in the content are valid
Parameters: onlyDNS:
Supported On:
Checks if the content contains code of format SQL, Python, TypeScript, etc.
Parameters: format:
Supported On:
Check if the given string is lowercase or not.
Parameters: format:
Supported On:
Check if the content ends with a specified string.
Parameters: Suffix:
Supported On:
Makes a webhook request for custom guardrails
Parameters: webhookURL:
Supported On:
Check if the JWT token is valid.
Parameters: JWKS URI:
Check if the inference model to be used is in the whitelist.
Parameters: Models:
Checks if the content passes the mentioned content moderation checks.
Parameters: categories:
Supported On:
Checks if the response content is in the mentioned language.
Parameters: language:
Supported On:
Detects Personally Identifiable Information (PII) in the content.
Parameters: categories:
Supported On:
Detects if the content is gibberish.
Parameters:
Supported On:
Partner Guardrails
Acuvity
- Scan prompts and responses for security threats
- Detect PII, toxicity, and prompt injections
- Real-time content analysis and filtering
Aporia
- Validate custom Aporia policies via project ID
- Define policies on your Aporia dashboard
- Seamless integration with Portkey checks
Azure
- Detect and redact sensitive PII data
- Apply Azure’s comprehensive content safety checks
- Enterprise-grade security compliance
AWS Bedrock
- Analyze and redact PII to prevent manipulation
- Integrate AWS Guardrails directly in Portkey
- Advanced security policy enforcement
Lasso Security
- Analyze content for security risks and jailbreaks
- Enforce custom policy violations detection
- AI-powered threat detection and prevention
Mistral
- Detect and filter harmful content automatically
- Multi-dimensional content safety analysis
- Real-time moderation capabilities
Pangea
- Guard LLM inputs and outputs with Text Guard
- Detect malicious content and data transfers
- Prevent model manipulation attempts
Patronus
- Detect hallucinations and factual errors
- Assess quality: conciseness, helpfulness, tone
- Identify gender and racial bias in outputs
Pillar
- Scan prompts and responses comprehensively
- Detect PII, toxicity, and injection attacks
- Enterprise security and compliance features
Prompt Security
- Scan prompts for security vulnerabilities
- Analyze responses for policy violations
- Advanced threat detection and mitigation
Bring Your Own Guardrail
We have built Guardrails in a very modular way, and support bringing your own Guardrail using a custom webhook! Learn more here.Portkey’s Guardrails
Along with the partner Guardrails, there are also deterministic as well as LLM-based Guardrails supported natively by Portkey.BASIC Guardrails are available on all Portkey plans.PRO Guardrails are available on Portkey Pro & Enterprise plans.BASIC — Deterministic Guardrails
Regex Match
Checks if the request or response text matches a regex pattern.
Parameters: rule:
stringSupported On:
input_guardrails, output_guardrailsSentence Count
Checks if the content contains a certain number of sentences. Ranges allowed.
Parameters: minSentences:
number, maxSentences: numberSupported On:
input_guardrails, output_guardrailsWord Count
Checks if the content contains a certain number of words. Ranges allowed.
Parameters: minWords:
number, maxWords: numberSupported On:
input_guardrails, output_guardrailsCharacter Count
Checks if the content contains a certain number of characters. Ranges allowed.
Parameters: minCharacters:
number, maxCharacters: numberSupported On:
input_guardrails, output_guardrailsJSON Schema
Check if the response JSON matches a JSON schema.
Parameters: schema:
jsonSupported On:
output_guardrails onlyJSON Keys
Check if the response JSON contains any, all or none of the mentioned keys.
Parameters: keys:
array, operator: stringSupported On:
output_guardrails onlyContains
Checks if the content contains any, all or none of the words or phrases.
Parameters: words:
array, operator: stringSupported On:
output_guardrails onlyValid URLs
Checks if all the URLs mentioned in the content are valid
Parameters: onlyDNS:
booleanSupported On:
output_guardrails onlyContains Code
Checks if the content contains code of format SQL, Python, TypeScript, etc.
Parameters: format:
stringSupported On:
output_guardrails onlyLowercase Detection
Check if the given string is lowercase or not.
Parameters: format:
stringSupported On:
input_guardrails, output_guardrailsEnds With
Check if the content ends with a specified string.
Parameters: Suffix:
stringSupported On:
input_guardrails, output_guardrailsWebhook
Makes a webhook request for custom guardrails
Parameters: webhookURL:
string, headers: jsonSupported On:
input_guardrails, output_guardrailsJWT Token Validator
Check if the JWT token is valid.
Parameters: JWKS URI:
string,
Header Key: string,
Cache Max Age: number,
Clock Tolerance: number,
Max Token Age: number (in seconds)Supported On: input_guardrailsModel Whitelist
Check if the inference model to be used is in the whitelist.
Parameters: Models:
array, Inverse: booleanSupported On: input_guardrailsPRO — LLM Guardrails
Moderate Content
Checks if the content passes the mentioned content moderation checks.
Parameters: categories:
arraySupported On:
input_guardrails onlyCheck Language
Checks if the response content is in the mentioned language.
Parameters: language:
stringSupported On:
input_guardrails onlyDetect PII
Detects Personally Identifiable Information (PII) in the content.
Parameters: categories:
arraySupported On:
input_guardrails, output_guardrailsDetect Gibberish
Detects if the content is gibberish.
Parameters:
booleanSupported On:
input_guardrails, output_guardrailsYou can now have configurable timeouts for Partner & Pro Guardrails!