Sentence Count Guardrail

The Sentence Count Guardrail is a custom Synapse mediator for the WSO2 API Manager Universal Gateway that validates incoming AI API requests and GenAI responses by enforcing minimum and/or maximum sentence count constraints on a selected JSON field. It helps control verbosity, reduce prompt injection surface via overly long inputs, and enforce style or compliance limits.

Features

• Validate payload content by counting sentences.
• Enforces configurable minimum and maximum sentence thresholds.
• Target specific fields in JSON payloads using JSON Path
• Optionally invert validation logic (e.g., allow only content outside the specified sentence range)
• Provides an optional Show Guardrail Assessment feature to include detailed validation results in error responses for enhanced observability.

How to Use

Follow these steps to integrate the Sentence Count Guardrail policy into your AI API:

• Open the API Publisher Portal (https://<host>:<port>/publisher)
• Select your AI API
• Go to Develop > API Configurations > Policies
• Expand Common Policies in the Policy List
• Drag and drop the Sentence Count Guardrail policy into your desired mediation flow
• Fill in the required policy configuration (name, minimum and maximum sentence count)
• Save and Deploy the AI API

Example Policy Configuration

Click to expand configuration steps

Example: Restrict user prompts to between 2 and 5 sentences.

1. Create an AI API using Mistral AI.
2. Add the Sentence Count Guardrail policy to the API with the following configuration:

Field	Example
Guardrail Name	Sentence Limiter
Minimum Sentence Count	2
Maximum Sentence Count	5
JSON Path	$.messages[-1].content
Invert the Guardrail Decision	false
Show Guardrail Assessment	false

1. Save and re-deploy the API.
2. Invoke the API's chat/completion endpoint with a prompt that violates the sentence count, such as having only one sentence (if min is set to 2):

Test Request (violates minimum of 2 sentences):

{
    "messages": [
        {
            "role": "user",
            "content": "This is a short prompt."
        }
    ]
}

The following guardrail error response will be returned with HTTP status code 446:

{
    "code": "900514",
    "type": "SENTENCE_COUNT_GUARDRAIL",
    "message": {
        "interveningGuardrail": "Sentence Limiter",
        "action": "GUARDRAIL_INTERVENED",
        "actionReason": "Violation of applied sentence count constraints detected.",
        "direction": "REQUEST"
    }
}

Limitations

The Sentence Count Guardrail uses the following regular expression to split sentences from the inspected content:

\{.*?\}

This pattern splits sentences based on punctuation marks such as periods, exclamation points, and question marks. This approach may not handle all sentence boundaries accurately, especially in cases involving abbreviations, decimals, or other complex sentence structures.