> ## Documentation Index
> Fetch the complete documentation index at: https://docs.plaisolutions.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Guardrails

> API endpoints for managing AI safety and content filtering guardrails

## List Available Guardrails

<api-endpoint method="GET" path="/guardrails" />

List all available guardrails that can be applied to agents for content filtering and AI safety.

### Response

Returns an array of available guardrail configurations that can be applied to agents.

<Note>
  Guardrails help ensure AI safety by filtering inappropriate content, preventing harmful outputs, and enforcing content policies.
</Note>

<CodeGroup>
  ```bash cURL theme={null}
  curl --location --request GET 'https://api.plaisolutions.com/guardrails' \
  --header 'Authorization: Bearer YOUR_TOKEN'
  ```

  ```javascript JavaScript theme={null}
  const response = await fetch('https://api.plaisolutions.com/guardrails', {
    method: 'GET',
    headers: {
      'Authorization': 'Bearer YOUR_TOKEN',
      'Content-Type': 'application/json'
    }
  });

  const guardrails = await response.json();
  console.log(guardrails);
  ```

  ```python Python theme={null}
  import requests

  url = "https://api.plaisolutions.com/guardrails"
  headers = {
      "Authorization": "Bearer YOUR_TOKEN",
      "Content-Type": "application/json"
  }

  response = requests.get(url, headers=headers)
  guardrails = response.json()
  print(guardrails)
  ```
</CodeGroup>

***

## Add Guardrail to Agent

<api-endpoint method="POST" path="/agents/{agent_id}/guardrails" />

Add a guardrail to an agent to implement content filtering and AI safety measures.

### Path Parameters

<ParamField path="agent_id" type="string" required>
  The unique identifier of the agent
</ParamField>

### Request Body

<ParamField body="id" type="string" required>
  The unique identifier of the guardrail to add
</ParamField>

<ParamField body="priority" type="integer" required>
  Priority level for this guardrail (lower numbers = higher priority)

  * `1-10` - High priority (critical safety measures)
  * `11-50` - Medium priority (content filtering)
  * `51-100` - Low priority (style and tone guidance)
</ParamField>

### Response

Returns a confirmation of the guardrail addition.

<CodeGroup>
  ```bash cURL theme={null}
  curl --location --request POST 'https://api.plaisolutions.com/agents/agent_123/guardrails' \
  --header 'Authorization: Bearer YOUR_TOKEN' \
  --header 'Content-Type: application/json' \
  --data-raw '{
    "id": "content_filter_v1",
    "priority": 5
  }'
  ```

  ```javascript JavaScript theme={null}
  const agentId = 'agent_123';
  const response = await fetch(`https://api.plaisolutions.com/agents/${agentId}/guardrails`, {
    method: 'POST',
    headers: {
      'Authorization': 'Bearer YOUR_TOKEN',
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      id: "content_filter_v1",
      priority: 5
    })
  });

  const result = await response.json();
  console.log(result);
  ```

  ```python Python theme={null}
  import requests

  agent_id = "agent_123"
  url = f"https://api.plaisolutions.com/agents/{agent_id}/guardrails"
  headers = {
      "Authorization": "Bearer YOUR_TOKEN",
      "Content-Type": "application/json"
  }
  data = {
      "id": "content_filter_v1",
      "priority": 5
  }

  response = requests.post(url, headers=headers, json=data)
  result = response.json()
  print(result)
  ```
</CodeGroup>

***

## List Agent Guardrails

<api-endpoint method="GET" path="/agents/{agent_id}/guardrails" />

List all guardrails currently applied to a specific agent.

### Path Parameters

<ParamField path="agent_id" type="string" required>
  The unique identifier of the agent
</ParamField>

### Response

Returns an array of guardrail configurations currently active for the agent, ordered by priority.

<CodeGroup>
  ```bash cURL theme={null}
  curl --location --request GET 'https://api.plaisolutions.com/agents/agent_123/guardrails' \
  --header 'Authorization: Bearer YOUR_TOKEN'
  ```

  ```javascript JavaScript theme={null}
  const agentId = 'agent_123';
  const response = await fetch(`https://api.plaisolutions.com/agents/${agentId}/guardrails`, {
    method: 'GET',
    headers: {
      'Authorization': 'Bearer YOUR_TOKEN',
      'Content-Type': 'application/json'
    }
  });

  const agentGuardrails = await response.json();
  console.log(agentGuardrails);
  ```

  ```python Python theme={null}
  import requests

  agent_id = "agent_123"
  url = f"https://api.plaisolutions.com/agents/{agent_id}/guardrails"
  headers = {
      "Authorization": "Bearer YOUR_TOKEN",
      "Content-Type": "application/json"
  }

  response = requests.get(url, headers=headers)
  agent_guardrails = response.json()
  print(agent_guardrails)
  ```
</CodeGroup>

***

## Remove Guardrail from Agent

<api-endpoint method="DELETE" path="/agents/{agent_id}/guardrails/{guardrail_id}" />

Remove a specific guardrail from an agent.

### Path Parameters

<ParamField path="agent_id" type="string" required>
  The unique identifier of the agent
</ParamField>

<ParamField path="guardrail_id" type="string" required>
  The unique identifier of the guardrail to remove
</ParamField>

### Response

Returns a confirmation of guardrail removal.

<Warning>
  Removing guardrails may reduce AI safety measures. Ensure you understand the implications before removing critical safety guardrails.
</Warning>

<CodeGroup>
  ```bash cURL theme={null}
  curl --location --request DELETE 'https://api.plaisolutions.com/agents/agent_123/guardrails/content_filter_v1' \
  --header 'Authorization: Bearer YOUR_TOKEN'
  ```

  ```javascript JavaScript theme={null}
  const agentId = 'agent_123';
  const guardrailId = 'content_filter_v1';
  const response = await fetch(`https://api.plaisolutions.com/agents/${agentId}/guardrails/${guardrailId}`, {
    method: 'DELETE',
    headers: {
      'Authorization': 'Bearer YOUR_TOKEN',
      'Content-Type': 'application/json'
    }
  });

  const result = await response.json();
  console.log('Guardrail removed:', result);
  ```

  ```python Python theme={null}
  import requests

  agent_id = "agent_123"
  guardrail_id = "content_filter_v1"
  url = f"https://api.plaisolutions.com/agents/{agent_id}/guardrails/{guardrail_id}"
  headers = {
      "Authorization": "Bearer YOUR_TOKEN",
      "Content-Type": "application/json"
  }

  response = requests.delete(url, headers=headers)
  result = response.json()
  print("Guardrail removed:", result)
  ```
</CodeGroup>

***

## Guardrail Types and Categories

Understanding the different types of guardrails available for AI safety and content control:

### Content Safety Guardrails

<AccordionGroup>
  <Accordion title="Harmful Content Filter">
    Prevents generation of harmful, toxic, or dangerous content

    ```json theme={null}
    {
      "id": "harmful_content_filter",
      "name": "Harmful Content Filter",
      "category": "safety",
      "description": "Blocks harmful, toxic, or dangerous content generation",
      "recommended_priority": 1
    }
    ```
  </Accordion>

  <Accordion title="Personal Information Protection">
    Prevents sharing of personal identifiable information (PII)

    ```json theme={null}
    {
      "id": "pii_protection",
      "name": "PII Protection",
      "category": "privacy", 
      "description": "Prevents sharing of personal identifiable information",
      "recommended_priority": 2
    }
    ```
  </Accordion>

  <Accordion title="Profanity Filter">
    Filters inappropriate language and profanity

    ```json theme={null}
    {
      "id": "profanity_filter",
      "name": "Profanity Filter",
      "category": "content",
      "description": "Filters inappropriate language and profanity",
      "recommended_priority": 10
    }
    ```
  </Accordion>
</AccordionGroup>

### Business Policy Guardrails

<AccordionGroup>
  <Accordion title="Brand Guidelines">
    Ensures responses align with brand voice and guidelines

    ```json theme={null}
    {
      "id": "brand_guidelines",
      "name": "Brand Guidelines",
      "category": "business",
      "description": "Maintains brand voice and messaging consistency",
      "recommended_priority": 20
    }
    ```
  </Accordion>

  <Accordion title="Legal Compliance">
    Prevents responses that could create legal liability

    ```json theme={null}
    {
      "id": "legal_compliance",
      "name": "Legal Compliance",
      "category": "compliance",
      "description": "Ensures responses comply with legal requirements",
      "recommended_priority": 3
    }
    ```
  </Accordion>

  <Accordion title="Industry Regulations">
    Enforces industry-specific regulatory compliance

    ```json theme={null}
    {
      "id": "industry_regulations",
      "name": "Industry Regulations", 
      "category": "compliance",
      "description": "Enforces sector-specific regulatory requirements",
      "recommended_priority": 5
    }
    ```
  </Accordion>
</AccordionGroup>

### Quality and Style Guardrails

<AccordionGroup>
  <Accordion title="Response Length Control">
    Controls the length and verbosity of responses

    ```json theme={null}
    {
      "id": "response_length_control",
      "name": "Response Length Control",
      "category": "quality",
      "description": "Ensures appropriate response length and conciseness",
      "recommended_priority": 30
    }
    ```
  </Accordion>

  <Accordion title="Factual Accuracy Check">
    Verifies factual claims and prevents misinformation

    ```json theme={null}
    {
      "id": "factual_accuracy",
      "name": "Factual Accuracy Check",
      "category": "quality",
      "description": "Verifies factual claims and prevents misinformation",
      "recommended_priority": 8
    }
    ```
  </Accordion>

  <Accordion title="Language Appropriateness">
    Ensures appropriate language level and tone

    ```json theme={null}
    {
      "id": "language_appropriateness",
      "name": "Language Appropriateness",
      "category": "style",
      "description": "Maintains appropriate language level and professional tone",
      "recommended_priority": 25
    }
    ```
  </Accordion>
</AccordionGroup>

***

## Guardrail Management Best Practices

<Steps>
  <Step title="Start with Essential Safety">
    Begin with critical safety guardrails (harmful content, PII protection) at high priority
  </Step>

  <Step title="Layer by Priority">
    Add guardrails in priority order, with safety measures taking precedence over style preferences
  </Step>

  <Step title="Test Thoroughly">
    Test agent responses after adding guardrails to ensure they work as expected
  </Step>

  <Step title="Monitor Performance">
    Monitor how guardrails affect response quality and user satisfaction
  </Step>

  <Step title="Regular Review">
    Periodically review and update guardrail configurations based on usage patterns
  </Step>
</Steps>

### Priority Guidelines

<CardGroup cols={2}>
  <Card title="High Priority (1-10)" icon="shield-alert">
    **Critical Safety Measures**

    * Harmful content prevention
    * PII protection
    * Legal compliance
    * Industry regulations
  </Card>

  <Card title="Medium Priority (11-50)" icon="filter">
    **Content Quality Control**

    * Profanity filtering
    * Brand guidelines
    * Factual accuracy
    * Bias prevention
  </Card>

  <Card title="Low Priority (51-100)" icon="edit">
    **Style and Preferences**

    * Response length control
    * Tone adjustment
    * Language level
    * Formatting preferences
  </Card>
</CardGroup>

### Configuration Examples

<Tip>
  **Recommended Guardrail Stack for Customer Service:**

  ```json theme={null}
  [
    {"id": "harmful_content_filter", "priority": 1},
    {"id": "pii_protection", "priority": 2},
    {"id": "legal_compliance", "priority": 3},
    {"id": "brand_guidelines", "priority": 15},
    {"id": "professional_tone", "priority": 20},
    {"id": "response_length_control", "priority": 30}
  ]
  ```
</Tip>

### Implementation Strategy

```javascript theme={null}
// Example: Setting up guardrails for a new agent
async function configureAgentGuardrails(agentId, useCase) {
  const guardrailConfigs = {
    'customer_service': [
      {id: 'harmful_content_filter', priority: 1},
      {id: 'pii_protection', priority: 2},
      {id: 'brand_guidelines', priority: 15},
      {id: 'professional_tone', priority: 20}
    ],
    'content_creation': [
      {id: 'harmful_content_filter', priority: 1},
      {id: 'copyright_protection', priority: 5},
      {id: 'factual_accuracy', priority: 8},
      {id: 'creative_guidelines', priority: 25}
    ],
    'technical_support': [
      {id: 'harmful_content_filter', priority: 1},
      {id: 'technical_accuracy', priority: 5},
      {id: 'solution_focused', priority: 10}
    ]
  };
  
  const configs = guardrailConfigs[useCase] || guardrailConfigs['customer_service'];
  
  for (const config of configs) {
    await fetch(`/agents/${agentId}/guardrails`, {
      method: 'POST',
      headers: {'Authorization': 'Bearer YOUR_TOKEN'},
      body: JSON.stringify(config)
    });
  }
}
```

### Monitoring and Analytics

<Note>
  **Guardrail Effectiveness Metrics:**

  * **Trigger Rate**: How often guardrails activate
  * **False Positives**: Safe content incorrectly filtered
  * **User Satisfaction**: Impact on user experience
  * **Response Quality**: Overall response appropriateness
  * **Performance Impact**: Effect on response time
</Note>

### Troubleshooting Common Issues

<AccordionGroup>
  <Accordion title="Over-Filtering">
    **Problem**: Guardrails are too strict, blocking appropriate content

    **Solutions**:

    * Review guardrail priorities and adjust as needed
    * Consider less restrictive guardrail variants
    * Add exceptions for specific use cases
    * Test with diverse content samples
  </Accordion>

  <Accordion title="Under-Filtering">
    **Problem**: Inappropriate content is still getting through

    **Solutions**:

    * Add additional relevant guardrails
    * Increase priority of existing safety guardrails
    * Review guardrail configuration for gaps
    * Consider custom guardrail rules
  </Accordion>

  <Accordion title="Performance Impact">
    **Problem**: Guardrails are slowing down response times

    **Solutions**:

    * Optimize guardrail priority ordering
    * Remove redundant or low-value guardrails
    * Consider batching guardrail checks
    * Monitor and profile individual guardrail performance
  </Accordion>
</AccordionGroup>

### Custom Guardrail Development

<Warning>
  **Enterprise Features:**

  * Custom guardrail development available for enterprise customers
  * Industry-specific compliance packages
  * Advanced configuration options
  * Dedicated support and consultation
  * Contact support for custom guardrail requirements
</Warning>

### Compliance and Governance

<CardGroup cols={2}>
  <Card title="Audit Trail" icon="file-text">
    All guardrail changes are logged for compliance and governance requirements
  </Card>

  <Card title="Version Control" icon="git-branch">
    Guardrail configurations can be versioned and rolled back if needed
  </Card>

  <Card title="Policy Enforcement" icon="gavel">
    Centralized policy management across all agents in your organization
  </Card>

  <Card title="Reporting" icon="chart-bar">
    Detailed reporting on guardrail effectiveness and usage patterns
  </Card>
</CardGroup>