AI Semantic Prompt Guard: Only deny messages about a specific topic - Plugin

Only deny messages about a specific topicv3.8+

Only deny messages about a specific topic. For example, only deny messages about exploiting vulnerabilities.

For a detailed walkthrough, see Use AI Semantic Prompt Guard plugin to govern your LLM traffic.

If you use the text-embedding-ada-002 as an embedding model, you must set a fixed dimension of 1536, as required by the official model specification. Alternatively, use the text-embedding-3-small model, which supports dynamic dimensions and works without specifying a fixed value.

Prerequisites

AI Proxy plugin or AI Proxy Advanced plugin configured with an LLM service.
A Redis instance.
Port 6379, or your custom Redis port is open and reachable from Kong Gateway.

Environment variables

OPENAI_API_KEY: Your OpenAI API key
REDIS_HOST: The host where your Redis instance runs

Set up the plugin

Add this section to your declarative configuration file:

_format_version: "3.0"
plugins:
  - name: ai-semantic-prompt-guard
    config:
      embeddings:
        auth:
          header_name: Authorization
          header_value: Bearer ${{ env "DECK_OPENAI_API_KEY" }}
        model:
          name: text-embedding-3-small
          provider: openai
      search:
        threshold: 0.7
      vectordb:
        strategy: redis
        distance_metric: cosine
        threshold: 0.5
        dimensions: 1024
        redis:
          host: ${{ env "DECK_REDIS_HOST" }}
          port: 6379
      rules:
        match_all_conversation_history: true
        deny_prompts:
        - Hacking techniques or penetration testing without authorization
        - Bypassing software licensing or digital rights management
        - Instructions on exploiting vulnerabilities or writing malware
        - Circumventing security controls or access restrictions
        - Gathering personal or confidential employee information
        - Using AI to impersonate or phish others
        - Social engineering tactics or manipulation techniques
        - Guidance on violating company IT policies
        - Content unrelated to work, such as entertainment or dating
        - Political, religious, or sensitive non-work-related discussions

Make the following request:

curl -i -X POST http://localhost:8001/plugins/ \
    --header "Accept: application/json" \
    --header "Content-Type: application/json" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make the following request:

curl -X POST https://{region}.api.konghq.com/v2/control-planes/{controlPlaneId}/core-entities/plugins/ \
    --header "accept: application/json" \
    --header "Content-Type: application/json" \
    --header "Authorization: Bearer $KONNECT_TOKEN" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

region: Geographic region where your Kong Konnect is hosted and operates.
controlPlaneId: The id of the control plane.
KONNECT_TOKEN: Your Personal Access Token (PAT) associated with your Konnect account.

See the Konnect API reference to learn about region-specific URLs and personal access tokens.

echo "
apiVersion: configuration.konghq.com/v1
kind: KongClusterPlugin
metadata:
  name: ai-semantic-prompt-guard
  namespace: kong
  annotations:
    kubernetes.io/ingress.class: kong
  labels:
    global: 'true'
config:
  embeddings:
    auth:
      header_name: Authorization
      header_value: Bearer $OPENAI_API_KEY
    model:
      name: text-embedding-3-small
      provider: openai
  search:
    threshold: 0.7
  vectordb:
    strategy: redis
    distance_metric: cosine
    threshold: 0.5
    dimensions: 1024
    redis:
      host: '$REDIS_HOST'
      port: 6379
  rules:
    match_all_conversation_history: true
    deny_prompts:
    - Hacking techniques or penetration testing without authorization
    - Bypassing software licensing or digital rights management
    - Instructions on exploiting vulnerabilities or writing malware
    - Circumventing security controls or access restrictions
    - Gathering personal or confidential employee information
    - Using AI to impersonate or phish others
    - Social engineering tactics or manipulation techniques
    - Guidance on violating company IT policies
    - Content unrelated to work, such as entertainment or dating
    - Political, religious, or sensitive non-work-related discussions
plugin: ai-semantic-prompt-guard
" | kubectl apply -f -

Prerequisite: Configure your Personal Access Token

terraform {
  required_providers {
    konnect = {
      source  = "kong/konnect"
    }
  }
}

provider "konnect" {
  personal_access_token = "$KONNECT_TOKEN"
  server_url            = "https://us.api.konghq.com/"
}

Add the following to your Terraform configuration to create a Konnect Gateway Plugin:

resource "konnect_gateway_plugin_ai_semantic_prompt_guard" "my_ai_semantic_prompt_guard" {
  enabled = true

  config = {

    embeddings = {

      auth = {
        header_name = "Authorization"
        header_value = "Bearer var.openai_api_key"
      }

      model = {
        name = "text-embedding-3-small"
        provider = "openai"
      }
    }

    search = {
      threshold = 0.7
    }

    vectordb = {
      strategy = "redis"
      distance_metric = "cosine"
      threshold = 0.5
      dimensions = 1024

      redis = {
        host = var.redis_host
        port = 6379
      }
    }

    rules = {
      match_all_conversation_history = true
      deny_prompts = ["Hacking techniques or penetration testing without authorization", "Bypassing software licensing or digital rights management", "Instructions on exploiting vulnerabilities or writing malware", "Circumventing security controls or access restrictions", "Gathering personal or confidential employee information", "Using AI to impersonate or phish others", "Social engineering tactics or manipulation techniques", "Guidance on violating company IT policies", "Content unrelated to work, such as entertainment or dating", "Political, religious, or sensitive non-work-related discussions"]
    }
  }

  control_plane_id = konnect_gateway_control_plane.my_konnect_cp.id
}

This example requires the following variables to be added to your manifest. You can specify values at runtime by setting TF_VAR_name=value.

variable "redis_host" {
  type = string
}

Add this section to your declarative configuration file:

_format_version: "3.0"
plugins:
  - name: ai-semantic-prompt-guard
    service: serviceName|Id
    config:
      embeddings:
        auth:
          header_name: Authorization
          header_value: Bearer ${{ env "DECK_OPENAI_API_KEY" }}
        model:
          name: text-embedding-3-small
          provider: openai
      search:
        threshold: 0.7
      vectordb:
        strategy: redis
        distance_metric: cosine
        threshold: 0.5
        dimensions: 1024
        redis:
          host: ${{ env "DECK_REDIS_HOST" }}
          port: 6379
      rules:
        match_all_conversation_history: true
        deny_prompts:
        - Hacking techniques or penetration testing without authorization
        - Bypassing software licensing or digital rights management
        - Instructions on exploiting vulnerabilities or writing malware
        - Circumventing security controls or access restrictions
        - Gathering personal or confidential employee information
        - Using AI to impersonate or phish others
        - Social engineering tactics or manipulation techniques
        - Guidance on violating company IT policies
        - Content unrelated to work, such as entertainment or dating
        - Political, religious, or sensitive non-work-related discussions

Make sure to replace the following placeholders with your own values:

serviceName|Id: The id or name of the service the plugin configuration will target.

Make the following request:

curl -i -X POST http://localhost:8001/services/{serviceName|Id}/plugins/ \
    --header "Accept: application/json" \
    --header "Content-Type: application/json" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

serviceName|Id: The id or name of the service the plugin configuration will target.

Make the following request:

curl -X POST https://{region}.api.konghq.com/v2/control-planes/{controlPlaneId}/core-entities/services/{serviceId}/plugins/ \
    --header "accept: application/json" \
    --header "Content-Type: application/json" \
    --header "Authorization: Bearer $KONNECT_TOKEN" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

region: Geographic region where your Kong Konnect is hosted and operates.
controlPlaneId: The id of the control plane.
KONNECT_TOKEN: Your Personal Access Token (PAT) associated with your Konnect account.
serviceId: The id of the service the plugin configuration will target.

See the Konnect API reference to learn about region-specific URLs and personal access tokens.

echo "
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: ai-semantic-prompt-guard
  namespace: kong
  annotations:
    kubernetes.io/ingress.class: kong
config:
  embeddings:
    auth:
      header_name: Authorization
      header_value: Bearer $OPENAI_API_KEY
    model:
      name: text-embedding-3-small
      provider: openai
  search:
    threshold: 0.7
  vectordb:
    strategy: redis
    distance_metric: cosine
    threshold: 0.5
    dimensions: 1024
    redis:
      host: '$REDIS_HOST'
      port: 6379
  rules:
    match_all_conversation_history: true
    deny_prompts:
    - Hacking techniques or penetration testing without authorization
    - Bypassing software licensing or digital rights management
    - Instructions on exploiting vulnerabilities or writing malware
    - Circumventing security controls or access restrictions
    - Gathering personal or confidential employee information
    - Using AI to impersonate or phish others
    - Social engineering tactics or manipulation techniques
    - Guidance on violating company IT policies
    - Content unrelated to work, such as entertainment or dating
    - Political, religious, or sensitive non-work-related discussions
plugin: ai-semantic-prompt-guard
" | kubectl apply -f -

Next, apply the KongPlugin resource by annotating the service resource:

kubectl annotate -n kong service SERVICE_NAME konghq.com/plugins=ai-semantic-prompt-guard

Prerequisite: Configure your Personal Access Token

terraform {
  required_providers {
    konnect = {
      source  = "kong/konnect"
    }
  }
}

provider "konnect" {
  personal_access_token = "$KONNECT_TOKEN"
  server_url            = "https://us.api.konghq.com/"
}

Add the following to your Terraform configuration to create a Konnect Gateway Plugin:

resource "konnect_gateway_plugin_ai_semantic_prompt_guard" "my_ai_semantic_prompt_guard" {
  enabled = true

  config = {

    embeddings = {

      auth = {
        header_name = "Authorization"
        header_value = "Bearer var.openai_api_key"
      }

      model = {
        name = "text-embedding-3-small"
        provider = "openai"
      }
    }

    search = {
      threshold = 0.7
    }

    vectordb = {
      strategy = "redis"
      distance_metric = "cosine"
      threshold = 0.5
      dimensions = 1024

      redis = {
        host = var.redis_host
        port = 6379
      }
    }

    rules = {
      match_all_conversation_history = true
      deny_prompts = ["Hacking techniques or penetration testing without authorization", "Bypassing software licensing or digital rights management", "Instructions on exploiting vulnerabilities or writing malware", "Circumventing security controls or access restrictions", "Gathering personal or confidential employee information", "Using AI to impersonate or phish others", "Social engineering tactics or manipulation techniques", "Guidance on violating company IT policies", "Content unrelated to work, such as entertainment or dating", "Political, religious, or sensitive non-work-related discussions"]
    }
  }

  control_plane_id = konnect_gateway_control_plane.my_konnect_cp.id
  service = {
    id = konnect_gateway_service.my_service.id
  }
}

This example requires the following variables to be added to your manifest. You can specify values at runtime by setting TF_VAR_name=value.

variable "redis_host" {
  type = string
}

Add this section to your declarative configuration file:

_format_version: "3.0"
plugins:
  - name: ai-semantic-prompt-guard
    route: routeName|Id
    config:
      embeddings:
        auth:
          header_name: Authorization
          header_value: Bearer ${{ env "DECK_OPENAI_API_KEY" }}
        model:
          name: text-embedding-3-small
          provider: openai
      search:
        threshold: 0.7
      vectordb:
        strategy: redis
        distance_metric: cosine
        threshold: 0.5
        dimensions: 1024
        redis:
          host: ${{ env "DECK_REDIS_HOST" }}
          port: 6379
      rules:
        match_all_conversation_history: true
        deny_prompts:
        - Hacking techniques or penetration testing without authorization
        - Bypassing software licensing or digital rights management
        - Instructions on exploiting vulnerabilities or writing malware
        - Circumventing security controls or access restrictions
        - Gathering personal or confidential employee information
        - Using AI to impersonate or phish others
        - Social engineering tactics or manipulation techniques
        - Guidance on violating company IT policies
        - Content unrelated to work, such as entertainment or dating
        - Political, religious, or sensitive non-work-related discussions

Make sure to replace the following placeholders with your own values:

routeName|Id: The id or name of the route the plugin configuration will target.

Make the following request:

curl -i -X POST http://localhost:8001/routes/{routeName|Id}/plugins/ \
    --header "Accept: application/json" \
    --header "Content-Type: application/json" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

routeName|Id: The id or name of the route the plugin configuration will target.

Make the following request:

curl -X POST https://{region}.api.konghq.com/v2/control-planes/{controlPlaneId}/core-entities/routes/{routeId}/plugins/ \
    --header "accept: application/json" \
    --header "Content-Type: application/json" \
    --header "Authorization: Bearer $KONNECT_TOKEN" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

region: Geographic region where your Kong Konnect is hosted and operates.
controlPlaneId: The id of the control plane.
KONNECT_TOKEN: Your Personal Access Token (PAT) associated with your Konnect account.
routeId: The id of the route the plugin configuration will target.

See the Konnect API reference to learn about region-specific URLs and personal access tokens.

echo "
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: ai-semantic-prompt-guard
  namespace: kong
  annotations:
    kubernetes.io/ingress.class: kong
config:
  embeddings:
    auth:
      header_name: Authorization
      header_value: Bearer $OPENAI_API_KEY
    model:
      name: text-embedding-3-small
      provider: openai
  search:
    threshold: 0.7
  vectordb:
    strategy: redis
    distance_metric: cosine
    threshold: 0.5
    dimensions: 1024
    redis:
      host: '$REDIS_HOST'
      port: 6379
  rules:
    match_all_conversation_history: true
    deny_prompts:
    - Hacking techniques or penetration testing without authorization
    - Bypassing software licensing or digital rights management
    - Instructions on exploiting vulnerabilities or writing malware
    - Circumventing security controls or access restrictions
    - Gathering personal or confidential employee information
    - Using AI to impersonate or phish others
    - Social engineering tactics or manipulation techniques
    - Guidance on violating company IT policies
    - Content unrelated to work, such as entertainment or dating
    - Political, religious, or sensitive non-work-related discussions
plugin: ai-semantic-prompt-guard
" | kubectl apply -f -

Next, apply the KongPlugin resource by annotating the httproute or ingress resource:

kubectl annotate -n kong httproute  konghq.com/plugins=ai-semantic-prompt-guard

kubectl annotate -n kong ingress  konghq.com/plugins=ai-semantic-prompt-guard

Prerequisite: Configure your Personal Access Token

terraform {
  required_providers {
    konnect = {
      source  = "kong/konnect"
    }
  }
}

provider "konnect" {
  personal_access_token = "$KONNECT_TOKEN"
  server_url            = "https://us.api.konghq.com/"
}

Add the following to your Terraform configuration to create a Konnect Gateway Plugin:

resource "konnect_gateway_plugin_ai_semantic_prompt_guard" "my_ai_semantic_prompt_guard" {
  enabled = true

  config = {

    embeddings = {

      auth = {
        header_name = "Authorization"
        header_value = "Bearer var.openai_api_key"
      }

      model = {
        name = "text-embedding-3-small"
        provider = "openai"
      }
    }

    search = {
      threshold = 0.7
    }

    vectordb = {
      strategy = "redis"
      distance_metric = "cosine"
      threshold = 0.5
      dimensions = 1024

      redis = {
        host = var.redis_host
        port = 6379
      }
    }

    rules = {
      match_all_conversation_history = true
      deny_prompts = ["Hacking techniques or penetration testing without authorization", "Bypassing software licensing or digital rights management", "Instructions on exploiting vulnerabilities or writing malware", "Circumventing security controls or access restrictions", "Gathering personal or confidential employee information", "Using AI to impersonate or phish others", "Social engineering tactics or manipulation techniques", "Guidance on violating company IT policies", "Content unrelated to work, such as entertainment or dating", "Political, religious, or sensitive non-work-related discussions"]
    }
  }

  control_plane_id = konnect_gateway_control_plane.my_konnect_cp.id
  route = {
    id = konnect_gateway_route.my_route.id
  }
}

This example requires the following variables to be added to your manifest. You can specify values at runtime by setting TF_VAR_name=value.

variable "redis_host" {
  type = string
}

Add this section to your declarative configuration file:

_format_version: "3.0"
plugins:
  - name: ai-semantic-prompt-guard
    consumer: consumerName|Id
    config:
      embeddings:
        auth:
          header_name: Authorization
          header_value: Bearer ${{ env "DECK_OPENAI_API_KEY" }}
        model:
          name: text-embedding-3-small
          provider: openai
      search:
        threshold: 0.7
      vectordb:
        strategy: redis
        distance_metric: cosine
        threshold: 0.5
        dimensions: 1024
        redis:
          host: ${{ env "DECK_REDIS_HOST" }}
          port: 6379
      rules:
        match_all_conversation_history: true
        deny_prompts:
        - Hacking techniques or penetration testing without authorization
        - Bypassing software licensing or digital rights management
        - Instructions on exploiting vulnerabilities or writing malware
        - Circumventing security controls or access restrictions
        - Gathering personal or confidential employee information
        - Using AI to impersonate or phish others
        - Social engineering tactics or manipulation techniques
        - Guidance on violating company IT policies
        - Content unrelated to work, such as entertainment or dating
        - Political, religious, or sensitive non-work-related discussions

Make sure to replace the following placeholders with your own values:

consumerName|Id: The id or name of the consumer the plugin configuration will target.

Make the following request:

curl -i -X POST http://localhost:8001/consumers/{consumerName|Id}/plugins/ \
    --header "Accept: application/json" \
    --header "Content-Type: application/json" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

consumerName|Id: The id or name of the consumer the plugin configuration will target.

Make the following request:

curl -X POST https://{region}.api.konghq.com/v2/control-planes/{controlPlaneId}/core-entities/consumers/{consumerId}/plugins/ \
    --header "accept: application/json" \
    --header "Content-Type: application/json" \
    --header "Authorization: Bearer $KONNECT_TOKEN" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

region: Geographic region where your Kong Konnect is hosted and operates.
controlPlaneId: The id of the control plane.
KONNECT_TOKEN: Your Personal Access Token (PAT) associated with your Konnect account.
consumerId: The id of the consumer the plugin configuration will target.

See the Konnect API reference to learn about region-specific URLs and personal access tokens.

echo "
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: ai-semantic-prompt-guard
  namespace: kong
  annotations:
    kubernetes.io/ingress.class: kong
config:
  embeddings:
    auth:
      header_name: Authorization
      header_value: Bearer $OPENAI_API_KEY
    model:
      name: text-embedding-3-small
      provider: openai
  search:
    threshold: 0.7
  vectordb:
    strategy: redis
    distance_metric: cosine
    threshold: 0.5
    dimensions: 1024
    redis:
      host: '$REDIS_HOST'
      port: 6379
  rules:
    match_all_conversation_history: true
    deny_prompts:
    - Hacking techniques or penetration testing without authorization
    - Bypassing software licensing or digital rights management
    - Instructions on exploiting vulnerabilities or writing malware
    - Circumventing security controls or access restrictions
    - Gathering personal or confidential employee information
    - Using AI to impersonate or phish others
    - Social engineering tactics or manipulation techniques
    - Guidance on violating company IT policies
    - Content unrelated to work, such as entertainment or dating
    - Political, religious, or sensitive non-work-related discussions
plugin: ai-semantic-prompt-guard
" | kubectl apply -f -

Next, apply the KongPlugin resource by annotating the KongConsumer resource:

kubectl annotate -n kong  CONSUMER_NAME konghq.com/plugins=ai-semantic-prompt-guard

Prerequisite: Configure your Personal Access Token

terraform {
  required_providers {
    konnect = {
      source  = "kong/konnect"
    }
  }
}

provider "konnect" {
  personal_access_token = "$KONNECT_TOKEN"
  server_url            = "https://us.api.konghq.com/"
}

Add the following to your Terraform configuration to create a Konnect Gateway Plugin:

resource "konnect_gateway_plugin_ai_semantic_prompt_guard" "my_ai_semantic_prompt_guard" {
  enabled = true

  config = {

    embeddings = {

      auth = {
        header_name = "Authorization"
        header_value = "Bearer var.openai_api_key"
      }

      model = {
        name = "text-embedding-3-small"
        provider = "openai"
      }
    }

    search = {
      threshold = 0.7
    }

    vectordb = {
      strategy = "redis"
      distance_metric = "cosine"
      threshold = 0.5
      dimensions = 1024

      redis = {
        host = var.redis_host
        port = 6379
      }
    }

    rules = {
      match_all_conversation_history = true
      deny_prompts = ["Hacking techniques or penetration testing without authorization", "Bypassing software licensing or digital rights management", "Instructions on exploiting vulnerabilities or writing malware", "Circumventing security controls or access restrictions", "Gathering personal or confidential employee information", "Using AI to impersonate or phish others", "Social engineering tactics or manipulation techniques", "Guidance on violating company IT policies", "Content unrelated to work, such as entertainment or dating", "Political, religious, or sensitive non-work-related discussions"]
    }
  }

  control_plane_id = konnect_gateway_control_plane.my_konnect_cp.id
  consumer = {
    id = konnect_gateway_consumer.my_consumer.id
  }
}

This example requires the following variables to be added to your manifest. You can specify values at runtime by setting TF_VAR_name=value.

variable "redis_host" {
  type = string
}

Add this section to your declarative configuration file:

_format_version: "3.0"
plugins:
  - name: ai-semantic-prompt-guard
    consumer_group: consumerGroupName|Id
    config:
      embeddings:
        auth:
          header_name: Authorization
          header_value: Bearer ${{ env "DECK_OPENAI_API_KEY" }}
        model:
          name: text-embedding-3-small
          provider: openai
      search:
        threshold: 0.7
      vectordb:
        strategy: redis
        distance_metric: cosine
        threshold: 0.5
        dimensions: 1024
        redis:
          host: ${{ env "DECK_REDIS_HOST" }}
          port: 6379
      rules:
        match_all_conversation_history: true
        deny_prompts:
        - Hacking techniques or penetration testing without authorization
        - Bypassing software licensing or digital rights management
        - Instructions on exploiting vulnerabilities or writing malware
        - Circumventing security controls or access restrictions
        - Gathering personal or confidential employee information
        - Using AI to impersonate or phish others
        - Social engineering tactics or manipulation techniques
        - Guidance on violating company IT policies
        - Content unrelated to work, such as entertainment or dating
        - Political, religious, or sensitive non-work-related discussions

Make sure to replace the following placeholders with your own values:

consumerGroupName|Id: The id or name of the consumer group the plugin configuration will target.

Make the following request:

curl -i -X POST http://localhost:8001/consumer_groups/{consumerGroupName|Id}/plugins/ \
    --header "Accept: application/json" \
    --header "Content-Type: application/json" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

consumerGroupName|Id: The id or name of the consumer group the plugin configuration will target.

Make the following request:

curl -X POST https://{region}.api.konghq.com/v2/control-planes/{controlPlaneId}/core-entities/consumer_groups/{consumerGroupId}/plugins/ \
    --header "accept: application/json" \
    --header "Content-Type: application/json" \
    --header "Authorization: Bearer $KONNECT_TOKEN" \
    --data '
    {
      "name": "ai-semantic-prompt-guard",
      "config": {
        "embeddings": {
          "auth": {
            "header_name": "Authorization",
            "header_value": "Bearer '$OPENAI_API_KEY'"
          },
          "model": {
            "name": "text-embedding-3-small",
            "provider": "openai"
          }
        },
        "search": {
          "threshold": 0.7
        },
        "vectordb": {
          "strategy": "redis",
          "distance_metric": "cosine",
          "threshold": 0.5,
          "dimensions": 1024,
          "redis": {
            "host": "'$REDIS_HOST'",
            "port": 6379
          }
        },
        "rules": {
          "match_all_conversation_history": true,
          "deny_prompts": [
            "Hacking techniques or penetration testing without authorization",
            "Bypassing software licensing or digital rights management",
            "Instructions on exploiting vulnerabilities or writing malware",
            "Circumventing security controls or access restrictions",
            "Gathering personal or confidential employee information",
            "Using AI to impersonate or phish others",
            "Social engineering tactics or manipulation techniques",
            "Guidance on violating company IT policies",
            "Content unrelated to work, such as entertainment or dating",
            "Political, religious, or sensitive non-work-related discussions"
          ]
        }
      }
    }
    '

Make sure to replace the following placeholders with your own values:

region: Geographic region where your Kong Konnect is hosted and operates.
controlPlaneId: The id of the control plane.
KONNECT_TOKEN: Your Personal Access Token (PAT) associated with your Konnect account.
consumerGroupId: The id of the consumer group the plugin configuration will target.

See the Konnect API reference to learn about region-specific URLs and personal access tokens.

echo "
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: ai-semantic-prompt-guard
  namespace: kong
  annotations:
    kubernetes.io/ingress.class: kong
config:
  embeddings:
    auth:
      header_name: Authorization
      header_value: Bearer $OPENAI_API_KEY
    model:
      name: text-embedding-3-small
      provider: openai
  search:
    threshold: 0.7
  vectordb:
    strategy: redis
    distance_metric: cosine
    threshold: 0.5
    dimensions: 1024
    redis:
      host: '$REDIS_HOST'
      port: 6379
  rules:
    match_all_conversation_history: true
    deny_prompts:
    - Hacking techniques or penetration testing without authorization
    - Bypassing software licensing or digital rights management
    - Instructions on exploiting vulnerabilities or writing malware
    - Circumventing security controls or access restrictions
    - Gathering personal or confidential employee information
    - Using AI to impersonate or phish others
    - Social engineering tactics or manipulation techniques
    - Guidance on violating company IT policies
    - Content unrelated to work, such as entertainment or dating
    - Political, religious, or sensitive non-work-related discussions
plugin: ai-semantic-prompt-guard
" | kubectl apply -f -

Next, apply the KongPlugin resource by annotating the KongConsumerGroup resource:

kubectl annotate -n kong  CONSUMERGROUP_NAME konghq.com/plugins=ai-semantic-prompt-guard

Prerequisite: Configure your Personal Access Token

terraform {
  required_providers {
    konnect = {
      source  = "kong/konnect"
    }
  }
}

provider "konnect" {
  personal_access_token = "$KONNECT_TOKEN"
  server_url            = "https://us.api.konghq.com/"
}

Add the following to your Terraform configuration to create a Konnect Gateway Plugin:

resource "konnect_gateway_plugin_ai_semantic_prompt_guard" "my_ai_semantic_prompt_guard" {
  enabled = true

  config = {

    embeddings = {

      auth = {
        header_name = "Authorization"
        header_value = "Bearer var.openai_api_key"
      }

      model = {
        name = "text-embedding-3-small"
        provider = "openai"
      }
    }

    search = {
      threshold = 0.7
    }

    vectordb = {
      strategy = "redis"
      distance_metric = "cosine"
      threshold = 0.5
      dimensions = 1024

      redis = {
        host = var.redis_host
        port = 6379
      }
    }

    rules = {
      match_all_conversation_history = true
      deny_prompts = ["Hacking techniques or penetration testing without authorization", "Bypassing software licensing or digital rights management", "Instructions on exploiting vulnerabilities or writing malware", "Circumventing security controls or access restrictions", "Gathering personal or confidential employee information", "Using AI to impersonate or phish others", "Social engineering tactics or manipulation techniques", "Guidance on violating company IT policies", "Content unrelated to work, such as entertainment or dating", "Political, religious, or sensitive non-work-related discussions"]
    }
  }

  control_plane_id = konnect_gateway_control_plane.my_konnect_cp.id
  consumer_group = {
    id = konnect_gateway_consumer_group.my_consumer_group.id
  }
}

This example requires the following variables to be added to your manifest. You can specify values at runtime by setting TF_VAR_name=value.

variable "redis_host" {
  type = string
}

AI Semantic Prompt Guard

Only deny messages about a specific topicv3.8+

Prerequisites

Environment variables

Set up the plugin

Next Steps

Help us make these docs great!

Still need help