Private AI Integration

Private AI allows you to detect and mask Personally Identifiable Information (PII) in your data. This integration enables NeMo Guardrails to use Private AI for PII detection and masking in input, output, and retrieval flows.

Setup

1. Ensure that you have access to Private AI API server running locally or in the cloud. To get started with the cloud version, you can use the Private AI Portal. For containerized deployments, check out this Quickstart Guide.

2. Update your config.yml file to include the Private AI settings:

PII detection config

1
rails:
2
  config:
3
    privateai:
4
      server_endpoint: http://your-privateai-api-endpoint/process/text  # Replace this with your Private AI process text endpoint
5
      input:
6
        entities:  # If no entity is specified here, all supported entities will be detected by default.
7
          - NAME_FAMILY
8
          - LOCATION_ADDRESS_STREET
9
          - EMAIL_ADDRESS
10
      output:
11
        entities:
12
          - NAME_FAMILY
13
          - LOCATION_ADDRESS_STREET
14
          - EMAIL_ADDRESS
15
  input:
16
    flows:
17
      - detect pii on input
18
  output:
19
    flows:
20
      - detect pii on output

The detection flow will not let the input/output/retrieval text pass if PII is detected.

PII masking config

1
rails:
2
  config:
3
    privateai:
4
      server_endpoint: http://your-privateai-api-endpoint/process/text  # Replace this with your Private AI process text endpoint
5
      input:
6
        entities:  # If no entity is specified here, all supported entities will be detected by default.
7
          - NAME_FAMILY
8
          - LOCATION_ADDRESS_STREET
9
          - EMAIL_ADDRESS
10
      output:
11
        entities:
12
          - NAME_FAMILY
13
          - LOCATION_ADDRESS_STREET
14
          - EMAIL_ADDRESS
15
  input:
16
    flows:
17
      - mask pii on input
18
  output:
19
    flows:
20
      - mask pii on output

The masking flow will mask the PII in the input/output/retrieval text before they are sent to the LLM/user. For example, Hi John Doe, my email is john.doe@example.com will be converted to Hi [NAME], my email is [EMAIL_ADDRESS].

Replace http://your-privateai-api-endpoint/process/text with your actual Private AI process text endpoint and set the PAI_API_KEY environment variable if you’re using the Private AI cloud API.

3. You can customize the entities list under both input and output to include the PII types you want to detect. A full list of supported entities can be found here.

Usage

Once configured, the Private AI integration can automatically:

1. Detect or mask PII in user inputs before they are processed by the LLM.
2. Detect or mask PII in LLM outputs before they are sent back to the user.
3. Detect or mask PII in retrieved chunks before they are sent to the LLM.

The detect_pii and mask_pii actions in nemoguardrails/library/privateai/actions.py handle the PII detection and masking processes, respectively.

Customization

You can customize the PII detection behavior by modifying the entities lists in the config.yml file. Refer to the Private AI documentation for a complete list of supported entity types.

Error Handling

If the Private AI detection API request fails, the system will assume PII is present as a precautionary measure.

Notes

• Ensure that your Private AI process text endpoint is properly set up and accessible from your NeMo Guardrails environment.
• The integration currently supports PII detection and masking.

For more information on Private AI and its capabilities, please refer to the Private AI documentation.