[ENH]: Implement guardrail service with APIs to effectively manage guardrails in the system.

[vinayakbagal7]

Contact Details

[email protected]

Feature Description

Requirements:

1. Develop APIs to handle all CRUD operations for guardrail connections.

• Implement encryption support for Guardrail connections to ensure secure communication on the PAIG Cloud platform.
• Create a unique private-public key pair for each tenant.
• Use the public key for encrypting sensitive data during guardrail operations.
• Ensure the private key is securely stored and used for decryption on the server side.

2. Implement a new API to test connections to external services.

• API should check connection status.
• Return success or failure response.

3. Develop APIs to handle all CRUD operations for guardrails.

• Develop a flexible and scalable implementation for Bedrock guardrails creation, allowing future integration of various guardrails providers.
• Implement a guardrail functionality in PAIG to allow users to create, manage, and apply guardrails independently of individual applications. This will improve consistency, reduce redundancy, and expand the scope of protection to include various types of sensitive content.
• Users can create guardrails independent of applications.
• Users can associate multiple applications with a single guardrail.
• Users can configure customized responses for each guardrail configuration types

4. Provide a way to create different types of configurations listed below in the guardrail.

• Content Moderation - This can be configured with predefined categories and sensitivity levels (none, low, medium, high) using a sliding scale.
• Off Topics filters - Implement support in the Guardrail configuration to allow users to define and manage denied topics, enabling more granular control over restricted content.
• Denied Terms - Provide option to add specific words to guardrails. With this, the user can control the vocabulary used in applications and prevent the use of inappropriate terms.
• Sensitive Data Filtering - Sensitive information checks can be configured to include PII, sensitive information categories, and regex filters. Users can set the behavior for sensitive data guardrails to allow, mask, or deny.
• Prompt Safety - Add support for Prompt Safety in Guardrail configurations to mitigate prompt attacks by enforcing restrictions and safeguards on input prompts.

5. Create a CRUD API to manage response message templates.

• API allows creation, reading, updating, and deleting templates.
• Templates are stored and retrieved correctly.

6. Implement the "Get Guardrail" API specifically for the shield service.

• Ensure guardrail versioning is taken into account.
• Note that the shield service requires only the applicationKey as input parameters.

• APIs should adhere to RESTful standards.
• Unit tests must be included to validate functionality.

Code of Conduct

• I agree to follow this project's Code of Conduct