# Allow Group-Level Opt-Out of Anthropic Prompt Caching ## Overview This epic implements functionality allowing GitLab group administrators to opt out of Anthropic prompt caching at the group level, providing more granular control over AI-related data handling. ## Proposed Solution Implement a group-level setting that allows group administrators to disable Anthropic prompt caching for their specific group. This setting will override the instance-level configuration for that group only. ## Requirements - Add a new toggle in the group settings UI under AI/ML settings - Implement backend logic to respect group-level opt-out preference - Ensure the setting properly cascades to all projects within the group - Add appropriate API endpoints for managing this setting - Update relevant documentation - Add audit logging for setting changes ## Technical Considerations - Database schema updates needed for storing group preferences - Cache invalidation strategy for existing prompts when setting is toggled - Performance impact assessment for checking group-level settings - Security reviews for permission models ## Success Criteria - Group administrators can successfully opt out of prompt caching - Setting correctly applies to all projects within the group - No impact on other groups' settings - Proper audit trail of setting changes - Documentation updated to reflect new functionality ## Diagram ```mermaid sequenceDiagram actor GA as Group Admin participant UI as GitLab UI participant BE as Backend API participant DB as Database participant AI Gateway GA->>UI: Navigate to Group Settings UI->>BE: GET /groups/:id/settings BE->>DB: Fetch current group settings DB-->>BE: Return settings BE-->>UI: Display settings page GA->>UI: Toggle AI Prompt Cache UI->>BE: PATCH /groups/:id/settings/ai BE->>BE: Validate permissions BE->>DB: Update group settings BE->> AI Gateway : Send payload excluding/include caching AI Gateway -->>BE: Confirmation DB-->>BE: Update successful BE-->>UI: Success response UI-->>GA: Show success notification ```