Application Health Assistant (AHA)

Objective

AHA (Application Health Assistant) is an AI-powered tool integrated with CAST Imaging and CAST Highlight that automates the detection and remediation of Green Vulnerabilities using Gen AI. It streamlines application health management by providing intelligent, context-aware fixes and actionable insights to enhance software quality.

Expectations

• Utilize the tool for all onboarded applications to effectively address GREEN deficiencies
• Capture application health and GREEN indicator scores from CAST Highlights (before/after running the tool)
• Identify the accuracy of the tool to ensure reliable results
• Evaluate efficiency gains for developers when fixing issues with the tool

Prerequisites

• GitHub account required
• Must have at least one of these access levels on the GitHub code repository:
  • Write
  • Maintain
  • Admin (Any one required to create pull requests)

Onboarding Requirements

• Application must have a repository on GitHub
• Application must be onboarded on CAST HL and Imaging
• AHA admin can then onboard the application in AHA

AHA Functionality

Accessible via Angular default port (4200): http://localhost:4200

1. Landing Page

Provides tool overview with 'Login with GitHub' button for authentication.

2. Authentication

GitHub used for authentication. Three scenarios:

• Not logged into GitHub: Prompts for username and device verification
• Logged into GitHub: Direct authentication
• Token expired: Requires re-authorization

3. Authorization

Users need:

• GitHub account
• Write/Maintain/Admin role on at least one onboarded GitHub project repository

Admin menu appears only for admin/super admin roles.

4. Home Screen

After authentication, shows:

• Application dropdown (only accessible repositories)
• Repository name, CAST HL branch, last scan dates
• Green Impact issue numbers in orange box

5. Issue/Objects Details

Clicking orange box shows:

• Detailed issue information
• Impacted objects (occurrences count)

6. Object Selection

• Select one/multiple objects for AI fixing
• 'Add to request' enables 'Fix' button
• Disabled objects re-enabled after CAST Imaging re-scan if not fixed

7. Request Status

Shows requested objects count vs total for each issue.

8. Fix with AI

• Select prompt
• Click 'Fix with AI' to send to AHA Engine
• Engine combines prompt with CAST Imaging dependencies
• Returns fix code via Core API OpenAI (Len AI API)

9. Fix & PR Details

Six statuses:

1. Response Queued
2. Response Success (PR Awaited)
3. Partial Success
4. Unmodified
5. Failure
6. Completed

Notes:

• PR only creatable for statuses 2 & 3
• Retry button appears for partial success/unmodified/failure
• Check details via 'Fix Objects' link
• Error messages visible via status links

Pull Request Process:

1. Creates new branch from base
2. Updates fixed code in new branch
3. Creates PR (manual merge required)

Whitespace Tip: If formatting issues occur, enable 'Hide whitespace' in GitHub PR settings.

10. Feedback

Users can submit feedback.

11. Admin Section (Admin/Super Admin only)

Three screens:

1. Application List: Onboard new apps (add/edit/delete)
2. Prompt List: Create/modify prompts
3. User Management: Manage user roles

Roles:

• Developer: Full site access (no admin tab)
• Admin: Access to admin tab (app onboarding/prompts)
• Super Admin: Admin tab + user management

Upcoming Enhancements

1. Retry feature for completed status with prompt selection
2. Custom prompt creation on prompt selection screen
3. Expand prompts library
4. Dashboard for application health/GREEN indicator status