ADMINDASHBOARD_PRD.md

Product Requirement Document: Admin Dashboard & Manual Analysis Workflow

1. Overview

The goal is to create a comprehensive Admin Dashboard for "Vault" that allows operations admins to monitor key platform metrics (users, videos, products) and manage a "Wizard of Oz" style workflow for AI product analysis.

Currently, AI analysis can be tracked via technical logs, but this change shifts control to a user-friendly admin interface. It introduces a manual review step ("Wizard of Oz" method) where admins must approve or manually add products before they are visible to the end user.

2. User Roles & Permissions

• Admin: Users with emails listed in the ADMIN_EMAILS environment variable.
• User: Standard platform user (Creator/Viewer).

3. Features

3.1 Admin Dashboard (/admin/dashboard)

A central hub for monitoring platform health and activity.

Key Metrics:

1. User Stats:
   • Total registered users.
   • New users (last 24h/7d).
2. Video Stats:
   • Total videos submitted.
   • Status Breakdown: Pending Review, In Progress, Completed, Failed.
3. Product Stats:
   • Matched Products (With Links): Number of detected objects matched to a marketplace product.
   • Unmatched Products (Without Links): Detected objects pending a match.
   • Interest Pledges: Count of user "I want this" interactions.
   • Suggestions: Count of product suggestions from viewers (product requests) and affiliate proposals (suppliers).

3.2 Manual "AI" Analysis Workflow

Replace the fully automated trigger with a human-in-the-loop review process.

Workflow Steps:

1. User Trigger:

   • User clicks "Analyze Video" (or "Create Public Vault").
   • System Update: Sets video scan_status to pending_analysis.
   • User Visibility:
     • Manually added products remain visible.
     • A "Pending AI analysis" indicator is shown, informing users that AI-detected products will be available soon.

2. Admin Queue (/admin/analysis-queue):

   • Display list of videos with status pending_analysis or awaiting_approval.
   • Columns: Video Title, Creator, Submission Date, Current Status.
   • Action: "Review".

3. Analysis Interface:

   • Video Player: Watch the submitted video.
   • Manual Product Entry:
     • Form to add a "Detected Object" (Time, Category, Name).
     • Form to add "Marketplace Match" (Affiliate Link, Product Name, Price, Image).
     • Admins can manually add products at any time.
   • Cross-Vault Product Reuse:
     • List products already linked to this video in other creators' vaults.
     • Action: Move/Clone these suggestions directly into the current vault.
   • AI Assistance: "Run AI Analysis" button.
     • System Update: Sets scan_status to in_progress.
     • Triggers the AI pipeline.
     • Once finished, sets scan_status to awaiting_approval.
   • Review & Approval:
     • Product-Level Approval: Admin reviews individual AI results or manual entries and clicks "Approve" or "Reject" for each.
     • Immediate Visibility: Approved products appear in the vault immediately.
   • Complete Review:
     • Button: "Mark Video as Done".
     • System Update: Sets video scan_status to completed.
     • Notifications:
       • Logged-in Users: Trigger an in-app notification (and optional email) "Your AI Analysis is ready!".
       • Anonymous Users: Since we can't notify them directly, the "Pending AI analysis" UI should encourage them to bookmark the page or (new feature) optionally provide an email for a one-time alert.
     • Action: Video is removed from the admin's active review queue.

3.3 Mobile Admin Experience

Ensure the /admin interface is fully responsive.

• Queue Table: Use a card-based layout on mobile.
• Analysis Interface: Stack the video player and detection list vertically.
• Controls: Large, touch-friendly buttons for "Approve/Reject".

3.3 Product Status Monitoring (/admin/products)

A view to manage and improve product data quality.

• Filterable List:
  • "Missing Link": Objects without marketplace matches.
  • "Has Interest": Objects with high user interest pledges.
  • "User Suggestion": Pending product requests/proposals.
• Actions:
  • Add Link: Quick entry for affiliate URL.
  • Approve/Reject Suggestion: Move a suggestion to a confirmed match.

4. Technical Requirements

4.1 Database Changes

• Schema:
  • Ensure video_scan_status enum supports a manual review state (e.g., pending_review).
  • Reuse existing admin_moderation table for logging admin actions, or create admin_analysis_log.
• API:
  • GET /api/admin/stats: Aggregate counts for dashboard.
  • GET /api/admin/queue: Fetch pending videos.
  • POST /api/admin/analysis/complete: Publish results.

4.2 UI Components

• Dashboard: Use shadcn/ui Cards for metrics.
• Data Table: Use shadcn/ui DataTable for video/product lists with filtering/sorting.
• Forms: react-hook-form and zod for validation of manual product entry.

5. Future / Additional Suggestions

1. Revenue & Click Tracking:
   • Visualize affiliateRevenue and productClicks data.
   • Chart: Daily Clicks vs. Revenue.
2. User Management:
   • Searchable user list.
   • "Ban User" functionality for spam prevention.
3. Performance Metrics:
   • Track "Time to Review" (Submission Time vs. Completion Time) to identify bottlenecks in the manual workflow.