Scalable AI Training, password reset and infra

[afrinxnahar]

Key Changes:

• Scalable AI Training:
  • Introduced BullMQ + Redis queue system for asynchronous AI model training.
  • Added dedicated train-ai-worker package with processors for video/audio download (yt-dlp), token management, and style analysis.
  • Real-time progress tracking via Server-Sent Events (SSE) and new status endpoints + progress bar in the dashboard.
• Password Reset Flow:
  • Secure OTP-based reset using email (Resend + new email-templates package).
  • Updated backend controllers/services and rebuilt frontend UX (forgot/reset pages + OTP input component).
• Infrastructure & Best Practices:
  • Enhanced Supabase integration (admin client, migrations for dubbing_jobs, user_style, OTP fields).
  • Centralized Zod validation (new validations package + validation pipe).
  • Added typed API client (api-client package) for safe backend calls.
  • Docker support (Dockerfile, docker-compose with Redis & worker).
  • Strict type safety across new packages (with .d.ts builds via tsup).

Benefits:

• Handles high-volume AI jobs reliably without blocking the API.
• Improved user experience with progress feedback and modern auth flows.
• Cleaner, modular monorepo structure following DRY principles.

Resolves long-running training issues and lays foundation for production-scale features.

[vercel]

@afrinxnahar is attempting to deploy a commit to the afrin127329's projects Team on Vercel.

A member of the Team first needs to authorize it.