TikTok Clone with Next.js, Drizzle ORM, Neon, and Upstash Redis

This repository contains a TikTok-like short-form video application built using Next.js, Drizzle ORM ( connecting to a Neon PostgreSQL database), and Upstash Redis for caching recommendations. The app features a vertical video feed, user interactions (likes, views, watch times), and a recommendation system.

Project Overview

• Next.js (App Router): Frontend + SSR architecture for the TikTok-style interface.
• Drizzle ORM + Neon: Manages PostgreSQL tables for user_profiles, videos, user_interactions.
• Upstash Redis: Caches recommended video IDs for each user, enabling quick retrieval by the feed API.
• Collaborative Filtering (Python): Offline/scheduled script computes recommendations based on user interactions and stores them in Redis.

---

Features

Video Feed

• Scrollable TikTok-like short videos.
• Each video has captions, tags, user ID, likes count, etc.

User Interactions

• Table user_interactions tracks likes, views, watch times, etc.
• CF logic uses Python to generate a user-based similarity matrix and recommend unseen videos.

Redis Caching

• A Python job (see recommend_scheduler.py) runs every few minutes using APScheduler.
• Computes recommended videos, pushing them to Upstash Redis under recommendations:userId.
• The Next.js feed reads these cached video IDs for a fast, personalized feed experience.