How lordzedith/tiktok works
Project type and purpose
• This is a full-stack video-sharing platform (a TikTok clone) with two React SPAs (user frontend + admin dashboard), a FastAPI backend, a background processing service for moderation & recommendation (in backend_constant), and a relational database. It implements AI-powered content moderation (video & comments) and a self-training recommendation engine.
File structure
• Top-level directories:
• frontend/ – user-facing React app
• admin/ – admin React app for human overrides and dashboards
• backend/ – FastAPI HTTP server (APIs for video upload, streaming, user auth, comments)
• backend_constant/ – long-running Python process that executes scheduled jobs (video moderation via Gemini AI, comment moderation with DistilBERT, recommendation model training, user preference analysis)
• testing/ – scripts to seed/test database & upload sample videos
• Root configs (.env, tailwind.config.js, vite.config.ts) govern environment variables, styling, and builds.
• backend/ contains auth.py, database.py, models.py, schemas.py, recommend_videos.py, and main.py (entry point).
• backend_constant/constant_run.py orchestrates daily and continuous tasks.
README for additional insights
• The README outlines:
• AI-moderation pipeline stages (automated screening → human review)
• Video transcoding & thumbnail generation (via FFmpeg)
• Content-based + engagement-based hybrid recommendation, with daily batch retraining
• Continuous user-preference analysis
• Admin dashboard features (override decisions, view logs)
• It even includes a high-level directory sketch under “Architecture.”