QuestXP: Eliminating Playlist Fatigue
"Tutorial Hell" is real. Self-taught learners spend thousands of dollars on bootcamps that are essentially curated YouTube playlists. Elite instructional content is free, but structure, accountability, and testing are locked behind paywalls.
QuestXP turns unstructured YouTube content into active, gamified university-level courses.
🛠️ The Architecture
QuestXP is a decoupled system built for extreme scalability:
- Frontend: React 18, Vite, Zustand, Tailwind CSS.
- Backend: Node.js (v22), Express.js, MongoDB Atlas.
- Queue/Cache: Valkey, BullMQ.
- Vector DB/AI: Pinecone Serverless, OpenAI GPT-4o-mini.
⚡ Core Technical Innovations
1. Multi-Stage Event-Driven Queue (BullMQ + Valkey)
Processing a course requires fanning out metadata fetch, transcript parsing, chunk embedding, and quiz generation.
We offloaded all heavy AI processing to background workers. The API returns instantly, while workers process concurrently via Valkey-backed queues.
2. Precise RAG Doubt Bot (Pinecone + OpenAI)
Transcripts are embedded using text-embedding-3-small and indexed in Pinecone using individual namespaces for every video lecture. The AI Doubt Chatbot queries this index with a high cosine similarity threshold (0.75) for precise, grounded contextual answers.
3. Pure-JS Adaptive Study Planner
A deterministic Capacity Midpoint greedy-forward-fill algorithm. It applies a 75% efficiency factor to account for note-taking, and distributes lectures proportionally. If you miss a target, it automatically recalculates future dates.
🚀 Impact & Traction
During our closed beta, QuestXP successfully onboarded 80+ highly active users who generated courses, took active-recall quizzes, and climbed the XP leaderboards.
- Live Application: questxp.in
- Demo Video: YouTube Demo
- Core Codebase: GitHub Repository
Top comments (0)