Upload documents in 80+ formats, videos (paste any link — YouTube · Bilibili · Vimeo · TED · Dailymotion · TikTok · Facebook · Instagram · X · Reddit · LinkedIn, auto-detected — or your own files), podcasts (paste any episode link, RSS feed, or MP3), and web articles — AI generates summaries, flashcards, quizzes, glossaries, and mind maps. Master any topic with spaced repetition, one-button daily smart sessions, an exam planner with AI mock exams and cram sheets, and a hands-free voice tutor.
| Category | What you get | |
|---|---|---|
| 📄 | Content Sources | 80+ document formats — PDF (scanned PDFs & images get AI OCR), Office incl. legacy .doc/.ppt/.xls, OpenDocument, iWork, eBooks (EPUB / MOBI / FB2), email (EML / MSG / MHTML), XPS / Visio / SVG, Jupyter notebooks, subtitles, LaTeX / Markdown / HTML / RTF, source code — plus online video from 11 sites auto-detected from the pasted link (YouTube & Bilibili with playlist / multi-part import; Vimeo, TED, Dailymotion; public TikTok, Facebook, Instagram, X/Twitter, Reddit, LinkedIn posts), uploaded videos & audio (most containers), web article clipping, podcasts from any episode link (Apple Podcasts, Overcast, Castro, Pocket Casts, Podbean, Buzzsprout & more — plus RSS feeds with an episode picker and direct MP3 links). Re-adding the same file or link is caught by content hash and takes you to the existing copy instead of duplicating it |
| 🤖 | AI Generation | Summaries, flashcards (basic / cloze / chart), adaptive quizzes (difficulty tunes to your performance on the material), glossaries, mind maps, worked problems |
| 🎯 | Study Tools | Rich-text notes, hands-free voice tutor (speaks, listens, and answers back), graded teach-back, photo problem capture, FSRS-4.5 spaced repetition, question bank, knowledge graph & learning paths |
| 🗓️ | Today & Practice | Daily study plan with goal budgeting, one-button smart session (due reviews + mistake redos + weak concepts, interleaved), unified Practice / Exam mode (quiz · flashcard · glossary · problem), timed runner |
| 🎓 | Exam Prep | Exam planner with countdown & day-by-day schedule, AI cram sheets built from your weak spots, timed AI mock exams, mistakes notebook, reinforcement center |
| 📊 | Insights | Time-on-task & accuracy analytics, per-course mastery, knowledge-gap detection, AI recommendations, AI usage & estimated cost, XP & levels |
| 🔔 | Reminders | Browser push reminders when flashcards come due (Web Push), notification digest — due cards, streak-at-risk, daily-goal progress, top knowledge gaps |
| 📴 | Offline PWA | Installable app shell; flashcards & glossary cached for offline review with background sync |
| 🔊 | Extras | PDF annotations, text-to-speech, semantic full-text search across documents, videos, notes, flashcards & glossary (finds related concepts even when the words differ) + ask-your-library AI answers with clickable source citations, public share links |
| 🔗 | Citations | Every generated card, term, and answer links back to the exact source passage; re-uploading a document flags the artifacts generated from the old version as stale |
| 🔄 | Import / Export | Anki import & export, Notion / Markdown ZIP import, study-pack export (PDF / CSV), ICS calendar feed, web clipper, highlight-to-flashcard from any web page |
| 👥 | Groups | Shared courses & documents, real-time chat, live co-study rooms with shared focus timer, assignments, live quiz battles, XP leaderboard |
| 🏫 | Classrooms | Teacher-run classrooms with rosters, assignments, and a gradebook |
| 🧰 | Tools | AI essay grading against rubrics, in-browser Python code cells (Pyodide), handwriting grading, language-learning mode |
| 💳 | Plans | Free / Pro / Team subscriptions with usage quotas; a Team plan covers every member of an organization |
| 🔐 | Auth | Email + OTP, Google / GitHub OAuth, JWT sessions, guided onboarding checklist, English & Spanish UI |
| 📱 | Mobile App | React Native (Expo) app in rn/ with full web-feature parity, plus biometric app lock, camera scan-to-summarize, haptic study feedback, keep-awake exams, offline review, and push reminders |
AI Providers — Gemini · OpenAI · Claude · Grok · DeepSeek · Kimi · Doubao · Qwen · Wenxin Yiyan (multi-provider routing, switchable from settings)
Backend — .NET 10 · ASP.NET Core · EF Core 9 · MediatR · FluentValidation · SignalR · PostgreSQL · Redis · Amazon S3 · yt-dlp · ffmpeg · Whisper.net · MailKit · JWT
Frontend — React 19 · TypeScript 5.8 · Vite 6 · TailwindCSS 4 · React Router 7 · Tiptap · D3.js + Markmap · Pyodide · Service Worker + idb-keyval (offline PWA)
Mobile — React Native 0.86 · Expo SDK 57 · expo-router · SignalR · WebView-hosted pdf.js / KaTeX / markmap (see rn/README.md)
Shared — packages/core (@study/core): platform-agnostic API services, SSE stream reader, and DTOs consumed by web/, rn/, and extension/
Build systems: Vite (
npm run dev/npm run build) is the production path —deploy.shshipsweb/distto S3/CloudFront. The Next.js scripts (dev:next/build:next/start:next) are an alternative SSR entry (web/src/app) added for landing/share-page SEO experiments and are not deployed; keep changes working under Vite first.
Architecture — Clean Architecture · CQRS · Repository + Unit of Work · SSE streaming
Prerequisites — .NET SDK 10 · Node.js 18+ · PostgreSQL 17 with pgvector (migrations create the vector extension — stock Postgres won't migrate) · Redis 7+ · ffmpeg · AWS CLI
Required services — Google Gemini API key, Google & GitHub OAuth apps, SMTP/SES email, S3 or MinIO storage (other AI providers optional)
# 1. Clone
git clone https://github.com/your-username/Study_Platform.git
cd Study_Platform
# 2. Create database
psql postgres -c "CREATE USER studyplatform WITH PASSWORD 'yourpassword';"
psql postgres -c "CREATE DATABASE studyplatform OWNER studyplatform;"
# 3. Start local services (MinIO console: http://localhost:9001, minioadmin / minioadmin123)
docker compose up -d redis minio minio-init
# 4. Configure server/StudyPlatform.API/appsettings.Development.json (see below)
# 5. Migrate & run backend
cd server
dotnet ef database update --project StudyPlatform.Infrastructure --startup-project StudyPlatform.API
dotnet run --project StudyPlatform.API # → http://localhost:5001
# 6. Run frontend
cd web && npm install && npm run dev # → http://localhost:3000
# 7. (Optional) Run the mobile app — see rn/README.md for env setup
cd rn && npm install && npx expo startserver/StudyPlatform.API/appsettings.Development.json
web/.env.local (and admin/.env.local with just VITE_API_URL)
VITE_API_URL=http://localhost:5001
VITE_GOOGLE_CLIENT_ID=xxxx.apps.googleusercontent.com
VITE_GITHUB_CLIENT_ID=Ov23lic...Clip any web page into your library as a cleaned-up Markdown article:
- Browser extension — load
extension/as an unpacked extension (Chrome / Edge / Brave). Also adds a right-click Save selection as flashcard action: the highlighted text becomes the front, AI writes the back. See extension/README.md. - Bookmarklet — Settings → Export → Web Clipper bookmarklet; no install needed.
Bundles PostgreSQL, Redis, and MinIO — no external database, cache, or storage account needed.
cp .env.example .env # fill in all values
docker compose up --build -d
docker compose exec api dotnet ef database update \
--project StudyPlatform.Infrastructure --startup-project StudyPlatform.APIWeb :3000 · Admin :4200 · API + Swagger :5001 · MinIO console :9001
VITE_*variables are baked in at build time — rebuild frontend images after changing them. MinIO credentials and bucket come from.env(MINIO_ROOT_USER/MINIO_ROOT_PASSWORD/S3_BUCKET_NAME);S3_PUBLIC_SERVICE_URLmust be reachable from the host browser.
./deploy.sh provisions ECS on a low-cost EC2 instance (t3.micro, ECS_MEMORY=768 by default — override for more headroom), an ALB for the API, RDS PostgreSQL, ElastiCache Redis, S3 buckets, and static web / admin frontends. Export DB_PASS, JWT_SECRET, GOOGLE_CLIENT_ID/SECRET, GITHUB_CLIENT_ID/SECRET, SMTP_USER, and SMTP_PASSWORD before running it.
The API runs multiple replicas: SignalR picks up a Redis backplane whenever Redis is configured and answers a startup ping, and degrades to instance-local delivery if Redis dies later. Set Api:RequireScaleOutBackplane=true so a missing backplane fails startup instead of silently half-delivering hub messages. AI generation jobs stay pinned to the replica that accepted them (their provider credentials never touch the database).
Online videos (YouTube, Bilibili, Vimeo, TED, Dailymotion, TikTok, and public social posts) use native captions when available (YouTube json3 or WebVTT), then fall back to yt-dlp + Whisper; uploaded videos are transcribed with ffmpeg + Whisper. YouTube may block cloud IPs — set YOUTUBE_PROXY_URL (SOCKS/HTTP proxy) and optionally YOUTUBE_COOKIES_B64="$(base64 < cookies.txt | tr -d '\n')" before running ./deploy-backend.sh.
{ "ConnectionStrings": { "DefaultConnection": "Host=localhost;Port=5432;Database=studyplatform;Username=studyplatform;Password=yourpassword", }, "Redis": { "Enabled": false, "ConnectionString": "localhost:6379" }, "JwtSettings": { "SecretKey": "your-32-char-secret", "AccessTokenExpiryMinutes": 15, "RefreshTokenExpiryDays": 7, }, "EmailSettings": { "Provider": "Ses", "FromEmail": "you@gmail.com", "SesRegion": "ap-southeast-2", }, // or SMTP fallback "S3": { "BucketName": "documents-dev", "ServiceUrl": "http://localhost:9000", // local MinIO "PublicServiceUrl": "http://localhost:9000", "ForcePathStyle": true, "AccessKey": "minioadmin", "SecretKey": "minioadmin123", }, "GoogleOAuth": { "ClientId": "xxxx.apps.googleusercontent.com", "ClientSecret": "GOCSPX-..." }, "GitHubOAuth": { "ClientId": "Ov23lic...", "ClientSecret": "..." }, "Cors": { "AllowedOrigins": ["http://localhost:3000", "http://localhost:3001"] }, "AppLimits": { "DocumentUploadLimit": -1 }, // -1 = unlimited; hosted default is 10 documents / audio / video uploads per account // Optional — enables browser push reminders for due flashcards. // Generate a key pair once with: npx web-push generate-vapid-keys "Vapid": { "PublicKey": "", "PrivateKey": "", "Subject": "mailto:you@example.com" }, // Optional — AI token metering. DailyTokenLimit 0 = unlimited; Pricing is per-million // tokens keyed by model-name prefix (longest match wins) and drives the cost estimates // in the AI-usage view. Unpriced models still log tokens, just with a $0 estimate. "AiUsage": { "DailyTokenLimit": 0, "Pricing": { "gemini-2.5-flash": { "InputPerMillion": 0.30, "OutputPerMillion": 2.50 } }, }, }