Capsule

Turn any YouTube playlist into a daily email course with spaced repetition. Submit a playlist URL and your email — get one lesson a day until the course is done.

Live at: mindos.fly.dev

How it works

1. Submit a YouTube playlist URL + email
2. App fetches all video transcripts (3-tier pipeline)
3. LLM (OpenRouter) breaks each transcript into lessons with key takeaways
4. Lessons stored in SQLite with a spaced repetition schedule
5. APScheduler emails one lesson per day via Resend until the course is complete
6. Track progress live via SSE stream or poll the job status endpoint

Tech stack

Layer	Tech
Backend	FastAPI + APScheduler
Transcript	youtube-transcript-api → yt-dlp → Groq Whisper (3-tier fallback)
LLM	OpenRouter (configurable model)
Database	SQLite (WAL mode)
Email	Resend
Admin reporting	gspread + Google Auth
Monitoring	Sentry
Deployment	DigitalOcean (Docker)
Frontend	Single-page HTML/JS (index.html, admin.html)

API

POST /api/enroll                    — submit playlist URL + email, starts job
GET  /api/job/{id}                  — poll job progress
GET  /api/job/{id}/stream           — SSE stream for live progress
GET  /api/courses                   — list enrolled courses (by email)
DELETE /api/courses/{id}            — cancel a course
GET  /api/courses/{id}/schedule     — full lesson schedule
GET  /api/videos/{id}/emails        — emails generated for a video
POST /api/emails/{id}/send-now      — manually trigger a specific email
POST /api/courses/{id}/send-first   — trigger first email for a course
GET  /api/cron                      — manual cron trigger (CRON_SECRET required)

Admin

/admin — password-protected dashboard for monitoring enrollments, viewing job status, and manually triggering emails.

Transcript pipeline

1. youtube-transcript-api — fast, uses YOUTUBE_PROXY if set
2. yt-dlp — auto-generated captions fallback
3. Groq Whisper — audio transcription; ffmpeg auto-chunks files >24MB

Processing architecture

• JIT: enrollment processes 1 video immediately
• advance_processing runs at 2am IST nightly to queue the next video per course
• send_emails job runs every 5 minutes
• generate_reviews runs hourly

Setup

pip install -r requirements.txt
cp .env.example .env
uvicorn main:app --reload

Required env vars:

OPENROUTER_API_KEY
GROQ_API_KEY
RESEND_API_KEY
CRON_SECRET
ADMIN_USER
ADMIN_PASSWORD
YOUTUBE_PROXY       # optional, WebShare format
LLM_BUDGET_CAP      # optional, hard USD cap (default 1.0)
SENTRY_DSN          # optional