Ebook System
Ebook System
YouTube transcription → essay → HTML book + EPUB for SeggySaid and Yasar Ahmad.
Architecture
Pipeline
- Video download (yt-dlp / tiktok-dlp)
- Transcription (whisper)
- LLM summarization → 300-600 word essay (GLM 5.2, max_tokens=8000)
- HTML book deployed to Cloudflare Pages
- EPUB generation via ebooklib
- Master landing page:
omair.pages.dev/library
Channels
| Channel | Essays | Words | Subdirectory |
|---|---|---|---|
| SeggySaid | 90 | 41K | seggy-book |
| Yasar Ahmad | 671 | 317K | yas-book |
Monitoring
- SeggySaid: cron every 6h (
seggy_monitor.py) - Yasar Ahmad: cron every 6h (
yas_ahmed_monitor.py) - Both use fork-to-background pattern (120s Hermes cron timeout)
Quirks
- TikTok: no
-f best(HEVC) - YouTube: glob
.webm, needs WARP socks5://127.0.0.1:40000 - GLM 5.2 reasoning overhead: max_tokens must be 8000+
- ~45s per essay, 3 workers