Python
Prototype
Hybrid Media Downloader
A local CLI that authenticates, sniffs streams, and falls back to recording
CLI works end to end; tracked step by step in a living spec
Overview
A local command-line tool for saving media you already have access to. It logs in with Playwright and persists the session, then hands off to yt-dlp to sniff and download the stream. When a stream cannot be pulled directly, it falls back to recording the screen.
It is deliberately local-only and single-user. The interesting part is the layering: browser automation for the parts that need a real session, a downloader for the parts that do not, and a recorder as the last resort.
Stack
- Language
- Python 3
- Browser automation
- Playwright (Chromium), session saved to disk
- Download
- yt-dlp
- Fallback
- Screen recording
- Packaging
- Dockerfile alongside a local venv
Decisions Worth Explaining
Three strategies behind one command
Direct download where possible, authenticated download where a session is needed, and recording where neither works, chosen per source rather than per run.
A living step tracker instead of a static spec
The spec directory carries a tracker with per-step status and a dated changelog, so work resumes without re-reading the whole design.
What Is Not Done
- Personal use only; it does not attempt to handle sources you have no right to.
- No test suite; correctness is verified by running it.
- Saved credentials and session files live on disk in the working directory.