Request
Users submit a media URL, watch live progress, preview completed files, and save single videos or playlists.
Python utility · Media pipelines / 2026
AnyDownloader
Wrapping an unpredictable media extraction process in a clear, observable workflow.
01 / The problem
Complexity should live in the system,not in the experience.
Media extraction depends on changing third-party platforms, authentication cookies, available formats, long-running downloads, and local files—conditions that make a simple request-response interface brittle.
02 / System flow
From input to a reliable outcome.
Stream
A background thread sends logs and results to the browser over server-sent events.
Extract
yt-dlp selects and retries formats while FFmpeg merges or converts outputs to portable media.
Serve
Docker packages both services and mounted downloads, with Render configuration for cloud self-hosting.
03 / Decisions that shaped the build
01Engineering choice
Stream progress, not silence
Download hooks publish start, completion, and error events so long-running work remains visible in the interface.
02Engineering choice
Use ordered format fallbacks
The downloader steps from preferred merged formats toward broader compatibility when a platform lacks the first choice.
03Engineering choice
Handle playlists as collections
Playlist entries retain individual filenames, previews, and save actions instead of collapsing into one opaque job.
04 / Reliability
Designed for imperfect conditions.
- Typed URL validation at the API boundary
- Configurable retries, timeout, and maximum file size
- Cookie health checks with cloud environment support
- Resolved-path validation before serving local files
05 / Reflection
“AnyDownloader showed that the useful product layer around an open-source engine is observability: explicit progress, sensible fallbacks, safe file serving, and errors a person can act on.”Explore selected work