Fetch YouTube and friends — anything yt-dlp supports
liquidlm fetch runs yt-dlp on your machine, so it reaches YouTube and the roughly 1,700 other sites yt-dlp supports. Captions become a searchable note, and audio or video can land in your vault alongside it.
Why this exists
The talks, interviews, and recorded sessions a team actually learns from are scattered across hundreds of hosts. Pasting one of those links into LiquidLM is the natural move, and for much of the web it works. Large video hosts are the exception: they challenge datacenter IPs with the same bot checks that break anonymous yt-dlp on servers. We already ship a linked-media pipeline with yt-dlp in the cloud, but YouTube stays blocked there on purpose so the product fails honestly instead of looping retries.
Your laptop is a different story. Browser cookies, residential egress, and a normal yt-dlp install usually work. CLI fetch puts that advantage behind one command that lands knowledge where chat, MCP, API, and search already look.
Because it is yt-dlp underneath, the reach is yt-dlp's reach — roughly 1,700 sites, catalogued at github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md. YouTube is the one people ask about first; the friends are everything else on that list.
What `liquidlm fetch` does
Point it at a vault and a URL. It probes with yt-dlp, creates a note titled from the media metadata, and embeds the source URL. Captions are pulled when the extractor exposes them — no separate flag.
- Default: note only (URL + captions when available + optional `--notes` context).
- `--audio` or `--video`: download on your machine, upload under `fetched/`, and relate the file to the note so the normal media pipeline can index it.
- `--cookies-from-browser` / `--cookies`: pass through yt-dlp auth when YouTube asks you to sign in.
- Progress: stage spinners for probe/upload, and live yt-dlp download output.
Try it
Install or upgrade the CLI, sign in, then fetch into a vault label or id:
- npm: `npm install -g @liquidlm/cli@0.1.4`
- Homebrew: `brew upgrade liquidlm/tap/liquidlm` (after the tap formula updates)
- Captions/note: `liquidlm fetch my-vault 'https://youtu.be/…' --notes 'standup clip'`
- Audio upload: `liquidlm fetch my-vault 'https://youtu.be/…' --audio --cookies-from-browser chrome`
What stays server-side
Direct uploads, notes, links that we can fetch safely, and GitHub connector sync are unchanged. Cloud linked-media still works for providers that do not require browser cookies. Fetch is the escape hatch for the hosts that only cooperate from a real user environment.