←back to thread

1222 points phantomathkg | 1 comments | | HN request time: 0s | source
Show context
kepano ◴[] No.44065986[source]
I built Obsidian Web Clipper (open source, MIT) to replace my read-it-later app and save everything to local markdown files. Now that Obsidian Bases is available, it makes for a very nice web archival tool and reading experience. Here's a video:

https://mastodon.social/@kepano/114553164915046938

You can use Web Clipper with any app that supports Markdown, not just Obsidian.

Defuddle is the underlying HTML-to-Markdown library I made for Web Clipper, and can also be used as a CLI:

https://github.com/kepano/defuddle

https://github.com/kepano/defuddle-cli

replies(4): >>44067250 #>>44067701 #>>44067853 #>>44067859 #
jonahx ◴[] No.44067250[source]
This is very cool.

The thing I really want is this, combined with some automated local background LLM training / rag (not sure what the right approach is) process. So that, at the end of the day, everything I bookmark get saved locally, can be read in a nice format like you have the video, and be semantically queried, and it's all local:

"What was that article I saw read 1-3 months ago some new type of LLM training?"

"Find that really nice explanation of determinants article"

etc...

Have you investigated anything like that?

replies(1): >>44067317 #
1. kepano ◴[] No.44067317[source]
Since the content is saved to Markdown you can use it with pretty much any tool that will ingest that content.

There's also Obsidian Web Clipper's Interpreter feature, which lets you run prompts on a web page before saving:

https://help.obsidian.md/web-clipper/interpreter