[prashantsengar]

An open-source pluggable version of Omnivore (read-it-later app).

The idea came since Omnivore decided to shut down. It is open source and self-hostable but the whole system has so many bells and whistles that nobody would want to self-host it. Even more, even if I am able to export all my data, I will be unable to use their beautiful client to view my data or build anything of my own because I am not good at that.

The idea is to have a read-it-later app that has modular components each of which can be self-hosted independently which talk in protocols. So a different data crawler module that can take many forms - browser extension, Telegram bot, or something else, that takes an URL and posts the page content to a server.

The server will have modular parsers that are extensible to support different websites and pages, contributed by community (similar to Telegram's parser or Calibre's). It then converts it into a common format to be stored.

The storage should also be pluggable so that I can easily store my data wherever I want - whether locally as MD files, on S3, or on an SQL Database. This also should be extensible to allow different storage types like so many other tools do.

Then a web server that serves common endpoints to read, add tags and annotations, and other operations. It just needs the user's storage endpoint and authorization to allow them to access their data. I feel this is going to be the most challenging.

Let users create their own clients that render this for users. Just like Telegram and Matrix.

This idea resurfaced recently because I wanted to try getrecall.ai but it looks too expensive and I am worried about privacy. Having a setup like this would allow me to plugin a local LLM to this system as well.