Homelib
I have a ton of bookmarks, starred repos, and links that are impossible to manage and find things with. I wanted a way to collect and archive all this information, locally, under my own control. And I wanted to make it accessible to AI models for researcha and discovery
The library, with saved items ready to browse and search.
Homelib is your own private library for the things you read online.
When I import a link, Homelib grabs a clean copy and does analysis, writes a summary, assigns topics, and records the people, companies and places mentioned (the "entities"). The source material stay as files on a machine I control, so the library is still there if a link disappears or a service changes.
When I want more context, I can profile or research an item across other sources, connect it to related material, or ask questions in the chat. The assistant shows the sources it used, so I can follow the trail rather than take an answer on trust.
It has a ton of useful features: git-backed storage and sync between several instances, the ability to process audio and youtube videos as transcripts, finding offline articles via archive sources, fetching content behind paywalls, and more. The relations and discovery features turned out to be super useful. There's even a built-in agent chat with RAG and search.
The assistant finding useful sources in the library and explaining why they fit.
I made The Library at Home, a one-minute animated film about the idea behind Homelib. The project is in a small closed beta; the site has the details and a way to ask to try it. Im still locking down the architecture and standards, and adding better classification.