atomtanstudio/scrollback

Scrollback is a self hosted personal archive for saving X, formerly Twitter, posts, threads, article links, RSS feeds, media, and AI art prompts into a private, searchable library that you run on your own machine or server. Nothing is stored on a third party service, all saved content lives in a database you control. Content gets into Scrollback through two main paths. A browser extension lets you capture X posts and threads with one click while browsing, and an RSS or Atom feed syncer pulls articles into your library on a schedule. Once content is saved, you can search it with regular keyword search, or with semantic search when the app is set up with a vector database, meaning you can find items by their meaning rather than exact matching words. Optionally, you can connect an OpenAI or Google Gemini API key to get AI generated summaries, tags, categories, translated text, and image descriptions attached to saved items. Media files can be stored locally on disk or in Cloudflare R2 storage. The app also exposes a remote query API so other tools or scripts can pull from your archive without touching the database directly, and it can build a searchable index meant for AI agents to query your saved content. For bulk importing X activity like Likes and Bookmarks, the README recommends using an official X API bearer token rather than relying only on the browser extension's page based capture, since page based capture depends on X's web app behavior and carries more account risk. The tech stack is Next.js 16, React 18, Prisma 7, Tailwind CSS, and either SQLite for fast local installs or PostgreSQL with pgvector for hosted installs and vector powered semantic search. It requires Node.js 22.x. An onboarding wizard walks through database setup, admin account creation, optional AI provider setup, optional X API setup, and browser extension pairing after the first run.