ReadAny Wants to Fix the Fact That You Forget Everything You Read
You finish a book, feel smarter for about a week, and then someone asks what it was about and you can only remember the cover. Your highlights are scattered across three apps. And when you try to find that one passage about attention or habit formation, you're stuck typing keywords into a search box that only matches exact strings. ReadAny is an open-source e-book reader built around a different premise: that an e-reader should help you query your books, not just display them.
What It Does
ReadAny is a local-first AI e-book reader for desktop and mobile. It supports macOS, Windows, Linux, iOS, and Android, and it's built around an AI-native reading workflow rather than bolting AI onto a traditional reader.
The core loop goes like this: you read, you highlight, and then you can ask questions about the current book or a selected passage and get answers grounded in your reading context. Behind that is a RAG (retrieval-augmented generation) pipeline using hybrid retrieval—vector search plus BM25—so you can find ideas by meaning even when you don't remember the exact words. Embeddings and the vector store run locally, which means your books, highlights, and notes stay on your machine and work offline.
On the model side, ReadAny doesn't lock you in. You can connect OpenAI, Claude, Gemini, Ollama, DeepSeek, or any custom-compatible provider, so you can run everything locally through Ollama or point it at a hosted API depending on your preferences. It also ships with TTS, reading statistics, WebDAV sync, a skills system, and note export in five formats, with support for more than ten book formats.
Why It's Cool
It treats search as a semantic problem, not a string-matching one. The README's own framing is telling: "Why can I only search by keywords?" Hybrid vector retrieval plus BM25 is a genuinely better approach for the way people actually remember books—by concept, not by exact phrasing. If you half-remember an idea but not the words, this is the difference between finding it and giving up.
Local-first isn't a buzzword here, it's structural. Local embeddings and a local vector store mean your reading library doesn't have to leave your device to get AI features. That's a meaningful tradeoff compared to cloud-based reading assistants, and it's the kind of architecture decision that determines whether you'd actually put your personal notes into a tool.
The comparison table is refreshingly honest. ReadAny lines itself up against Calibre, KOReader, and Apple Books, and it doesn't pretend to win everywhere. Calibre still beats it on format support (15+ versus 10+). KOReader and Apple Books have TTS too, just limited. What ReadAny claims uniquely is the combination of AI chat, semantic search, local vector storage, WebDAV sync, a skills system, and open source—all in one reader. That's a specific, defensible niche rather than a vague "better than everything" claim.
The workflow is coherent end to end. Selecting text to highlight, asking about that passage, and then exporting your notes in one of five formats is a complete loop. A lot of AI reading tools stop at the chat interface and leave you to figure out the knowledge management part yourself.
Mobile arrived in v2.0. The iOS and Android apps are now available, which matters because a reader that only lives on your desktop is a reader you use half as often. Combined with WebDAV sync, you can plausibly move between devices without losing your place or your notes.
How to Try It
- Head to the releases page and grab the build for your platform. Desktop builds cover macOS, Windows, and Linux; mobile builds cover iOS and Android.
- Install and open the app, then add a book in one of the supported formats.
- Configure a model provider. If you want to stay fully local, point it at Ollama. If you'd rather use a hosted model, connect OpenAI, Claude, Gemini, or DeepSeek—or a custom-compatible endpoint.
- Start reading. Select a passage, highlight it, and ask a question about it to see the RAG chat in action.
- If you want the same library on multiple devices, set up WebDAV sync.
The source lives at github.com/codedogQBY/ReadAny, and there's a Chinese README (README_CN.md) if that's more useful to you.
Final Thoughts
ReadAny is best suited to people who read seriously enough to accumulate notes and then get frustrated when those notes become unusable. If you're a casual reader who finishes a book and moves on, the AI features won't earn their keep. But if you've ever wished you could ask your library a question, the combination of local vector storage, hybrid retrieval, and provider flexibility makes this worth a look. It's open source, it's cross-platform, and it's solving a problem most readers have quietly accepted as unavoidable.