Readr launches open-source AI ebook reader

Readr’s strongest idea is keeping AI inside the reading flow instead of turning every question into a copy-paste detour. Its open, local-first architecture also makes it a useful reference for privacy-conscious AI app builders.

• –Adaptive context routing combines whole-book prompting for smaller books with hybrid BM25 and embedding retrieval for larger ones
• –Source citations and page-aware context help reduce hallucinations and spoiler leakage
• –On-device Kokoro-82M narration keeps books private, though the roughly 104 MB model and Apple-silicon requirement add friction
• –BYO provider support and local Ollama avoid vendor lock-in, while MIT licensing makes the Swift package useful for experimentation