I put Gemma 4 E4B and E2B into an e-reader so I can ask my weird questions and share my thoughts in private directly in app.

Here's how it works in the app: Framework: Runs on LiteRT-LM (like Google's AI Edge). Models: Downloads either the E2B (~2.5 GB) or E4B (~3.6 GB) INT4 quantized models directly from ungated litert-community repos without API keys, tokens, or accounts. Execution & Memory: Defaults to GPU execution with a CPU fallback. To preserve RAM, the model only initializes into memory while the AI chat UI is active and unloads when closed. Context Pipeline: Book's metadata and current passage position is automatically injected so it provides accurate answers without user prompt engineering. Features: Includes a Deep Think toggle (trades speed for context depth), a Spoilers toggle to prevent plot leaks in story books and automatic response language matching based on the selection. Other stuff: You can save responses and add them to the note & highlight system. The app is not just AI, it's a full-featured e-reader. More details here: gardenreads.app I'd love your feedback on this!