Skip to content

release_notes.md

Latest commit

 

History

History
15 lines (11 loc) · 1.12 KB

release_notes.md

File metadata and controls

15 lines (11 loc) · 1.12 KB
Raw

The official release of a premium, offline-first Android client for running Large Language Models (LLMs) privately on native mobile hardware.

🔑 Key Features

  • 100% Private & Offline: No cloud APIs, telemetry, or data logging. Everything stays on your device.
  • Vulkan GPU Acceleration: High-speed streaming generation with seamless, automated CPU fallback.
  • Smart Storage: Models save to getExternalFilesDir(null)/models to save internal space while complying with SELinux sandboxing.
  • Fail-Safe Downloads: Strict Content-Length verification prevents corrupted or truncated model loads.
  • Model Manager: Includes a one-tap "Delete" feature to easily purge local storage and re-download.
  • Diagnostic UI: Beautiful error banners in the onboarding view showing exact stack traces for easy troubleshooting.
  • Multimodal OCR: Offline text extraction from images, videos, and PDFs via Google ML Kit.

📦 Included Assets

  • app-release.apk: Unified production package that automatically optimizes for GPU/CPU runtime (built without Qualcomm dependencies, falling back directly to Vulkan GPU).