AURA-models

AURA

Android Utility for Runtime AI

AURA is a native Android app for running open-source Small Language Models (SLMs) locally on your phone. It is built for private, on-device research and internal use. Once models are downloaded, inference runs fully offline and the app stays ad-free.

Developed by HawkFranklin Research.

Highlights

• Pure local inference using LiteRT-LM (no cloud required for chat).
• Model hub + downloads from Hugging Face.
• Core tasks: Chat, Ask Image, and Prompt Lab.
• Open source & ad-free by design.

Models (examples)

AURA supports open-source SLMs such as:

• Gemma 3
• Qwen 2.5
• DeepSeek R1 Distill
• Phi-4 Mini

(Exact models can change over time based on the allowlist.)

System Requirements

• Android 12+ (minSdk 31)
• RAM: 8GB+ strongly recommended (4–6GB may be unstable)
• Processor: Modern Snapdragon 8 Gen 1+ / Tensor G2 or newer recommended

Privacy & Offline Use

All inference runs on-device. Your prompts and data do not leave your phone. Internet is only needed to download models or access gated repositories, and after download the app works fully offline.

Roadmap

• Chat with PDF & documents (RAG)
• Real-time multi-lingual translation
• Voice-to-action workflows
• Optional hovering assistant icon

Developer Build

cd android
./gradlew assembleDebug

Output:

• android/app/build/outputs/apk/debug/app-debug.apk

Play Store Prep (first-time)

If this is your first Play Store submission, you will need:

• App signing (keystore) and Play App Signing enrollment
• Store listing copy (title, short/full descriptions)
• Feature graphic, screenshots, and icon assets
• Privacy policy URL (even for no-data apps)
• Data Safety form answers
• Content rating questionnaire
• Target API compliance and permission justifications
• Internal/closed testing track setup before production

If you share the codebase, we can help generate the store listing text, privacy policy skeleton, and Data Safety answers needed for submission.