Skip to content

build_and_configure.md

Latest commit

 

History

History
113 lines (80 loc) · 3.59 KB

build_and_configure.md

File metadata and controls

113 lines (80 loc) · 3.59 KB
Raw

Build and Configuration Guide

Run the Android Voice Assistant app on a device or emulator with minimal ML experience. The app records speech, transcribes it, generates a response using an LLM, and plays it back using Android Text-to-Speech.

flowchart LR
    mic[(Microphone)] --> stt[Speech to Text]
    stt -- Prompt --> llm[LLM]
    llm -- LLM Decode / Response --> tts[Android Text to Speech]
    tts --> speaker[(Speaker)]
Loading

Table of Contents

Quick Start

To get the Android Voice Assistant app running quickly:

  1. Open the repo in Android Studio.
  2. Let Gradle sync and download dependencies.
  3. Select an arm64 or x86_64 device/emulator.
  4. Click Run.

Note

  • If you're using a physical device, make sure USB debugging is enabled in Developer Options.

  • The first build may take several minutes as dependencies and models are downloaded.

For command-line builds and model configuration details, continue below.

Prerequisites

Item Why you need it
Android Studio Build and run the app
Android NDK r29 Native C++ modules
CMake 3.27+ Native builds
Python 3 Push resources/models
Device or emulator (arm64 or x86_64) Run the app

Create local.properties in the repo root if you do not already have one:

cmake.dir=/path/to/cmake

Build from the CLI

./gradlew assembleDebug

To select a different LLM backend at build time:

./gradlew assembleDebug -PllmFramework=onnxruntime-genai

Models and Resources

Model files are downloaded during the build and deployed to the device using the app's resource push script. If you need to manually push resources later, use:

STT:

python3 app/pushAppResources.py <local_dir> <device_dir>

LLM:

python3 app/pushAppResources.py <local_dir> <device_dir> --llm_framework <LLM_FRAMEWORK>>

For more information please see the documentation in the LLM Configuration and our pushAppResources script script.

Key Locations

Path Description
app/ Android application and UI code
stt/ Speech-to-text module including models and configuration files
llm/ LLM module including models and configuration files
resources/ Resources used by the application

Next Steps

  • Read docs/benchmarking.md to measure latency and throughput.
  • Review README.md for configuration options (KleidiAI, LLM backends, custom model configs).