Skip to content

README.md

Latest commit

 

History

History
213 lines (144 loc) · 4.43 KB

README.md

File metadata and controls

213 lines (144 loc) · 4.43 KB

AlphaAI

AlphaAI is an Android AI assistant prototype with a pluggable skill system, behavior learning, and a rule-based recommendation engine.

Highlights

  • Skill scheduling via SimpleScheduler
  • Built-in skills:
    • Toast (system.toast)
    • Camera photo capture (camera.take_photo)
    • Location (location.get)
    • Notification (notification.show)
    • Intent execution (intent.execute)
  • Intent-based commands:
    • Open apps
    • Open URLs
    • Dial numbers
    • Share text
    • Open map intents
  • Local behavior learning with Room:
    • User action logs
    • Recommendation hit logs
    • Basic profile fields
  • Rule-based recommendation engine (WorkManager + JSON rules)
  • In-app debug trigger for immediate recommendation checks
  • Self-hosted cloud API gateway integration via CloudApiSkill (cloud.api)

Project Structure

  • app: UI, app lifecycle, runtime integration
  • core: scheduler, resolver, core execution contracts
  • skills: built-in skill implementations
  • data: Room entities, DAO, database, profile generator
  • recommendation: rule model, loader, matcher, worker, notifier
  • common: shared Android library placeholder

Quick Start

Requirements

  • Android Studio (Giraffe+ recommended)
  • JDK 17
  • Android SDK with API 33
  • Emulator or Android device

Build & Install

./gradlew :app:assembleDebug
./gradlew :app:installDebug

On Windows PowerShell:

.\gradlew.bat :app:assembleDebug
.\gradlew.bat :app:installDebug

Run

Launch the app from Android Studio, or with adb:

adb shell am start -n com.example.alphaai/com.yourname.alphaai.MainActivity

Example Commands

Core Skills

  • toast hello
  • photo
  • location
  • notification

Intent Skills

  • open settings
  • open chrome
  • open https://www.github.com
  • visit github.com
  • dial 10086
  • share hello to chrome

Cloud API Skills

  • weather london
  • translate hello world to fr
  • news ai
  • stock AAPL
  • cloud weather current city=tokyo

Cloud Gateway Deployment

AlphaAI provides a lightweight API gateway prototype so users can self-host and connect third-party cloud services securely.

Option A: Cloudflare Workers (recommended)

Gateway template path:

  • gateway/cloudflare-worker

Steps:

  1. Install Node.js 20+ and Wrangler.
  2. Edit gateway/cloudflare-worker/wrangler.toml:
  • Set APP_GATEWAY_API_KEY.
  • Configure service credentials (WEATHER_*, TRANSLATE_*, NEWS_*, STOCK_*).
  1. Deploy:
cd gateway/cloudflare-worker
npm install
npm run deploy
  1. Copy your Worker URL, for example:
https://alphaai-gateway.<your-subdomain>.workers.dev

Option B: Docker Compose (VPS or local server)

Gateway template path:

  • gateway/docker
  • docker-compose.gateway.yml

Steps:

  1. Create env file:
cp gateway/docker/.env.example gateway/docker/.env
  1. Fill in API keys in gateway/docker/.env.
  2. Start gateway:
docker compose -f docker-compose.gateway.yml up -d --build
  1. Gateway endpoint (default):
http://<your-host>:8080/api

AlphaAI App Configuration for Cloud Gateway

Set gateway values in app/build.gradle:

  • CLOUD_GATEWAY_BASE_URL
  • CLOUD_GATEWAY_API_KEY

Example:

buildConfigField "String", "CLOUD_GATEWAY_BASE_URL", '"https://alphaai-gateway.example.com"'
buildConfigField "String", "CLOUD_GATEWAY_API_KEY", '"your-shared-gateway-key"'

Rebuild app after editing these fields.

For production hardening notes and deployment pitfalls, see Remind.md.

Recommendation Rules

Default rules are loaded from:

  • app/src/main/assets/recommendations.json

Periodic recommendation checks are scheduled in:

  • AlphaAIApplication via WorkManager

You can trigger recommendation evaluation immediately from the UI button:

  • Trigger recommendation now

Data & Privacy

  • All learning data is stored locally in alphaai.db
  • You can clear learning data from the app UI (Clear learning data)

Development Status

Current implementation includes:

  • Core scheduling + system skills
  • Learning data collection and history UI
  • Recommendation engine v1 with notification output

Planned improvements:

  • Feedback-based recommendation weight updates
  • Better context providers (foreground app, richer location types)
  • Dynamic/registry-based skill loading for open-source contributors

Contributing

Please read:

  • CONTRIBUTING.md

License

This project is licensed under the MIT License.

  • LICENSE