🌌 AETHER AI

A Production-Grade, Cross-Platform AI Chat Application

---

⚠️ Notice: Personal Learning Project

Please note that this repository was created strictly for personal learning and educational purposes. It serves as a practical implementation of advanced Python concepts, Clean Architecture, and SOLID principles.

🚀 Overview

AETHER AI is an advanced, cross-platform AI chat application built to production-grade engineering standards. It features a premium, responsive user interface utilizing glassmorphism aesthetics and provides raw HTTP LLM routing across multiple providers.

More than just an application, Aether AI is designed around strict Clean Architecture and SOLID principles, utilizing a pure Supabase PostgREST persistence layer with client-side zero-knowledge encrypted cloud syncing.

✨ Key Features

• Premium Flet UI: A highly polished, responsive interface featuring layered compositing and glassmorphism aesthetics (--aether-bg-secondary with backdrop blur and custom glow effects).
• Multi-Provider LLM Routing: Seamless, raw HTTP routing to multiple AI providers including Groq, Gemini, and OpenRouter using SSE (Server-Sent Events) for real-time streaming.
• Zero-Knowledge Key Syncing: Client-side AES-256-GCM encryption for API keys. Plaintext keys never leave the device, ensuring secure cross-device syncing via Supabase.
• Intelligent Context Engine: A sliding context window utilizing collections.deque that gracefully evicts older messages based on model token limits while strictly preserving system prompts.
• Command Palette: A global, Cmd+K accessible overlay providing real-time fuzzy search across chats, snippets, and documents.
• Token & Cost Analytics: A comprehensive Usage Dashboard featuring interactive Pie and Bar charts to track token consumption, provider distribution, and cumulative costs.
• Multi-Format Export Engine: High-performance, multiprocessing-backed conversation exports supporting PDF, Markdown, and JSON.
• Voice Mode UI: Real-time audio processing featuring dynamic waveform visualization rendered on a Canvas at 30+ FPS.

🏗️ Architecture & Engineering

Aether AI is strictly decoupled into four primary layers, ensuring that business logic remains completely independent of frameworks, UI, or databases.

1. Domain Layer: The pure core containing dataclass entities, value objects, domain exceptions, and repository interfaces. Zero external dependencies.
2. Application Layer: Contains use cases (e.g., SendMessageUseCase) orchestrating the domain objects via input/output ports.
3. Infrastructure Layer: Implements adapters for Supabase (via raw httpx PostgREST), AES encryption, multiprocessing export engines, and external LLM APIs.
4. Presentation Layer: Flet-specific UI components, custom routers, MVVM state management, and an Observer/Pub-Sub EventBus.

Core Patterns Utilized

• Clean Architecture & SOLID
• Dependency Injection (DI)
• Repository & Factory Patterns
• Observer/Pub-Sub (EventBus)
• Strategy Pattern (Failover Routing)

🛠️ Technology Stack

• Language: Python 3.12+
• UI Framework: Flet 0.25+
• Networking: httpx 0.27+ (Async REST/SSE)
• Backend / BaaS: Supabase (Auth, PostgREST, RLS)
• Cryptography: cryptography 44+ (AES-256-GCM)
• Data Validation: Pydantic 2.0+ (Settings)
• Testing & QA: pytest, pytest-asyncio, hypothesis, ruff, mypy
• Packaging: PyInstaller 6.0+

⚙️ Environment Configuration

The application is configured via environment variables. Ensure the following are set before launching:

Variable	Required	Description
AETHER_SUPABASE_URL	Yes	Supabase project URL
AETHER_SUPABASE_ANON_KEY	Yes	Supabase anonymous public key
AETHER_ENCRYPTION_PASSWORD	No	Master password for key vault
AETHER_LOG_LEVEL	No	Logging level (default: INFO)
AETHER_MAX_CONTEXT_TOKENS	No	Default context window size (default: 8192)

(Note: API keys for Groq, Gemini, and OpenRouter can be configured directly through the in-app Settings UI and will be securely encrypted).

🧪 CI/CD & Testing

Aether AI enforces rigorous CI/CD checks via GitHub Actions:

• Linting & Formatting: ruff check .
• Static Type Checking: mypy --strict .
• Testing: Property testing via hypothesis and comprehensive unit/integration tests with pytest.
• Coverage: Hard gate requiring >= 85% test coverage.

📚 Documentation

For an in-depth dive into the architecture, pedagogy, and design manifesto of the project, please refer to the official documentation: 👉 Aether AI Reference Documentation

---

Engineered for resilience, crafted for usability.