Aion Chat is a self-hosted AI companion designed for local network and remote access (via Tailscale), offering multi-device synchronization. It addresses the need for a private, integrated AI assistant by combining long-term memory, voice interaction, camera vision, and smart home capabilities. The project targets technically savvy users seeking a customizable, privacy-focused AI experience with extensive integration possibilities.

How It Works

The system employs a Python FastAPI backend with a native JavaScript frontend, communicating via WebSockets for real-time synchronization across PC and mobile browsers. It utilizes a modular architecture, integrating various AI models (Gemini, SiliconFlow) for core chat, image generation, and embedding functions. Key technologies include SQLite for local data storage, OpenCV for camera vision, WebRTC for voice, and native Android development for its companion app. Its novelty lies in the deep integration of diverse functionalities—from RAG memory and streaming TTS to camera monitoring, TRPGs, and music playback—all managed locally and accessible via PWA or a native app.

Quick Start & Requirements

• Installation: Primarily Python-based. Requires cloning the repository, installing dependencies via pip install -r requirements.txt, and running the FastAPI server (e.g., uvicorn main:app --reload). A Windows batch script (一键启动.bat) is provided.
• Prerequisites: Python 3.x, pip, SQLite, OpenCV, Node.js (for potential CLI tools), Android Studio for the companion app.
• Dependencies: Extensive list including fastapi, uvicorn, opencv-python, webrtcvad-wheels, pyncm, akshare, ebooklib. Android app requires compileSdk 34 / minSdk 24.
• Hardware: Local machine for hosting; GPU recommended for efficient local AI model inference.
• Links: Primary documentation is the README.

Highlighted Details

• Advanced RAG Memory: Features manual/auto summarization, vector storage (Gemini embedding), and a weighted recall score.
• Streaming TTS & Voice: Supports server-side streaming TTS (CosyVoice2) and WebRTC VAD for robust voice activity detection, enabling half-duplex voice calls.
• Multi-modal AI: Integrates camera monitoring (OpenCV, ESP32-CAM), AI image generation (Gemini), and video call capabilities with Gemini's video understanding.
• Broad Integrations: Includes NetEase Cloud Music, EPUB reading with AI annotation, fund monitoring, a TRPG engine, BLE toy control, location tracking (Gaode Maps), and device activity logging.
• Local-First & PWA/Native: Emphasizes local data storage and offers PWA installation alongside a native Android app with WebView and native bridges.

Maintenance & Community

The project's modularity and feature depth suggest active development. No specific community channels (Discord, Slack) or contributor information were found in the provided text.

Licensing & Compatibility

No license information is provided in the README. This is a critical omission for assessing commercial use or derivative works.

Limitations & Caveats

• Setup Complexity: The extensive feature set and numerous dependencies present a significant setup challenge, requiring familiarity with Python environments, AI model integration, and potentially Android development.
• Hardware Demands: Efficient AI/video processing may require a GPU.
• Licensing Uncertainty: The absence of a declared license poses a major adoption blocker, preventing clear understanding of usage rights.
• Android Permissions: The companion app requires extensive user permissions (location, microphone, usage access, background activity), which may deter some users.