Aetherius AI Assistant is a locally-operated, modular framework for building private AI assistants and multi-agent systems. It targets users seeking deep control over their AI, offering realistic long-term memory simulation and agentic capabilities powered by open-source LLMs.

How It Works

Aetherius employs a custom Retrieval Framework simulating human memory, utilizing Qdrant for vector storage. It processes user input through a multi-stage loop involving input expansion, knowledge domain extraction, semantic term separation, and multiple memory database searches to generate an "inner monologue" and "intuition." This approach aims for more creative and realistic thought processes than standard chatbots, enabling complex agentic behavior and task execution via sub-agents.

Quick Start & Requirements

• Installation: Download project zip or git clone. Run install_aetherius_client_windows.bat or follow manual steps.
• Prerequisites: Python 3.10.6, Git (optional), Docker (for local Qdrant), FFmpeg. CUDA 11.8+ recommended for GPU acceleration (specific PyTorch install command provided). Tesseract OCR for image processing.
• Dependencies: OpenAI API key (for GPT Vision), Google CSE Key/API Key (optional), Qdrant Cloud or local instance. Supports local LLM backends like KoboldCpp, Oobabooga, or AetherNode.
• Setup: Detailed Windows installation guide available. Estimated setup time varies based on local LLM setup and Qdrant configuration.

Highlighted Details

• Modular design with separate repositories for core techniques (e.g., Hierarchical RAG, Long-Term Memory).
• Supports multi-agent frameworks, allowing sub-agents to trigger Python scripts or access external data.
• Includes GPT Vision for image analysis and TTS/Voice Input capabilities.
• Features a "Reflective Journal" and "Cognitive Offload" functionality for private thought organization.

Maintenance & Community

• Project is under active, iterative development by a single maintainer, with recent refactoring efforts.
• Discord server available for support and information: https://discord.gg/pb5zcNa7zE
• YouTube channel planned for tutorials.

Licensing & Compatibility

• The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is undetermined.

Limitations & Caveats

The project is in an "iterative development" state, with explicit warnings about bugs and constant changes. The README mentions that usability has diminished due to outdated prompting techniques, and a significant refactor is underway. Some features like TTS and Voice Input in the experimental UI are noted as needing updates.