Ailice is a fully autonomous, general-purpose AI agent designed to function as a standalone assistant, akin to JARVIS. It targets users seeking advanced capabilities in thematic research, coding, system management, and complex hybrid tasks, aiming to unleash LLM potential into real-world applications.

How It Works

Ailice operates as a "text computer" with a Large Language Model (LLM) as its core processor. It employs an Interactive Agents Call Tree (IACT) architecture, where tasks are decomposed into smaller subtasks handled by specialized agents. These agents communicate bidirectionally, allowing for error correction and iterative refinement, with an embedded interpreter parsing LLM outputs into executable actions.

Quick Start & Requirements

• Install via pip install -e . after cloning the repository.
• Recommended: Anaconda for virtual environments.
• Chrome browser is required for web browsing capabilities.
• For local LLM execution: minimum 2x RTX 4090 (48GB VRAM) for 70B+ models.
• Ubuntu is the best-supported OS; macOS is similar; Windows users should use Docker or WSL.
• Official Docs: Installation and Usage, LLM Configuration

Highlighted Details

• Supports voice interaction and multimodal models.
• Integrates seamlessly with open-source and commercial LLMs (e.g., GPT-4, Qwen-2.5-72B).
• Features self-expansion capabilities for dynamic module and agent creation.
• Offers flexible LLM output parsing for diverse function call mechanisms.

Maintenance & Community

• Active development with recent updates (Mar 2025, Jan 2025).
• Community links: Twitter, Reddit.

Licensing & Compatibility

• License: MIT.
• Compatible with commercial and closed-source applications.

Limitations & Caveats

• Lacks a runtime control mechanism, potentially leading to loops or extended execution.
• Performance can vary based on the chosen LLM, code version, and load.
• Frequent Google usage may require API key configuration due to search engine restrictions on AI agents.