AIlice

Description

AIlice is a fully autonomous, general-purpose AI agent project aiming to create a standalone artificial intelligence assistant akin to JARVIS, built upon open-source Large Language Models (LLMs). Its core innovation lies in the Interactive Agents Call Tree (IACT) architecture, which allows AIlice to decompose complex tasks into dynamically constructed agents. These agents then integrate their results with a high degree of fault tolerance, enabling robust task completion.

Currently, AIlice demonstrates proficiency across a wide spectrum of tasks, including in-depth thematic research, coding assistance, system management, literature reviews, and complex hybrid tasks that extend beyond these foundational capabilities. The ultimate vision for AIlice is the self-evolution of AI agents, where agents autonomously develop their own feature expansions and create new agent types, seamlessly unleashing the knowledge and reasoning power of LLMs into the real world.

Key technical features include advanced research capabilities on specialized subjects, the ability to read and analyze scholarly works, and sophisticated automation for programming and system management, functioning as an AI-powered operating system. It supports voice interaction, is compatible with both open-source and commercial LLMs, and boasts native multi-modal support across all agents. The user interface is rich, handling image, video, and audio, LaTeX formulas, code highlighting, and file upload/download. Its Interactive Agents Call Tree architecture is designed for natural interaction and high fault tolerance, with flexible parsing of LLM outputs enabling a broader range of function call mechanisms. AIlice can also self-construct and dynamically load modules for environmental interaction, offering endless possibilities for feature expansion.

AIlice can be tried online via kragent.ai for instant experience without local setup. For local deployment, users can clone the repository and install dependencies. The project emphasizes flexibility in LLM selection, supporting various open-source and commercial models, with recommendations for optimal performance and cost-effectiveness. The development roadmap includes enhancing long-term memory mechanisms, improving multimodal support, enabling self-expansion capabilities, and enriching the UI interface.

AIlice's Core Features

• Fully autonomous, general-purpose AI agent

• Interactive Agents Call Tree (IACT) architecture for task decomposition

• Dynamic agent construction and result integration

• Proficiency in research, coding, system management, and literature reviews

• Goal of self-evolution for AI agents

• Voice interaction support

• Compatibility with open-source and commercial LLMs

• Native multi-modal support across agents

• Rich media UI with image/video/audio, LaTeX, code highlighting, file I/O

• High fault tolerance in agent interactions

• Flexible LLM output parsing for function calls

• Self-construction and dynamic loading of external interaction modules

Getting Started with AIlice

1. 1

   Clone: Clone the AIlice GitHub repository.

2. 2

   Install: Install dependencies using pip.

3. 3

   Configure: Set up LLM configurations and other parameters in config.json.

4. 4

   Execute: Run AIlice from the command line with desired parameters.

5. 5

   Interact: Use the web interface or command line to issue commands.

6. 6

   Extend: Develop and load custom extension modules for new functionalities.

AIlice's Use Cases

• Autonomous Research
• Code Generation
• System Management
• Content Creation
• Complex Task Decomposition
• Literature Review
• Multimodal Interaction
• AI Agent Development

FAQ from AIlice

• What is AIlice?

  AIlice is a fully autonomous, general-purpose AI agent designed to act as a sophisticated assistant, similar to JARVIS, leveraging open-source LLMs.

• How does AIlice handle complex tasks?

  AIlice uses an Interactive Agents Call Tree (IACT) architecture to decompose complex tasks into smaller, dynamically constructed agents that collaborate to achieve the final goal.

• Can AIlice run locally?

  Yes, AIlice can be run locally. For users planning to run LLMs locally, significant GPU VRAM is recommended (e.g., two RTX 4090s for 70B+ models).

• What kind of tasks can AIlice perform?

  AIlice can perform tasks such as thematic research, coding, system management, literature reviews, and complex hybrid tasks. It can also generate images, write code, and perform simulations.

• Does AIlice support voice interaction?

  Yes, AIlice includes support for voice dialogue, with practical implementation thanks to advanced text-to-speech and speech-to-text models.

• Can I use my own LLMs with AIlice?

  Yes, AIlice is compatible with both open-source and commercial LLMs. Configuration options are available to integrate various models, including those via inference services like OpenRouter and LM Studio.

• What are the system requirements for running AIlice?

  If not running LLMs locally, AIlice has minimal hardware requirements. Running large LLMs locally requires substantial GPU VRAM, with recommendations for specific hardware configurations.

• How can I extend AIlice's capabilities?

  AIlice is designed for extensibility. Users can develop and load custom extension modules (ext-modules) and can even contribute to the self-expansion of AI agents.

• Is there an online demo for AIlice?

  Yes, you can experience AIlice instantly online at kragent.ai, which allows exploration of its capabilities without local setup.

• What is the IACT architecture?

  IACT stands for Interactive Agents Call Tree. It's AIlice's core computational model that breaks down tasks into a tree of interacting agents, enabling complex problem-solving and fault tolerance.