AIlice

What is AIlice?

AIlice is a fully autonomous, general-purpose AI agent designed for developers, researchers, and system administrators who need a self-directed assistant that can handle complex, multi-step tasks. Built on the Interactive Agents Call Tree (IACT) architecture, AIlice decomposes tasks into dynamically constructed sub-agents, enabling it to research, code, manage systems, and perform academic literature reviews with high fault tolerance. It supports both open-source LLMs (like Qwen2.5-72B) and commercial models, and can run locally or via Docker for privacy.

What are the features of AIlice?

• IACT Architecture: Dynamically breaks down complex tasks into callable sub-agents with bidirectional communication, allowing error correction and contextual handoffs between agents.
• Multi-Agent Cooperation: Different agents (e.g., coder, researcher, search engine) can be assigned distinct LLMs, balancing cost and performance on a per-task basis.
• Self-Expansion via Dynamic Modules: AIlice can autonomously write and load new external interaction modules (ext-modules) or agent types, extending its capabilities through simple prompts.
• Native Multimodal Support: All agents handle rich media including images, video, audio, LaTeX formulas, and code highlighting, with automatic variable passing between agents.
• MCP Tool Integration: Wraps any MCP-compatible server (stdio or SSE) as a standard extension module, allowing AIlice to leverage a growing ecosystem of tools.
• Voice Interaction: Supports text-to-speech (TTS) and speech-to-text (STT) with configurable devices, enabling hands-free operation.

What are the use cases of AIlice?

• Thematic research: Ask AIlice to investigate a topic, gather sources, and produce a structured report with citations.
• Automated coding and debugging: Delegate software development tasks—clone repos, analyze architecture, write code, and run tests.
• System administration: Install software, configure environments, and run security scans on a target machine.
• Academic literature review: Search for recent papers, read them, and compile a summary of field progress.
• Document generation: Create formatted PDFs, LaTeX slides, or web pages with self-determined content and layout.

How to use AIlice?

• Quick start: Clone the repo, run pip install -e ., then launch AIlice with ailice --contextWindowRatio=0.2.
• Configure an LLM: Use --modelID with a prefix (e.g., anthropic:claude-sonnet-4) or set agentModelConfig in config.json for per-agent model selection.
• Enable voice: Add --speechOn=1 --ttsDevice=cuda --sttDevice=cuda to activate voice dialogue.
• Run in Docker: Build the image and use docker run with exposed ports; optionally add CUDA or GUI support.
• Extend with MCP: Add an MCP server to config.json using ailice_mcp_wrapper to bring in external tools.