🦀 LLM Agent Rust

English | 中文

LLM-powered intelligent Rust code development assistant with multi-agent collaboration.

Introduction

This project provides AI-powered tools for Rust development, including code generation, automatic error fixing, code review, and experience learning. Built with Go and powered by LLM APIs.

📁 Project Structure

llm_agent_rust/
├── rust_agent/        # v1 - Basic Rust code agent
└── rust_agent_v2/     # v2 - Multi-agent collaboration system

🛠️ Versions

rust_agent (v1)

Basic LLM-powered Rust code generation and fixing tool.

• Code generation from natural language
• Auto error fixing (up to 5 iterations)
• Code review (correctness, security, performance)
• Code explanation and test generation

rust_agent_v2 (v2)

Advanced multi-agent collaboration system with experience learning.

• Supervisor mode: coordinates multiple specialized agents
• Experience learning: records success/failure patterns for reuse
• Autonomous coding: from requirements to working code
• MCP integration: fetch, filesystem services
• Rich toolset: full cargo commands, file ops, web search

🚀 Quick Start

v1 - Basic Agent

cd rust_agent
go build -o rust-agent.exe ./cmd
rust-agent.exe -i

Commands:

• gen <requirement> - Generate code
• fix - Fix last generated code
• review - Review code
• explain - Explain code
• test - Generate tests

v2 - Multi-Agent System

cd rust_agent_v2
go build -o rust-agent.exe ./cmd
rust-agent.exe -i

Commands:

• /create <desc> - Create project with supervisor mode
• /direct <desc> - Direct coding mode
• /fix <project> - Auto-fix project
• /search <keyword> - Search crates.io
• /exp - View experience store

🧠 Experience Learning (v2)

The agent automatically records experiences from each task:

• Success: correct imports, API usage, code patterns
• Failure: errors encountered, solutions, lessons learned

These experiences are retrieved for similar future tasks.

🔧 Requirements

• Go 1.22+
• Rust toolchain (cargo, rustc)
• Zhipu API Key

⚙️ Configuration

v1

Edit config/config.go:

• ZhipuAPIKey: API Key
• Model: Model name
• WorkspaceDir: Workspace directory

v2

Create config.json:

{
  "api": {
    "zhipu_api_key": "your-api-key",
    "model": "glm-4-flash"
  },
  "workspace": {
    "root_dir": "./rust_workspace"
  }
}

📄 License

MIT