r_agent

r_agent is an agent framework for building AI workflows with rust.

Note

This project is currently in development. The exposed APIs may change in future releases. The inner implementation may directly use unwrapped calls, which may lead to panics and will be improved in future releases. Please use with caution.

Features To Implement

• OpenAI (OpenAI Compatible) backends support...
• More LLM backends (Azure, local models, etc.)
• Tool Calling: Equip agents with the ability to utilize self-designed tools
• Memory - Sliding window + auto-summarization strategies for context management
• ReAct Loop - Standard think → act → observe reasoning cycle
• DAG Pipeline - Directed Acyclic Graph workflow orchestration with parallel execution
• Debug Friendly - Utilize rust's error handling for transparent debugging, and logging for traceability, with request content and token usage tracking.

Usage Example

use r_agent::agent::base::BaseAgent;
use r_agent::tool::base::Tool;
use serde_json::Value;
use serde_json::json;
use r_agent::config::config::*;
use serde_json;

#[derive(Debug)]
struct CalculatorTool{
    config: Value
}

impl Tool for CalculatorTool {
    fn load(&self) -> &serde_json::Value {
        &self.config
    }

    fn init(&mut self) {
        // Initialization logic if needed
    }

    fn execute(&self, input: &str) -> String {
        // A very basic implementation that only handles addition for demonstration
        let args = serde_json::from_str::<serde_json::Value>(input).unwrap();
        String::from(
            format!("{}", args["num1"].as_f64().unwrap() + args["num2"].as_f64().unwrap())
        )
    }
}

#[tokio::main]
async fn main() {
    let tool = CalculatorTool {
        config: json!({
            "name": "sumOfTwoNumbers",
            "type": "object",
            "description": "A tool to calculate the sum of two numbers.",
            "parameters": {
                "type": "object",
                "properties": {
                    "num1": {
                        "type": "number",
                        "description": "The first number."
                    },
                    "num2": {
                        "type": "number",
                        "description": "The second number."
                    }
                },
                "required": ["num1", "num2"]
            }
        })
    };

    let config = load_config(None);
    let model_name = "gpt-4o-mini";
    
    let memory = r_agent::memory::summary::SummaryMemory::new("tool_agent_example", 0.3, &config, model_name, "", 8192, "./workspace/");
    
    let system_prompt = "You are a React Agent. Use tools to answer user queries.";
    let tool_manager = r_agent::tool::manager::ToolManager::new(vec![Box::new(tool)]);
    let tools = vec!["sumOfTwoNumbers".to_string()];

    let mut agent = r_agent::agent::react_agent::ReactAgent::new(
        &config,
        model_name,
        system_prompt,
        3,
        tool_manager,
        memory,
        tools,
    );

    let user_prompt = "What is the sum of 1000 and 10000? Use the tool and finally give me the answer.";

    let answer = agent.run(user_prompt).await;
    println!("Agent Answer: {}", answer);
}

Contributing

Contributions are welcome! Please open issues and submit pull requests for bug fixes and new features!

Thanks

Thanks to the following open source projects:

• EasyAgent
• llm
• tiktoken-rs