AxonerAI

A type-safe, high-performance agentic AI framework in Rust.

Why AxonerAI?

• 🦀 Type-safe - Catch errors at compile time, not runtime
• ⚡ Fast - Native Rust performance, no Python overhead
• 🔧 Simple - Clean API, minimal boilerplate
• 📦 Single binary - No dependency hell, easy deployment

Installation

Add to your Cargo.toml:

[dependencies]
axonerai = "0.1"
tokio = { version = "1", features = ["full"] }
uuid = { version = "1", features = ["v4"] }

Quick Start

use axonerai::agent::Agent;
use axonerai::groq::GroqProvider;
use axonerai::tool::ToolRegistry;
use axonerai::tools::{Calculator, WebSearch};
use axonerai::file_session_manager::FileSessionManager;
use std::path::PathBuf;
use std::time::Instant;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    // 1. Create a provider
    let provider = GroqProvider::new(
        std::env::var("GROQ_API_KEY")?
    );

    // 2. Register tools
    let mut registry = ToolRegistry::new();
    registry.register(Box::new(Calculator));
    registry.register(Box::new(WebSearch::new()));

    // 3. Create session manager (optional - pass None for stateless)
    let session_manager = FileSessionManager::new(
        uuid::Uuid::new_v4().to_string(),
        PathBuf::from("./sessions"),
    )?;

    // 4. Create agent
    let system_prompt = "You are a helpful assistant. You have several tools at your disposal. \
        Do not give information without proper usage of tools.".to_string();
    
    let agent = Agent::new(
        Box::new(provider),
        registry,
        Some(system_prompt),
        Some(session_manager), // Or None for stateless
    );

    // 5. Interactive loop
    let mut input = String::new();
    println!("🤖 Agent starting... (type 'quit' to exit)");
    
    loop {
        println!("\nYou 🦁: ");
        println!("---------------");
        std::io::stdin().read_line(&mut input).expect("Failed to read line");
        
        if input.trim().to_lowercase() == "quit" || input.trim().to_lowercase() == "exit" {
            println!("Bye! 👋");
            break;
        }

        let start_time = Instant::now();
        let response = agent.run(input.trim()).await?;
        let elapsed = start_time.elapsed();
        
        println!("----------------------------------------");
        println!("{}", response);
        println!("Time taken: {:?}", elapsed);
        
        input.clear();
    }

    Ok(())
}

Supported Providers

The Framework is currently hardcoded to use :

• groq : openai/gpt-oss-20b
• openai: gpt-5-mini

Feel free to modify this in the respective files in the src crate

Provider	Model Examples
Groq	openai/gpt-oss-20b(currently hardcoded, but can be changed to llama-3.3-70b-versatile or something else,)
Anthropic	claude-sonnet-4-20250514, claude-3-haiku-20240307
OpenAI	gpt-5-mini (currently hardcoded), gpt-4o-mini

// Groq (free tier available!)
use axonerai::groq::GroqProvider;
let provider = GroqProvider::new(api_key, "llama-3.3-70b-versatile".to_string());

// Anthropic
use axonerai::anthropic::AnthropicProvider;
let provider = AnthropicProvider::new(api_key, "claude-sonnet-4-20250514".to_string());

// OpenAI
use axonerai::openai::OpenAIProvider;
let provider = OpenAIProvider::new(api_key, "gpt-4o".to_string());

Built-in Tools

• Calculator - Basic arithmetic operations
• WebSearch - Search the web via Google Custom Search API
• WebScraper - Scrape content from URLs

Creating Custom Tools

use axonerai::tool::{Tool, ToolResult};
use serde_json::{json, Value};
use anyhow::Result;

pub struct MyTool;

impl Tool for MyTool {
    fn name(&self) -> String {
        "my_tool".to_string()
    }

    fn description(&self) -> String {
        "Does something useful".to_string()
    }

    fn input_schema(&self) -> Value {
        json!({
            "type": "object",
            "properties": {
                "input": {
                    "type": "string",
                    "description": "The input to process"
                }
            },
            "required": ["input"]
        })
    }

    fn execute(&self, input: Value) -> Result<String> {
        let input_str = input["input"].as_str().unwrap_or("");
        Ok(format!("Processed: {}", input_str))
    }
}

Register it:

use axonerai::tool::ToolRegistry;

let mut registry = ToolRegistry::new();
registry.register(Box::new(MyTool));

Environment Variables

# Required for your chosen provider
GROQ_API_KEY=your_groq_key
ANTHROPIC_API_KEY=your_anthropic_key
OPENAI_API_KEY=your_openai_key

# For WebSearch tool
GOOGLE_API_KEY=your_google_key
GOOGLE_CX=your_search_engine_id

Features

• Multi-provider support (Groq, Anthropic, OpenAI)
• Tool system with custom tool support
• Session management (file-based)
• System prompts

Comparison with Python Frameworks

Feature	AxonerAI	LangChain
Language	Rust	Python
Type Safety	Compile-time	Runtime
Performance	Fast	Slow
Learning Curve	Moderate	Easy

Responsible Use

Important: AxonerAI provides tools for web search and web scraping. Users are responsible for ensuring their use complies with applicable laws, terms of service, and ethical guidelines.

Web Tools Guidelines

WebSearch Tool:

• Requires a Google Custom Search API key and Search Engine ID
• Subject to Google's Custom Search JSON API Terms of Service
• Respect rate limits (100 queries/day on free tier)
• Commercial use may require paid quota

WebScraper Tool:

• Always respect robots.txt directives
• Check website Terms of Service before scraping
• Implement appropriate rate limiting to avoid overloading servers
• Some websites explicitly prohibit automated scraping
• Consider legal implications in your jurisdiction

API Provider Terms

When using AxonerAI with LLM providers, you must comply with their respective terms:

• Anthropic Terms of Service
• OpenAI Terms of Use
• Groq Terms of Service

Your Responsibility

By using AxonerAI, you agree that:

• You are responsible for your use of the framework and its tools
• You will comply with all applicable laws and third-party terms of service
• The maintainers of AxonerAI are not liable for misuse of the framework

Use responsibly. Respect the web.

License

MIT

Contributing

Contributions welcome! Please open an issue or PR on GitHub.