Kowalski

A comprehensive Rust-based agent framework for interacting with Ollama models and building AI-powered applications.

Overview

Kowalski is a modular framework that provides everything you need to build sophisticated AI agents. The framework is organized into specialized crates that can be used independently or together.

Architecture

Core Components

• kowalski-core: Basic agent infrastructure, types, and utilities
• kowalski-agent-template: Templates and builders for creating custom agents
• kowalski-tools: Collection of tools for web scraping, data processing, and more
• kowalski-federation: Multi-agent coordination and communication

Specialized Agents

• kowalski-academic-agent: Research and academic paper analysis
• kowalski-code-agent: Code analysis, refactoring, and generation
• kowalski-data-agent: Data analysis and processing (optional feature)
• kowalski-web-agent: Web research and information gathering

Installation

Add to your Cargo.toml:

[dependencies]
kowalski = "0.5.0"

# Optional: Enable data analysis capabilities
kowalski = { version = "0.5.0", features = ["data"] }

Quick Start

use kowalski::{Agent, AgentBuilder, Result};

#[tokio::main]
async fn main() -> Result<()> {
    // Create a basic agent
    let agent = AgentBuilder::new()
        .with_model("llama2")
        .with_system_prompt("You are a helpful AI assistant.")
        .build()?;

    // Send a message
    let response = agent.send_message("Hello, how are you?").await?;
    println!("Response: {}", response.content);

    Ok(())
}

Using Specialized Agents

Academic Research

use kowalski::academic_agent::AcademicAgent;

let agent = AcademicAgent::new("llama2")?;
let analysis = agent.analyze_paper("path/to/paper.pdf").await?;

Code Analysis

use kowalski::code_agent::CodeAgent;

let agent = CodeAgent::new("codellama")?;
let refactored = agent.refactor_code("src/main.rs").await?;

Web Research

use kowalski::web_agent::WebAgent;

let agent = WebAgent::new("llama2")?;
let research = agent.research_topic("Rust async programming").await?;

Features

• Modular Design: Use only the components you need
• Async/Await: Built on Tokio for high-performance async operations
• Tool Integration: Rich ecosystem of tools for various tasks
• Multi-Agent Coordination: Build complex systems with multiple agents
• Extensible: Easy to add custom tools and agents

Documentation

• API Documentation
• Examples
• Contributing

License

MIT License - see LICENSE file for details.