Crate agent_io 

BU Agent SDK

A Rust SDK for building AI agents with multi-provider LLM support.

Features

• Multi-provider LLM support (OpenAI, Anthropic, Google Gemini)
• Tool/function calling with dependency injection
• Streaming responses with event-based architecture
• Context compaction for long-running conversations
• Token usage tracking and cost calculation

Quick Start

use std::sync::Arc;
use agent_io::{Agent, llm::ChatOpenAI, tools::FunctionTool};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let llm = ChatOpenAI::new("gpt-4o")?;
    let agent = Agent::builder()
        .with_llm(Arc::new(llm))
        .build()?;
     
    let response = agent.query("Hello!").await?;
    println!("{}", response);
    Ok(())
}

Re-exports

pub use agent::Agent;

pub use agent::AgentEvent;

pub use llm::BaseChatModel;

pub use memory::EmbeddingProvider;

pub use memory::InMemoryStore;

pub use memory::MemoryManager;

pub use memory::MemoryStore;

pub use tokens::TokenCost;

pub use tools::Tool;

pub use observability::*;

Modules

agent

Agent core implementation

llm

LLM abstraction layer for multiple providers

memory

Memory system for AI agents

observability

Observability module for agent_io with tracing integration.

tokens

Token usage tracking and cost calculation

tools

Tool system with dependency injection

Macros

observe

Macro to create an observed function

observe_debug

Macro for debug-only observation

Enums

Error

Error types for the SDK

Type Aliases

Result

Result type alias for SDK operations