ai

AI library for Rust primarily targeting OpenAI and Ollama APIs with more to come. This is work in progress.

Using the library

cargo add ai

Example

Chat Completion API (OpenAI)

use ai::chat_completions::{ChatCompletion, ChatCompletionRequestBuilder, Messages};

#[tokio::main]
async fn main() -> ai::Result<()> {
    tracing_subscriber::fmt()
        .with_env_filter(tracing_subscriber::EnvFilter::from_default_env())
        .init();

    let openai =
        ai::clients::openai::Client::new("open_api_key")?;

    let request = &ChatCompletionRequestBuilder::default()
        .model("gpt-4o-mini".to_string())
        .messages(vec![
            Message::system("Your are a helpful assistant."),
            Message::user("Tell me a joke"),
        ])
        .build()?;

    let response = openai.complete(&request).await?;
    println!("{}", &response.choices[0].message.content);

    dbg!(&response);

    Ok(())
}

Dynamic Clients based on the runtime

let client: Box<dyn Client> = if let Ok(openai_api_key) = std::env::var("OPENAI_API_KEY") {
    let openai = ai::clients::openai::Client::new(&openai_api_key)?;
    Box::new(openai)
} else {
    let ollama = ai::clients::ollama::Client::new()?;
    Box::new(ollama)
};

let request = &ChatCompletionRequestBuilder::default()
    .model("llama3.2".into())
    .messages(vec![
        Message::system("Your are a helpful assistant."),
        Message::user("Tell me a joke".to_string()),
    ])
    .build()?;

let response = client.complete(&request).await?;

dbg!(&response);

Clients

OpenAI

cargo add ai --features=openai_client

let openai = ai::clients::openai::Client::new("open_api_key")?;
let openai = ai::clients::openai::Client::from_url("open_api_key", "http://api.openai.com/v1")?;
let openai = ai::clients::openai::Client::from_env()?;

Ollama

cargo add ai --features=ollama_client

let ollama = ai::clients::ollama::Client::new()?;
let ollama = ai::clients::ollama::Client::from_url("http://localhost:11434")?;