Documentation

Developer quickstart.

Rust and Go SDKs. Embed inference directly into your robot control loop.

Full SDK docs ship with production release

Quickstart

Set your API key and call inference. The SDK handles connection pooling, retries, and distributed tracing automatically.

1 · Add to Cargo.toml

[dependencies]
clarken-ai = "0.1"
tokio = { version = "1", features = ["full"] }

2 · Make your first call

Set CLARKEN_API_KEY in your environment, then:

use clarken_ai::{Client, InferenceRequest, Model};

#[tokio::main]
async fn main() -> Result<(), clarken_ai::Error> {
    let client = Client::new(
        std::env::var("CLARKEN_API_KEY")?
    )?;

    let response = client
        .infer(InferenceRequest {
            model:       Model::Edge,
            prompt:      "Obstacle at heading 045. Evaluate.",
            max_tokens:  512,
            temperature: 0.1,
        })
        .await?;

    println!("{}", response.completion);
    Ok(())
}

Building on Ankatos?

ClarkenAI Edge is embedded as Ankatos Cortex v1. Inference is handled by the CognitiveOS layer — no separate SDK integration needed. Ankatos docs →