Decision Liquid AI
D1 2B
Liquid d1-2B - decision model: typed choice, yes/no and score answers with probabilities over text or JSON, no generated text (16 languages)
d1-2b Make it part of your app.
One model. Your language.
use xybrid::{decode_decision_request, ModelLoader};
use xybrid::ir::Envelope;
// Decision models need the llama.cpp backend: features = ["llm-llamacpp"]
let model = ModelLoader::from_registry("d1-2b").load()?;
// Ask typed questions about a state: a choice, a yes/no or a score
let request = decode_decision_request(r#"{
"state": "I was charged twice this month, please refund one of them.",
"questions": {
"refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"}
}
}"#)?;
let result = model.run(&Envelope::decision_request(request), None)?;
// One answer per question, in order, with its probabilities; no text is generated
let decisions = result.decisions().ok_or("expected decisions")?;
println!("{} answers", decisions.answers().len());Built with Xybrid Quickstart