Skip to content

classify-message

This example classifies a message three ways: a string-literal union for its category, a string enum for sentiment, and an inline union for urgency.

Nola derives a JSON Schema enum from each type. The runtime checks the answer against the allowed values and includes them in a correction retry if the model returns an invalid label. Providers that support structured output also receive the schema.

The string enum is available at runtime, so callers can compare a result with Sentiment.Negative. The urgency answer is converted to a boolean in the return value. Nola does not yet support a separate description for each label.

src/classify.tsi
export type Category = "billing" | "refund" | "fraud" | "other";


export enum Sentiment {
  Positive = "positive",
  Neutral = "neutral",
  Negative = "negative",
}


export infer function classifyMessage(.message: string) {
  const category = ask `the category of the customer message`: Category;
  const sentiment = ask `the overall sentiment of the message`: Sentiment;
  const urgent = ask `does the message need urgent attention`: "yes" | "no";
  return { category, sentiment, urgent: urgent === "yes" };
}

Run it

Section titled “Run it”
Terminal window
npx nola-lang run src/main.ts                      # mock provider (deterministic)
# real model: edit nola.config.ts to openai({ model: "gpt-5-mini" }) (needs OPENAI_API_KEY)

Scaffold: npm create nola my-app -- --template classify-message Source: examples/classify-message on GitHub. The full walkthrough is Classification.

Next: Introduction — back to the start.