Structured Output

Define outputs in a prompt when the model should return data matching a known shape.

outputs:
  - name: city
    kind: string
  - name: temp
    kind: integer

StructuredResult

When outputs is defined, processors can return a StructuredResult: a dictionary/object-shaped value that also keeps the raw JSON string for efficient typed casting.

cast(result, target)

Deserialize a processed structured result to an application type.

Python

from dataclasses import dataclass
from prompty import cast

@dataclass
class Weather:
    city: str
    temperature: int

weather = cast(result, Weather)

TypeScript

import { cast } from "@prompty/core";
import { z } from "zod";

const Weather = z.object({ city: z.string(), temperature: z.number() });
const weather = cast<z.infer<typeof Weather>>(result, Weather.parse);

C#

using Prompty.Core;

record Weather(string City, int Temperature);

var weather = ((StructuredResult)result).Cast<Weather>();

Rust

use serde::Deserialize;

#[derive(Deserialize)]
struct Weather {
    city: String,
    temperature: i32,
}

let weather: Weather = prompty::cast(&result)?;

Typed invocation

Some runtimes provide generic invocation helpers that run the prompt and cast the final result in one step.

Python

weather = invoke(
    "weather.prompty",
    inputs={"city": "Seattle"},
    target_type=Weather,
)

TypeScript

const weather = await invoke(
  "weather.prompty",
  { city: "Seattle" },
  { validator: Weather.parse },
);

C#

var weather = await Pipeline.InvokeAsync<Weather>(
    "weather.prompty",
    new() { ["city"] = "Seattle" });

Rust

let weather: Weather = prompty::invoke_from_path_typed(
    "weather.prompty",
    Some(&json!({ "city": "Seattle" })),
).await?;

For output schema data shapes, see Property and Prompty.