Generate

generate(target, n, instructions) creates n values and returns them as a list, including when n=1.

import marvin

names = marvin.generate(str, n=3, instructions="Names for a fictional space station")
assert len(names) == 3

Pass a structured type to generate objects:

from pydantic import BaseModel
import marvin

class Product(BaseModel):
    name: str
    description: str

products = marvin.generate(Product, n=2, instructions="Fictional stationery products")

Generated data is model output, not a statistically random sample or a source of verified facts. Use instructions for constraints that are not represented by the type. generate_async is the async counterpart. See the API reference for optional arguments and schema-generation helpers.

Get Started

Concepts

Functions

Patterns

Guides

Community