LiteLLM

Curator can send requests through LiteLLM, which supports many LLM providers. Curator picks the LiteLLM backend on its own for most models that are not from OpenAI, Anthropic, or Mistral. You can also choose it with backend="litellm". This guide generates recipes with Gemini.

Prerequisites

Python 3.10 or later.
Curator, installed with pip install bespokelabs-curator.
An API key for your provider. For Gemini, get one from Google AI Studio.

Create an LLM subclass

from datasets import Dataset

from bespokelabs import curator


class RecipeGenerator(curator.LLM):
    """Generate recipes for different cuisines."""

    def prompt(self, input: dict) -> str:
        return f"Generate a random {input['cuisine']} recipe. Be creative but keep it realistic."

    def parse(self, input: dict, response: str) -> dict:
        return {"recipe": response, "cuisine": input["cuisine"]}

Create the input dataset

cuisines = Dataset.from_list([
    {"cuisine": cuisine}
    for cuisine in [
        "Chinese", "Italian", "Mexican", "French", "Japanese",
        "Indian", "Thai", "Korean", "Vietnamese", "Brazilian",
    ]
])

Set the API key

export GEMINI_API_KEY=<your-api-key>

Configure the LiteLLM backend

Use the model name in LiteLLM format, which starts with the provider prefix.

recipe_generator = RecipeGenerator(
    model_name="gemini/gemini-2.5-flash",
    backend="litellm",
    backend_params={
        "max_requests_per_minute": 2_000,
        "max_tokens_per_minute": 4_000_000,
    },
)

Set the rate limits to match your account. See Online processing for all the rate limit settings.

Generate the data

recipes = recipe_generator(cuisines)
print(recipes.dataset.to_pandas())

Other providers

Each provider needs its own API key and model prefix. The LiteLLM provider list has the details for every provider.

Together
DeepInfra

export TOGETHER_API_KEY=<your-api-key>

recipe_generator = RecipeGenerator(
    model_name="together_ai/meta-llama/Llama-3.3-70B-Instruct-Turbo",
    backend="litellm",
)

Find other models in the Together model list. Some models do not support structured output.

export DEEPINFRA_API_KEY=<your-api-key>

recipe_generator = RecipeGenerator(
    model_name="deepinfra/meta-llama/Llama-3.3-70B-Instruct",
    backend="litellm",
)

Find other models in the DeepInfra model list, and add the deepinfra/ prefix to the name.

Get started

Core features

Guides

Recipes

Fine-tuning

Reference

Prerequisites

Other providers

​Prerequisites

​Other providers

Prerequisites

Other providers