Backends

RLMs natively support a wide range of language model providers, including OpenAI, Anthropic, Portkey, OpenRouter, and LiteLLM. Additional providers can be supported with minimal effort. The backend_kwargs are named arguments passed directly to the backend client.

OpenAI

rlm = RLM(
    backend="openai",
    backend_kwargs={
        "api_key": os.getenv("OPENAI_API_KEY"),  # or set OPENAI_API_KEY env
        "model_name": "gpt-5-mini",
        "base_url": "https://api.openai.com/v1",  # optional
    },
)

Anthropic

rlm = RLM(
    backend="anthropic",
    backend_kwargs={
        "api_key": os.getenv("ANTHROPIC_API_KEY"),
        "model_name": "claude-sonnet-4-20250514",
    },
)

Portkey

Portkey is a client for routing to hundreds of different open and closed frontier models.

rlm = RLM(
    backend="portkey",
    backend_kwargs={
        "api_key": os.getenv("PORTKEY_API_KEY"),
        "model_name": "@openai/gpt-5-mini",  # Portkey format: @provider/model
    },
)

OpenRouter

OpenRouter is a multi-provider gateway for accessing a wide range of models from different providers through one API.

rlm = RLM(
    backend="openrouter",
    backend_kwargs={
        "api_key": os.getenv("OPENROUTER_API_KEY"),
        "model_name": "openai/gpt-5-mini",  # Format: provider/model
    },
)

LiteLLM

LiteLLM is a universal interface for 100+ model providers, with support for local models and custom endpoints.

rlm = RLM(
    backend="litellm",
    backend_kwargs={
        "model_name": "gpt-5-mini",
    },
)
# Set provider API keys in environment

vLLM (Local)

Local model serving.

# Start vLLM server
python -m vllm.entrypoints.openai.api_server \
    --model meta-llama/Llama-3-70b \
    --port 8000
rlm = RLM(
    backend="vllm",
    backend_kwargs={
        "base_url": "http://localhost:8000/v1",  # Required
        "model_name": "meta-llama/Llama-3-70b",
    },
)

Multiple Backends (Experimental)

Experimental: This feature allows you to specify ordered lists of backends and model kwargs, so that RLMs can sub-call different language models from within execution code. The order of other_backends and other_backend_kwargs must match: e.g., the 0th element of other_backends is used with the 0th dict in other_backend_kwargs.

This functionality is for advanced use and is currently experimental.It will become more useful as RLMs get the ability to orchestrate and delegate between different LMs within a workflow.

rlm = RLM(
    backend="openai",
    backend_kwargs={"model_name": "gpt-5-mini"},
    other_backends=["anthropic", "openai"],  # ORDER MATTERS!
    other_backend_kwargs=[
        {"model_name": "claude-sonnet-4-20250514"},
        {"model_name": "gpt-4o-mini"},
    ],  # ORDER MATCHES other_backends
)

Inside REPL (future releases):

llm_query("prompt")  # Uses default(gpt-5-mini)
llm_query("prompt", model="claude-sonnet-4-20250514")  # Uses Claude 
llm_query("prompt", model="gpt-4o-mini")  # Uses GPT-4o-mini