BerriAI/litellm · error · LiteLLMUnknownProvider

Unmapped LLM provider for this endpoint. You passed model={m

Error message

Unmapped LLM provider for this endpoint. You passed model={model}, custom_llm_provider={custom_llm_provider}. Check supported provider and route: https://docs.litellm.ai/docs/providers

What it means

During image generation, custom_llm_provider was found in litellm._custom_providers (registered custom provider namespace), but no matching entry exists in litellm.custom_provider_map, so no CustomLLM handler instance could be located; litellm raises LiteLLMUnknownProvider pointing at the provider docs (litellm/images/main.py:507 region).

Source

Thrown at litellm/images/main.py:507

                prompt=prompt,
                timeout=timeout,
                logging_obj=litellm_logging_obj,
                optional_params=optional_params,
                model_response=model_response,
                aimg_generation=aimg_generation,
                client=client,
                api_base=api_base,
                api_key=api_key,
            )
        elif custom_llm_provider in litellm._custom_providers:  # Assume custom LLM provider
            # Get the Custom Handler
            custom_handler: CustomLLM | None = None
            for item in litellm.custom_provider_map:
                if item["provider"] == custom_llm_provider:
                    custom_handler = item["custom_handler"]

            if custom_handler is None:
                raise LiteLLMUnknownProvider(model=model, custom_llm_provider=custom_llm_provider)

            ## ROUTE LLM CALL ##
            if aimg_generation is True:
                async_custom_client: AsyncHTTPHandler | None = None
                if client is not None and isinstance(client, AsyncHTTPHandler):
                    async_custom_client = client

                ## CALL FUNCTION
                model_response = custom_handler.aimage_generation(
                    model=model,
                    prompt=prompt,
                    api_key=api_key,
                    api_base=api_base,
                    model_response=model_response,
                    optional_params=optional_params,
                    logging_obj=litellm_logging_obj,
                    timeout=timeout,
                    client=async_custom_client,

View on GitHub (pinned to 6c2dcb801b)

Solutions

Register the handler before any call: litellm.custom_provider_map.append({'provider': 'my_provider', 'custom_handler': MyCustomHandler()})
Ensure the handler subclasses CustomLLM and implements image_generation/aimage_generation
Use the exact same provider string in the call (model='my_provider/model-x') as the registered key

Example fix

# before
litellm.image_generation(model="acme/img", prompt="a cat")  # acme never registered

# after
from litellm.integrations.custom_logger import CustomLLM
class AcmeImages(CustomLLM):
    def image_generation(self, model, prompt, **kwargs): ...
litellm.custom_provider_map.append({"provider": "acme", "custom_handler": AcmeImages()})
litellm.image_generation(model="acme/img", prompt="a cat")

Defensive patterns

Strategy: validation

Validate before calling

def custom_provider_registered(provider: str) -> bool:
    import litellm
    return any(item["provider"] == provider for item in litellm.custom_provider_map)

Try / catch

from litellm.exceptions import LiteLLMUnknownProvider
try:
    r = litellm.image_generation(model="acme/img", prompt=p)
except LiteLLMUnknownProvider:
    # handler not registered — degrade or register and retry once
    raise

Prevention

Register custom providers in litellm.custom_provider_map at process startup, before any request
Add a startup assertion that every provider referenced by your model list is registered
Keep provider names as constants shared between registration and call sites

When it happens

Trigger: Adding a provider name to litellm._custom_providers (or using a reserved custom name) without appending {'provider': name, 'custom_handler': CustomLLM()} to litellm.custom_provider_map; typo between the name used in the call and the name in custom_provider_map.

Common situations: Teams wrapping internal image endpoints with CustomLLM but initializing the map lazily/conditionally so the image call runs before registration; case-mismatch ('MyProvider' vs 'myprovider'); or code copied between projects that drops the registration block.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/84b14a87a35d4da5. Report an issue: GitHub.