BerriAI/litellm · error · ValueError

api_base is None. Please set AZURE_AI_API_BASE or dynamicall

Error message

api_base is None. Please set AZURE_AI_API_BASE or dynamically via `api_base` param, to make the request.

What it means

The Azure AI image-embeddings handler POSTs to {api_base}/images/embeddings. Without api_base the URL cannot be built, so it raises ValueError pointing you to the AZURE_AI_API_BASE env var or the api_base parameter. Raised in the sync and async embedding paths before any HTTP traffic.

Source

Thrown at litellm/llms/azure_ai/embed/handler.py:102

            stream=False,
            _response_headers=embedding_headers,
        )
        return returned_response

    def image_embedding(
        self,
        model: str,
        data: ImageEmbeddingRequest,
        timeout: float,
        logging_obj,
        model_response: EmbeddingResponse,
        optional_params: dict,
        api_key: str | None,
        api_base: str | None,
        client: HTTPHandler | AsyncHTTPHandler | None = None,
    ):
        if api_base is None:
            raise ValueError(
                "api_base is None. Please set AZURE_AI_API_BASE or dynamically via `api_base` param, to make the request."
            )
        if api_key is None:
            raise ValueError(
                "api_key is None. Please set AZURE_AI_API_KEY or dynamically via `api_key` param, to make the request."
            )

        if client is None or not isinstance(client, HTTPHandler):
            client = HTTPHandler(timeout=timeout, concurrent_limit=1)

        url: Final = f"{api_base}/images/embeddings"

        response: Final = client.post(
            url=url,
            json=data,
            headers={"Authorization": f"Bearer {api_key}"},
        )

View on GitHub (pinned to 6c2dcb801b)

Solutions

Pass api_base (your Foundry project endpoint, e.g. https://<resource>.services.ai.azure.com) to the embedding call.
Or export AZURE_AI_API_BASE in the runtime environment.
In proxy config, set api_base in the image-embedding deployment's litellm_params.
Confirm the resource actually exposes /images/embeddings (a Cohlete image-embedding deployment) with curl before retrying.

Example fix

# before
litellm.embedding(model='azure_ai/<image-embed-model>', input=[img_b64])

# after
litellm.embedding(
    model='azure_ai/<image-embed-model>',
    input=[img_b64],
    api_base='https://myres.services.ai.azure.com',
    api_key=AZURE_AI_KEY,
)

Defensive patterns

Strategy: validation

Validate before calling

import os

def image_embed_base() -> str:
    base = os.getenv('AZURE_AI_API_BASE')
    if not base:
        raise RuntimeError('image embeddings need AZURE_AI_API_BASE or api_base param')
    return base

Try / catch

try:
    litellm.embedding(model='azure_ai/img', input=imgs, api_base=image_embed_base())
except ValueError as e:
    if 'AZURE_AI_API_BASE' in str(e):
        raise ConfigurationError(str(e)) from e
    raise

Prevention

Extend your startup config check to image-embedding routes, not just chat.
Keep the Foundry endpoint in a shared secret/config store so all Azure AI calls agree.
Unit-test config resolution so missing env fails in CI, not in production.

When it happens

Trigger: litellm.embedding() on an azure_ai image-embedding model without api_base on the call and without AZURE_AI_API_BASE in the environment; config built for text embeddings (which may resolve a default) reused for the image route.

Common situations: Adding image embedding support to an existing text-embedding service and forgetting the Foundry endpoint; env var not propagated to k8s pod; mixing up AZURE_API_BASE vs AZURE_AI_API_BASE.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/aaae8cd22eb0dedf. Report an issue: GitHub.