BerriAI/litellm · error · ValueError

Azure Anthropic requests require an api_base. Set `api_base`

Error message

Azure Anthropic requests require an api_base. Set `api_base` or the AZURE_AI_API_BASE env var.

What it means

ValueError raised when a Claude model under the azure_ai provider routes to the Azure Anthropic handler (the code even normalizes the URL toward .../anthropic and /v1/messages) but AzureFoundryModelInfo.get_api_base finds no endpoint via api_base, litellm.api_base, or AZURE_AI_API_BASE. Claude-on-Foundry deployments are addressed by their deployment URL, so litellm aborts before calling.

Source

Thrown at litellm/main.py:1570

            messages=messages,
            api_base=api_base,
            api_key=api_key,
            model_response=model_response,
            logging_obj=logging,
            optional_params=optional_params,
            litellm_params=litellm_params,
            timeout=timeout,
            acompletion=acompletion,
            stream=stream,
            headers=headers or litellm.headers,
        )

    # Check if this is a Claude model - route to Azure Anthropic handler
    elif "claude" in model.lower():
        # Use Azure Anthropic handler for Claude models
        api_base = AzureFoundryModelInfo.get_api_base(api_base)
        if api_base is None:
            raise ValueError(
                "Azure Anthropic requests require an api_base. Set `api_base` or the AZURE_AI_API_BASE env var."
            )
        api_key = AzureFoundryModelInfo.get_api_key(api_key)

        # Ensure the URL ends with /v1/messages for Anthropic
        if api_base:
            api_base = api_base.rstrip("/")
            if not api_base.endswith("/v1/messages"):
                if "/anthropic" in api_base:
                    parts: Final = api_base.split("/anthropic", 1)
                    api_base = parts[0] + "/anthropic"
                else:
                    api_base = api_base + "/anthropic"
                api_base = api_base + "/v1/messages"

        response = azure_anthropic_chat_completions.completion(
            model=model,
            messages=messages,

View on GitHub (pinned to 77b7c6c40c)

Solutions

Pass api_base set to your Foundry Claude deployment URL (the code will append /v1/messages as needed)
Or export AZURE_AI_API_BASE with that URL
Verify the URL contains the anthropic/deployment segment you expect; the handler splits on '/anthropic' when normalizing
Confirm auth (api_key/AZURE_AI_API_KEY) is also set once the base is fixed

Example fix

# before
resp = litellm.completion(model='azure_ai/claude-3-5-sonnet', messages=m)  # ValueError

# after
import os
os.environ['AZURE_AI_API_BASE'] = 'https://<resource>.services.ai.azure.com/api/projects/<proj>/deployments/claude-3-5-sonnet'
os.environ['AZURE_AI_API_KEY'] = '...'
resp = litellm.completion(model='azure_ai/claude-3-5-sonnet', messages=m)

Defensive patterns

Strategy: validation

Validate before calling

import os
needs_foundry = model.startswith('azure_ai/') and 'claude' in model.lower()
if needs_foundry and not (api_base or os.getenv('AZURE_AI_API_BASE')):
    raise SystemExit('Claude on azure_ai needs api_base or AZURE_AI_API_BASE')

Type guard

def azure_anthropic_ready(model: str, api_base: str | None) -> bool:
    return 'claude' not in model.lower() or bool(api_base or os.getenv('AZURE_AI_API_BASE'))

Try / catch

try:
    resp = litellm.completion(model='azure_ai/claude-3-5-sonnet', messages=m)
except ValueError as e:
    if 'Azure Anthropic requests require an api_base' in str(e):
        raise RuntimeError('Configure the Foundry Claude deployment URL via api_base/AZURE_AI_API_BASE') from e
    raise

Prevention

Keep the Foundry Claude deployment URL in version-controlled config, not shell history
Verify the URL reaches the anthropic deployment; the handler appends /v1/messages itself
When migrating providers, audit env var names -- azure_ai uses AZURE_AI_* not AZURE_*
Add a startup check for every azure_ai model string in your config

When it happens

Trigger: completion(model='azure_ai/<claude-deployment>', ...) with none of api_base kwarg / litellm.api_base / AZURE_AI_API_BASE set; or only AZURE_API_BASE exported, which this path does not consult.

Common situations: Migrating from Azure OpenAI to a Claude model on Azure AI Foundry and reusing the old env setup; the deployment URL (which ends in /deployments/<name>) never configured in the new environment.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/2a95e3d5b2048e99. Report an issue: GitHub.