BerriAI/litellm · error · ValueError

api_key is required

Error message

api_key is required

What it means

Azure AI Search vector-store operations authenticate with an api-key header. get_auth_credentials reads litellm_params['api_key'] and raises immediately if absent — there is no env-var fallback in this path, unlike the chat/rerank configs.

Source

Thrown at litellm/llms/azure_ai/vector_stores/transformation.py:71

        ``/analyze`` inside the batch-write path); writes are classified before
        reads, so such a path demands the write grant rather than being
        shadowed into a read.
        """
        return {
            "read": [
                ("GET", "/indexes/"),
                ("POST", "/docs/search"),
                ("POST", "/docs/suggest"),
                ("POST", "/docs/autocomplete"),
                ("POST", "/analyze"),
            ],
            "write": [("POST", "/docs/index")],
        }

    def get_auth_credentials(self, litellm_params: dict) -> BaseVectorStoreAuthCredentials:
        api_key: Final = litellm_params.get("api_key")
        if api_key is None:
            raise ValueError("api_key is required")

        return {
            "headers": {
                "api-key": api_key,
            }
        }

    def validate_environment(self, headers: dict, litellm_params: GenericLiteLLMParams | None) -> dict:
        basic_headers: Final = self._base_validate_azure_environment(headers, litellm_params)
        basic_headers.update({"Content-Type": "application/json"})
        return basic_headers

    def get_complete_url(
        self,
        api_base: str | None,
        litellm_params: dict,
    ) -> str:
        """

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Add api_key to the vector store's litellm_params: litellm_params={'api_key': os.environ['AZURE_SEARCH_API_KEY'], 'azure_search_service_name': 'my-service'}
2. In proxy config, set vector_stores[i].litellm_params.api_key (supports os.environ/ references)
3. Verify with a quick print that the litellm_params dict actually contains a non-None api_key before the call

Example fix

# before
vs_config = {'provider': 'azure_ai_search', 'litellm_params': {'azure_search_service_name': 'my-service'}}

# after
vs_config = {'provider': 'azure_ai_search', 'litellm_params': {'api_key': os.environ['AZURE_SEARCH_ADMIN_KEY'], 'azure_search_service_name': 'my-service'}}

if not litellm_params.get('api_key'):
    raise ValueError('azure_ai_search vector store requires litellm_params["api_key"]')

When it happens

Trigger: Creating or using an azure_ai_search vector store (e.g. litellm.vector_store.create / search with provider azure_ai_search) where the vector store's litellm_params dict has no api_key entry.

Common situations: Configuring the vector store in proxy config.yaml with only azure_search_service_name set; migrating from a provider that pulled keys from env automatically; passing the key under a different name like azure_search_api_key.

Related errors

• Azure AI Search service name is required. Provide it via lit
• embedding_model is required in litellm_params for Azure AI S
• embedding_config is required in litellm_params for Azure AI
• api_key is required for Azure AI Speech transcription.
• AZURE_CLIENT_ID and AZURE_TENANT_ID must be set

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/e9391d3bc4fc5e4c. Report an issue: GitHub.