BerriAI/litellm · error · VertexAIError

vertexai import failed please run `pip install -U "google-cl

Error message

vertexai import failed please run `pip install -U "google-cloud-aiplatform>=1.38"`. Got error: {e}

What it means

This VertexAIError (HTTP 400) is raised by the Vertex Model Garden OpenAI-compatible route (`vertex_ai/openai/{MODEL_ID}` models) when importing the `vertexai` package (google-cloud-aiplatform SDK) or LiteLLM's OpenAILikeChatHandler fails at call time. The import is done lazily inside the completion function, so an ImportError for any reason — package missing, broken install, incompatible dependency — surfaces wrapped in this message with the original error appended.

Source

Thrown at litellm/llms/vertex_ai/vertex_model_garden/main.py:91

        litellm_params: dict,
        vertex_project=None,
        vertex_location=None,
        vertex_credentials=None,
        logger_fn=None,
        acompletion: bool = False,
        client=None,
    ):
        """
        Handles calling Vertex AI Model Garden Models in OpenAI compatible format

        Sent to this route when `model` is in the format `vertex_ai/openai/{MODEL_ID}`
        """
        try:
            import vertexai

            from litellm.llms.openai_like.chat.handler import OpenAILikeChatHandler
        except Exception as e:
            raise VertexAIError(
                status_code=400,
                message=f"""vertexai import failed please run `pip install -U "google-cloud-aiplatform>=1.38"`. Got error: {e}""",
            )

        if not (hasattr(vertexai, "preview") or hasattr(vertexai.preview, "language_models")):
            raise VertexAIError(
                status_code=400,
                message="""Upgrade vertex ai. Run `pip install "google-cloud-aiplatform>=1.38"`""",
            )
        try:
            model = get_vertex_base_model_name(model=model)

            access_token, project_id = self._ensure_access_token(
                credentials=vertex_credentials,
                project_id=vertex_project,
                custom_llm_provider="vertex_ai",
            )

View on GitHub (pinned to 77b7c6c40c)

Solutions

pip install -U "google-cloud-aiplatform>=1.38" in the same environment that runs litellm.
If it still fails, inspect the "Got error: ..." suffix — it contains the original ImportError (often a protobuf/grpcio conflict) and fix that dependency.
Recreate/repair the environment: pip install --force-reinstall google-cloud-aiplatform google-auth requests.
For Docker, add the package to the image rather than relying on runtime install.

Example fix

# before: import fails at call time
litellm.completion(model="vertex_ai/openai/meta-llama/Llama-3.1-405B", ...)

# after: install the SDK first
# pip install -U "google-cloud-aiplatform>=1.38"
litellm.completion(model="vertex_ai/openai/meta-llama/Llama-3.1-405B", ...)

Defensive patterns

Strategy: validation

Validate before calling

def model_garden_deps_ok() -> bool:
    try:
        import vertexai  # noqa: F401
        import litellm.llms.openai_like.chat.handler  # noqa: F401
        return True
    except Exception:
        return False

if not model_garden_deps_ok():
    raise RuntimeError("Run: pip install -U 'google-cloud-aiplatform>=1.38'")

Try / catch

try:
    litellm.completion(model="vertex_ai/openai/llama-3.1-405b-instruct", messages=msgs)
except VertexAIError as e:
    if "vertexai import failed" in str(e):
        raise RuntimeError("google-cloud-aiplatform not installed in this environment") from e
    raise

Prevention

Declare google-cloud-aiplatform>=1.38 as a direct dependency in your project, not an optional extra.
Add an import smoke test (import vertexai) to your deployment healthcheck.
Pin your environment (lockfile or image) so transitive downgrades can't remove the SDK.

When it happens

Trigger: Calling completion with a model like "vertex_ai/openai/llama-3.1-405b-instruct" on an environment where `pip show google-cloud-aiplatform` fails, or where a dependency conflict (e.g. protobuf, grpcio, google-api-core versions) makes `import vertexai` raise.

Common situations: Slim Docker images that installed litellm without the vertex extras; pip dependency resolution downgrading google-cloud-aiplatform to <1.38 or uninstalling it; virtualenvs shared between projects with conflicting google packages.

Understand the failure class

Background: "not installed", "pip install", "required for": how missing-dependency errors surface across open-source libraries — this error's family across 34 libraries.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/488774a7b266d055. Report an issue: GitHub.