Call Vertex AI models by using the OpenAI library

The Chat Completions API lets you send requests to Vertex AI models by using the OpenAI libraries for Python and REST. If you're already using the OpenAI libraries, you can use this API to switch between calling OpenAI models and Vertex AI hosted models to compare output, cost, and scalability, without changing your existing code. If you aren't already using the OpenAI libraries, we recommend that you call the Gemini API directly.

Supported models

The Chat Completions API supports both Gemini models and select self-deployed models from Model Garden.

Gemini models

The following table shows the Gemini models that are supported:

Model Version
Gemini 1.5 Flash google/gemini-1.5-flash-001
Gemini 1.5 Pro google/gemini-1.5-pro-001
Gemini 1.0 Pro Vision google/gemini-1.0-pro-vision
google/gemini-1.0-pro-vision-001
Gemini 1.0 Pro google/gemini-1.0-pro-002
google/gemini-1.0-pro-001
google/gemini-1.0-pro

Self-deployed models from Model Garden

The HuggingFace Text Generation Interface (HF TGI) and Vertex AI Model Garden prebuilt vLLM containers support the Chat Completions API. However, not every model deployed to these containers supports the Chat Completions API. The following table includes the most popular supported models by container:

HF TGI

vLLM

Authenticate

To use the OpenAI Python libraries, install the OpenAI SDK:

pip install openai

To authenticate with the Chat Completions API, you can either modify your client setup or change your environment configuration to use Google authentication and a Vertex AI endpoint. Choose whichever method that's easier, and follow the steps for setting up depending on whether you want to call Gemini models or self-deployed Model Garden models.

Certain models in Model Garden and supported Hugging Face models need to be deployed to a Vertex AI endpoint first before they can serve requests. When calling these self-deployed models from the Chat Completions API, you need to specify the endpoint ID. To list your existing Vertex AI endpoints, use the gcloud ai endpoints list command.

Client setup

To programmatically get Google credentials in Python, you can use the google-auth Python SDK:

pip install google-auth
pip install requests

Change the OpenAI SDK to point to the Vertex AI chat completions endpoint:

# Programmatically get an access token
creds, project = google.auth.default()
auth_req = google.auth.transport.requests.Request()
creds.refresh(auth_req)
# Note: the credential lives for 1 hour by default (https://cloud.google.com/docs/authentication/token-types#at-lifetime); after expiration, it must be refreshed.

# Pass the Vertex endpoint and authentication to the OpenAI SDK
PROJECT_ID = 'PROJECT_ID'
LOCATION = 'LOCATION'

##############################
# Choose one of the following:
##############################

# If you are calling a Gemini model, set the MODEL_ID variable and set
# your client's base URL to use openapi.
MODEL_ID = 'MODEL_ID'
client = openai.OpenAI(
    base_url = f'https://{LOCATION}-aiplatform.googleapis.com/v1beta1/projects/{PROJECT_ID}/locations/{LOCATION}/endpoints/openapi',
    api_key = creds.token)

# If you are calling a self-deployed model from Model Garden, set the
# ENDPOINT_ID variable and set your client's base URL to use your endpoint.
MODEL_ID = 'MODEL_ID'
client = openai.OpenAI(
    base_url = f'https://{LOCATION}-aiplatform.googleapis.com/v1beta1/projects/{PROJECT_ID}/locations/{LOCATION}/endpoints/{ENDPOINT}',
    api_key = creds.token)

By default, access tokens last for 1 hour. You can extend the life of your access token or periodically refresh your token and update the openai.api_key variable.

Environment variables

Install the Google Cloud CLI. The OpenAI library can read the OPENAI_API_KEY and OPENAI_BASE_URL environment variables to change the authentication and endpoint in their default client. Set the following variables:

$ export PROJECT_ID=PROJECT_ID
$ export LOCATION=LOCATION
$ export OPENAI_API_KEY="$(gcloud auth application-default print-access-token)"

To call a Gemini model, set the MODEL_ID variable and use the openapi endpoint:

$ export MODEL_ID=MODEL_ID
$ export OPENAI_BASE_URL="https://${LOCATION}-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_ID}/locations/${LOCATION}/endpoints/openapi"

To call a self-deployed model from Model Garden, set the ENDPOINT variable and use that in your URL instead:

$ export ENDPOINT=ENDPOINT_ID
$ export OPENAI_BASE_URL="https://${LOCATION}-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_ID}/locations/${LOCATION}/endpoints/${ENDPOINT}"

Next, initialize the client:

client = openai.OpenAI()

The Gemini Chat Completions API uses OAuth to authenticate with a short-lived access token. By default, access tokens last for 1 hour. You can extend the life of your access token or periodically refresh your token and update the OPENAI_API_KEY environment variable.

Call Gemini with the Chat Completions API

The following sample shows you how to send non-streaming requests:

curl

  curl -X POST \
    -H "Authorization: Bearer $(gcloud auth print-access-token)" \
    -H "Content-Type: application/json" \
  https://${LOCATION}-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_ID}/locations/${LOCATION}/endpoints/openapi/chat/completions \
  -d '{
    "model": "google/${MODEL_ID}",
    "messages": [{
      "role": "user",
      "content": "Write a story about a magic backpack."
    }]
  }'

Python

To learn how to install or update the Vertex AI SDK for Python, see Install the Vertex AI SDK for Python. For more information, see the Python API reference documentation. 

import vertexai
import openai

from google.auth import default, transport

# TODO(developer): Update and un-comment below lines
# project_id = "PROJECT_ID"
# location = "us-central1"

vertexai.init(project=project_id, location=location)

# Programmatically get an access token
credentials, _ = default(scopes=["https://www.googleapis.com/auth/cloud-platform"])
auth_request = transport.requests.Request()
credentials.refresh(auth_request)

# # OpenAI Client
client = openai.OpenAI(
    base_url=f"https://{location}-aiplatform.googleapis.com/v1beta1/projects/{project_id}/locations/{location}/endpoints/openapi",
    api_key=credentials.token,
)

response = client.chat.completions.create(
    model="google/gemini-1.5-flash-001",
    messages=[{"role": "user", "content": "Why is the sky blue?"}],
)

print(response)

The following sample shows you how to send streaming requests to a Gemini model by using the Chat Completions API:

curl

  curl -X POST \
    -H "Authorization: Bearer $(gcloud auth print-access-token)" \
    -H "Content-Type: application/json" \
  https://${LOCATION}-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_ID}/locations/${LOCATION}/endpoints/openapi/chat/completions \
  -d '{
    "model": "google/${MODEL_ID}",
    "stream": true,
    "messages": [{
      "role": "user",
      "content": "Write a story about a magic backpack."
    }]
  }'

Python

To learn how to install or update the Vertex AI SDK for Python, see Install the Vertex AI SDK for Python. For more information, see the Python API reference documentation. 

import vertexai
import openai

from google.auth import default, transport

# TODO(developer): Update and un-comment below lines
# project_id = "PROJECT_ID"
# location = "us-central1"

vertexai.init(project=project_id, location=location)

# Programmatically get an access token
credentials, _ = default(scopes=["https://www.googleapis.com/auth/cloud-platform"])
auth_request = transport.requests.Request()
credentials.refresh(auth_request)

# OpenAI Client
client = openai.OpenAI(
    base_url=f"https://{location}-aiplatform.googleapis.com/v1beta1/projects/{project_id}/locations/{location}/endpoints/openapi",
    api_key=credentials.token,
)

response = client.chat.completions.create(
    model="google/gemini-1.5-flash-001",
    messages=[{"role": "user", "content": "Why is the sky blue?"}],
    stream=True,
)
for chunk in response:
    print(chunk)

Call a self-deployed model with the Chat Completions API

The following sample shows you how to send non-streaming requests:

  curl -X POST \
    -H "Authorization: Bearer $(gcloud auth print-access-token)" \
    -H "Content-Type: application/json" \
  https://us-central1-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_ID}/locations/us-central1/endpoints/${ENDPOINT}/chat/completions \
  -d '{
    "messages": [{
      "role": "user",
      "content": "Write a story about a magic backpack."
    }]
  }'

The following sample shows you how to send streaming requests to a self-deployed model by using the Chat Completions API:

  curl -X POST \
    -H "Authorization: Bearer $(gcloud auth print-access-token)" \
    -H "Content-Type: application/json" \
  https://us-central1-aiplatform.googleapis.com/v1beta1/projects/${PROJECT_ID}/locations/us-central1/endpoints/${ENDPOINT}/chat/completions \
  -d '{
    "stream": true,
    "messages": [{
      "role": "user",
      "content": "Write a story about a magic backpack."
    }]
  }'

Supported parameters

For Google models, the Chat Completions API supports the following OpenAI parameters. For a description of each parameter, see OpenAI's documentation on Creating chat completions. Parameter support for third-party models varies by model. To see which parameters are supported, consult the model's documentation.

messages
  • System message
  • User message: The text and image_url types are supported. The image_url type supports images stored a Cloud Storage URI or a base 64 encoding in the form "data:<MIME-TYPE>;base64,<BASE64-ENCODED-BYTES>". To learn how to create a Cloud Storage bucket and upload a file to it, see Discover object storage. The detail option is not supported.
  • Assistant message
  • Tool message
  • Function message: This field is deprecated, but supported for backwards compatibility.
model
max_tokens
n
response_format
  • json_object: Interpreted as passing "application/json" to the Gemini API.
  • text: Interpreted as passing "text/plain" to the Gemini API.
  • Any other MIME type is passed as is to the model, such as passing "application/json" directly.
stop
stream
temperature
top_p
tools
  • type
  • function
    • name
    • description
    • parameters: Specify parameters by using the OpenAPI specification. This differs from the OpenAI parameters field, which is described as a JSON Schema object. To learn about keyword differences between OpenAPI and JSON Schema, see the OpenAPI guide.
tool_choice
  • none
  • auto
  • required: Corresponds to the mode ANY in the FunctionCallingConfig.
function_call This field is deprecated, but supported for backwards compatibility.
functions This field is deprecated, but supported for backwards compatibility.

If you pass any unsupported parameter, it is ignored.

Refresh your credentials

The following example shows how to refresh your credentials automatically as needed:

Python 

from typing import Any

import google.auth
import google.auth.transport.requests
import openai

class OpenAICredentialsRefresher:
    def __init__(self, **kwargs: Any) -> None:
        # Set a dummy key here
        self.client = openai.OpenAI(**kwargs, api_key="DUMMY")
        self.creds, self.project = google.auth.default(
            scopes=["https://www.googleapis.com/auth/cloud-platform"]
        )

    def __getattr__(self, name: str) -> Any:
        if not self.creds.valid:
            auth_req = google.auth.transport.requests.Request()
            self.creds.refresh(auth_req)

            if not self.creds.valid:
                raise RuntimeError("Unable to refresh auth")

            self.client.api_key = self.creds.token
        return getattr(self.client, name)

# TODO(developer): Update and un-comment below lines
# project_id = "PROJECT_ID"
# location = "us-central1"

client = OpenAICredentialsRefresher(
    base_url=f"https://{location}-aiplatform.googleapis.com/v1beta1/projects/{project_id}/locations/{location}/endpoints/openapi",
)

response = client.chat.completions.create(
    model="google/gemini-1.5-flash-001",
    messages=[{"role": "user", "content": "Why is the sky blue?"}],
)

print(response)

What's next