Vertex AI
In addition to the consumer-oriented Gemini API, Google provides an enterprise-grade machine learning platform called Vertex AI, which offers managed access to the Gemini models along with enhanced security, governance, and a full suite of MLOps tools.
The difference between Vertex and Gemini is that Gemini is a suite of advanced AI models designed to handle complex multimodal tasks like text, images, audio, and video, that is easy and fast to deploy. While Vertex AI focuses on big companies' infrastructure and tools for custom model development, Gemini delivers cutting-edge, pre-trained AI capabilities for various applications.
Usage
To use Vertex AI, require the Modules.Gemini
module as it is shown in the code example. Create a Gemini.LiveAPIClient
instance via the Gemini.createLiveAPIClient
method and provide your parameters.
Prepare the GEMINI_VERTEX_PROJECT
, GEMINI_VERTEX_LOCATION
, and GEMINI_VERTEX_CREDENTIALS
parameters for the Vertex AI model.
To send media between the Call
and the Gemini.LiveAPIClient
, use the *.sendMedia
or VoxEngine.sendMediaBetween
methods. Listen to the Gemini.LiveAPIClient
events, (see the Gemini.LiveAPIEvents
event list) and implement your application's business logic.
The following code example shows in detail how to connect incoming calls to Vertex AI:
Sessions may be terminated automatically once their limits are reached. Connection lifetime is also limited, but you can use session resumption to keep a session active across multiple connections, and context window compression to extend session duration.
For details, see the official documentation.