Google announced today the availability of Gemini, its latest and more powerful Large Language Model. Gemini is multimodal, which means it's able to consume not only text, but also images or videos.
I had the pleasure of working on the Java samples and help with the Java SDK, with wonderful engineer colleagues, and I'd like to share some examples of what you can do with Gemini, using Java!
First of all, you'll need to have an account on Google Cloud and created a project. The Vertex AI API should be enabled, to be able to access the Generative AI services, and in particular the Gemini large language model. Be sure to check out the
instructions.
Preparing your project build
To get started with some coding, you'll need to create a Gradle or a Maven build file that requires the Google Cloud libraries BOM, and the google-cloud-vertexai
library.
Here's an example with Maven:
...
<dependencyManagement>
<dependencies>
<dependency>
<artifactId>libraries-bom</artifactId>
<groupId>com.google.cloud</groupId>
<scope>import</scope>
<type>pom</type>
<version>26.29.0</version>
</dependency>
</dependencies>
</dependencyManagement>
<dependencies>
<dependency>
<groupId>com.google.cloud</groupId>
<artifactId>google-cloud-vertexai</artifactId>
</dependency>
...
</dependencies>
...
Your first queries
Now let's have a look at our first multimodal example, mixing text prompts and images:
try (VertexAI vertexAI = new VertexAI(projectId, location)) {
byte[] imageBytes = Base64.getDecoder().decode(dataImageBase64);
GenerativeModel model = new GenerativeModel("gemini-pro-vision", vertexAI);
GenerateContentResponse response = model.generateContent(
ContentMaker.fromMultiModalData(
"What is this image about?",
PartMaker.fromMimeTypeAndData("image/jpg", imageBytes)
));
System.out.println(ResponseHandler.getText(response));
}
You instantiate VertexAI
with your Google Cloud project ID, and the region location of your choice. To pass images to Gemini, you should either pass the bytes directly, or you can pass a URI of an image stored in a cloud storage bucket (like gs://my-bucket/my-img.jpg
). You create an instance of the model. Here, I'm using gemini-pro-vision
. But later on, a gemini-ultra-vision
model will also be available. Let's ask the model to generate content with the generateContent()
method, by passing both a text prompt, and also an image.
The ContentMaker
and PartMaker
classes are helpers to further simplify the creation of more advanced prompts that mix different modalities. But you could also just pass a simple string as argument of the generateContent()
method.
The ResponseHandler
utility will retrieve all the text of the answer of the model.
Instead of getting the whole output once all the text is generated, you can also adopt a streaming approach:
model.generateContentStream("Why is the sky blue?")
.stream()
.forEach(System.out::print);
You can also iterate over the stream with a for
loop:
ResponseStream<GenerateContentResponse> responseStream =
model.generateContentStream("Why is the sky blue?");
for (GenerateContentResponse responsePart: responseStream) {
System.out.print(ResponseHandler.getText(responsePart));
}
Let's chat!
Gemini is a multimodal model, and it's actually both a text generation model, but also a chat model. So you can chat with Gemini, and ask a series of questions in context. There's a handy ChatSession
utility class which simplifies the handling of the conversation:
try (VertexAI vertexAI = new VertexAI(projectId, location)) {
GenerateContentResponse response;
GenerativeModel model = new GenerativeModel(modelName, vertexAI);
ChatSession chatSession = new ChatSession(model);
response = chatSession.sendMessage("Hello.");
System.out.println(ResponseHandler.getText(response));
response = chatSession.sendMessage("What are all the colors in a rainbow?");
System.out.println(ResponseHandler.getText(response));
response = chatSession.sendMessage("Why does it appear when it rains?");
System.out.println(ResponseHandler.getText(response));
}
This is convenient to use ChatSession
as it takes care of keeping track of past questions from the user, and answers from the assistant.
Going further
This is just a few examples of the capabilities of Gemini. Be sure to check out some of the samples that are available on Github. Read more about Gemini and Generative AI in the Google Cloud documentation.
Top comments (0)