Awesome

Gemini in Java with Vertex AI and Spring AI

Gemini workshop for Java developers, using the Spring AI orchestration framework

[!NOTE] This is the code for Gemini in Java with Vertex AI and Spring AI codelab geared towards Java developers to discover Gemini and its open-source variant Gemma Large Language Model by Google using Spring AI framework.

Prerequisites

The code examples have been tested on the following environment:

• Java 21
• Maven >= 3.9.6

In order to run these examples, you need to have a Google Cloud account and project ready.

Before running the examples, you'll need to set up three environment variables:

export VERTEX_AI_GEMINI_PROJECT_ID=<your-project-id>
export VERTEX_AI_GEMINI_LOCATION=us-central1
export VERTEX_AI_GEMINI_MODEL=gemini-1.5-pro-001

# Note: you can test in another region or using the gemini-1.5-flash-001 model

[!IMPORTANT] Please update the project ID and location to match your project and select the model of your choice

Create the Maven wrapper:

mvn wrapper:wrapper

Codelab Samples

The samples in this codelab are grouped by various capabilities and patterns. You will find, in order:

• Chat
  • Simple Q&A with Gemini
  • Conversation with Gemini with chat history
  • Simple Q&A via streaming
• Multimodality
  • Analyzing & extracting image data using Multimodality
  • Transcribing audio data using Multimodality
  • Transcribing video data using Multimodality
• Capabilities
  • Structure prompts with prompt templates
  • Extracting structured data from unstructured text
  • Grounding responses with Web Search
  • Function Calling with Spring AI
• Document utilities
  • Document Readers and Splitters
• Embeddings
  • Generating Text Embeddings with Vertex AI
  • Generating Multimodal Embeddings with Vertex AI
• AI use-cases and patterns
  • Retrieval-augmented generation(RAG)
  • Text classification with Few-shot prompting
  • Sentiment analysis with few-shot prompting
  • Summarization Patterns with Gemini: Stuffing, Map-Reduce Patterns
• Local environments
  • Running Open-models with Ollama and Testcontainers

Build

[!TIP] Note the profiles complete used for the build

Build the samples in a single JAR, then run them individually for the respective use-case:

./mvnw clean package -Pcomplete

Run

[!TIP] List of samples, by use-case. Each sample can be run independently

• Chat

  • Simple Q&A with Gemini

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.SimpleChatExample
    

  • Conversation with Gemini with chat history

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.ConversationExample
    

  • Simple Q&A via streaming

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.SimpleChatStreamingExample
    

• Multimodality

  • Analyzing & extracting image data using Multimodality

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.MultimodalImagesExample
    

  • Transcribing audio data using Multimodality

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.MultimodalAudioExample
    

  • Transcribing video data using Multimodality

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.MultimodalVideoExample
    

• Capabilities

  • Structure prompts with prompt templates

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.WorkingWithTemplatesExample
    

  • Extracting structured data from unstructured text

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.StructuredOutputExample
    

  • Grounding responses with Web Search with Vertex SDK

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.GroundingWithWebsearchExample
    

  • Grounding responses with Web Search with SpringAI

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.GroundingWithWebsearchSpringAIExample
    

  • Function Calling with Spring AI

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.FunctionCallingExample
    

• Document utilities

  • Document Readers and Splitters

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.DocumentProcessingExample
    

• Embeddings

  • Generating Text Embeddings with Vertex AI

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.TextEmbeddingExample
    

  • Generating Multimodal Embeddings with Vertex AI

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.MultimodalEmbeddingExample
    

• AI use-cases and patterns

  • Retrieval-augmented generation(RAG)

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.RAGExample
    

  • Text classification with Few-shot prompting

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.TextClassificationExample
    

  • Sentiment analysis with Few-shot prompting

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.SentimentAnalysisExample
    

  • Summarization Patterns with Gemini: Stuffing, Map-Reduce Patterns

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.SummarizationExample
    

• Local environments

  • Running Open-models with Ollama and Testcontainers

    java -cp ./target/spring-ai-workshop-1.0.0-jar-with-dependencies.jar gemini.workshop.LocalTestingWithOllamaContainers
    

This is not an official Google product.