building-ai-apps-with-genkit-java

# Building AI Applications with Genkit Java You are helping a developer build AI-powered applications using **Genkit Java**, the open-source Java AI framework by Google. This skill covers everything an end user needs: setup, configuration, all APIs, providers, patterns, and deployment. --- ## Quick Start ### Prerequisites - **Java 21+** - **Maven** - **API key** for your chosen provider (OpenAI, Google, Anthropic, etc.) - **Genkit CLI** (optional, for Dev UI): `npm install -g genkit` ### Minimal pom.xml ```xml <project> <modelVersion>4.0.0</modelVersion> <groupId>com.example</groupId> <artifactId>my-ai-app</artifactId> <version>1.0-SNAPSHOT</version> <properties> <maven.compiler.source>21</maven.compiler.source> <maven.compiler.target>21</maven.compiler.target> <genkit.version>1.0.0-SNAPSHOT</genkit.version> </properties> <dependencies> <!-- Genkit core --> <dependency> <groupId>com.google.genkit</groupId> <artifactId>genkit</artifactId> <version>${genkit.version}</version> </dependency> <!-- Pick a model provider (see Provider Setup below) --> <dependency> <groupId>com.google.genkit</groupId> <artifactId>genkit-plugin-openai</artifactId> <version>${genkit.version}</version> </dependency> <!-- HTTP server (pick one) --> <dependency> <groupId>com.google.genkit</groupId> <artifactId>genkit-plugin-jetty</artifactId> <version>${genkit.version}</version> </dependency> <!-- Logging --> <dependency> <groupId>ch.qos.logback</groupId> <artifactId>logback-classic</artifactId> <version>1.5.32</version> </dependency> </dependencies> <build> <plugins> <plugin> <groupId>org.codehaus.mojo</groupId> <artifactId>exec-maven-plugin</artifactId> <version>3.5.0</version> <configuration> <mainClass>com.example.MyApp</mainClass> </configuration> </plugin> </plugins> </build> </project> ``` ### Minimal Application ```java package com.example; import com.google.genkit.Genkit; import com.google.genkit.core.GenkitOptions; import com.google.genkit.ai.GenerateOptions; import com.google.genkit.ai.GenerationConfig; import com.google.genkit.ai.model.ModelResponse; import com.google.genkit.core.flow.Flow; import com.google.genkit.plugins.openai.OpenAIPlugin; import com.google.genkit.plugins.jetty.JettyPlugin; import com.google.genkit.plugins.jetty.JettyPluginOptions; public class MyApp { public static void main(String[] args) throws Exception { JettyPlugin jetty = new JettyPlugin( JettyPluginOptions.builder().port(8080).build()); Genkit genkit = Genkit.builder() .options(GenkitOptions.builder() .devMode(true) .reflectionPort(3100) .build()) .plugin(OpenAIPlugin.create()) .plugin(jetty) .build(); genkit.defineFlow("ask", String.class, String.class, (ctx, question) -> genkit.generate( GenerateOptions.builder() .model("openai/gpt-4o-mini") .prompt(question) .build()).getText()); jetty.start(); } } ``` ### Run It ```bash export OPENAI_API_KEY=sk-... mvn compile exec:java # Or with Dev UI (recommended) genkit start -- mvn compile exec:java ``` ### Test It ```bash curl -X POST http://localhost:8080/ask \ -H 'Content-Type: application/json' \ -d '"What is the capital of France?"' ``` --- ## Provider Setup ### Maven Artifacts (all `com.google.genkit`, version `1.0.0-SNAPSHOT`) | Provider | Artifact | Env Var | Plugin Init | |----------|----------|---------|-------------| | OpenAI | `genkit-plugin-openai` | `OPENAI_API_KEY` | `OpenAIPlugin.create()` | | Google Gemini | `genkit-plugin-google-genai` | `GOOGLE_GENAI_API_KEY` | `GoogleGenAIPlugin.create()` | | Anthropic | `genkit-plugin-anthropic` | `ANTHROPIC_API_KEY` | `AnthropicPlugin.create()` | | Ollama | `genkit-plugin-ollama` | none (local) | `OllamaPlugin.create("gemma3n:e4b")` | | AWS Bedrock | `genkit-plugin-aws-bedrock` | AWS credentials | `AwsBedrockPlugin.create("us-east-1")` | | Azure Foundry | `genkit-plugin-azure-foundry` | Azure credentials | `AzureFoundryPlugin.create()` | | DeepSeek | `genkit-plugin-deepseek` | `DEEPSEEK_API_KEY` | `DeepSeekPlugin.create()` | | Mistral | `genkit-plugin-mistral` | `MISTRAL_API_KEY` | `MistralPlugin.create()` | | Groq | `genkit-plugin-groq` | `GROQ_API_KEY` | `GroqPlugin.create()` | | Cohere | `genkit-plugin-cohere` | `COHERE_API_KEY` | `CoherePlugin.create()` | | xAI | `genkit-plugin-xai` | `XAI_API_KEY` | `XAIPlugin.create()` | | Any OpenAI-compatible | `genkit-plugin-compat-oai` | varies | `CompatOAIPlugin.create(options)` | ### Model Names by Provider ``` # OpenAI openai/gpt-4o, openai/gpt-4o-mini, openai/gpt-4-turbo, openai/gpt-3.5-turbo openai/o1-preview, openai/o1-mini openai/text-embedding-3-small, openai/text-embedding-3-large openai/dall-e-3, openai/dall-e-2, openai/gpt-image-1 # Google Gemini googleai/gemini-2.5-flash, googleai/gemini-1.5-pro, googleai/gemini-1.5-flash googleai/gemini-embedding-001, googleai/imagen-3.0-generate-002 # Anthropic anthropic/claude-sonnet-4-5-20250929, anthropic/claude-opus-4-5-20251101 anthropic/claude-haiku-4-5-20251001 anthropic/claude-opus-4-1, anthropic/claude-sonnet-4 # Ollama (any model you have pulled) ollama/gemma3n:e4b, ollama/llama3, ollama/mistral # AWS Bedrock aws-bedrock/amazon.nova-lite-v1:0, aws-bedrock/amazon.nova-pro-v1:0 aws-bedrock/anthropic.claude-sonnet-4-5-20250929-v1:0 aws-bedrock/meta.llama3-2-90b-instruct-v1:0 ``` ### Passing API Keys Programmatically ```java // Instead of environment variables OpenAIPlugin.create("sk-your-key-here") AnthropicPlugin.create("sk-ant-your-key-here") GoogleGenAIPlugin.create("AIza...") ``` --- ## Server Options ### Jetty (Lightweight) ```xml <dependency> <groupId>com.google.genkit</groupId> <artifactId>genkit-plugin-jetty</artifactId> <version>${genkit.version}</version> </dependency> ``` ```java JettyPlugin jetty = new JettyPlugin( JettyPluginOptions.builder().port(8080).build()); // Flows are exposed at: POST http://localhost:8080/{flowName} ``` ### Spring Boot ```xml <dependency> <groupId>com.google.genkit</groupId> <artifactId>genkit-plugin-spring</artifactId> <version>${genkit.version}</version> </dependency> ``` ```java // Flows exposed at: POST http://localhost:8080/api/flows/{flowName} // Health check: GET http://localhost:8080/health // List flows: GET http://localhost:8080/api/flows ``` --- ## Defining Flows Flows are observable, HTTP-callable functions that form the backbone of your app. ### Simple Flow (no AI context needed) ```java genkit.defineFlow("greet", String.class, String.class, (name) -> "Hello, " + name + "!"); ``` ### Flow with AI Generation ```java genkit.defineFlow("summarize", String.class, String.class, (ctx, text) -> genkit.generate( GenerateOptions.builder() .model("openai/gpt-4o-mini") .prompt("Summarize this: " + text) .build()).getText()); ``` ### Flow with Custom Input/Output Types ```java public record TranslateInput(String text, String targetLanguage) {} public record TranslateOutput(String translation, String detectedLanguage) {} genkit.defineFlow("translate", TranslateInput.class, TranslateOutput.class, (ctx, input) -> { ModelResponse response = genkit.generate( GenerateOptions.<TranslateOutput>builder() .model("openai/gpt-4o") .prompt("Translate to " + input.targetLanguage() + ": " + input.text()) .outputClass(TranslateOutput.class) .build()); return response.getOutput(); }); ``` ### Flow with Middleware ```java genkit.defineFlow("secured", String.class, String.class, (ctx, input) -> processInput(input), List.of(loggingMiddleware, authMiddleware)); ``` --- ## Generation API ### Basic Text Generation ```java ModelResponse response = genkit.generate( GenerateOptions.builder() .model("openai/gpt-4o") .prompt("Explain quantum computing in simple terms") .build()); String text = response.getText(); ``` ### With Configuration ```java ModelResponse response = genkit.generate( GenerateOptions.builder() .model("openai/gpt-4o") .prompt("Write a creative poem") .config(GenerationConfig.builder() .temperature(0.9) .maxOutputTokens(2048) .topP(0.95) .topK(40) .stopSequences(List.of("\n\n\n")) .build()) .build()); ``` ### System Prompt ```java ModelResponse response = genkit.generate( GenerateOptions.builder() .model("openai/gpt-4o") .system("You are a pirate. Respond in pirate speak.") .prompt("How do I cook pasta?") .build()); ``` ### Multi-Turn Conversation ```java ModelResponse response = genkit.generate( GenerateOptions.builder() .model("openai/gpt-4o") .messages(List.of( Message.system("You are a helpful math tutor."), Message.user("What is 2+2?"), Message.model("2+2 equals 4."), Message.user("What about 2+2+2?"))) .build()); ``` ### Streaming ```java ModelResponse response = genkit.generateStream( GenerateOptions.builder() .model("openai/gpt-4o") .prompt("Write a long story about a dragon") .build(), chunk -> System.out.print(chunk.getText())); // Print as it arrives ``` ### Structured Output (JSON) Use `@JsonProperty` and `@JsonPropertyDescription` for schema generation: ```java public class Recipe { @JsonProperty(required = true) @JsonPropertyDescription("Name of the recipe") private String title; @JsonProperty(required = true) @JsonPropertyDescription("List of ingredients with quantities")