Language Model Examples

import dev.langchain4j.model.vertexai.VertexAiLanguageModel;
import dev.langchain4j.model.output.Response;

VertexAiLanguageModel model = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison@001")
    .build();

Response<String> response = model.generate("Write a short poem about clouds");
String text = response.content();
System.out.println(text);

VertexAiLanguageModel model = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison@002")
    .temperature(0.8)              // Controls randomness (0.0-1.0)
    .maxOutputTokens(1000)         // Maximum response length
    .topK(40)                      // Top-K sampling
    .topP(0.95)                    // Nucleus sampling
    .maxRetries(5)                 // Retry attempts on failure
    .build();

Response<String> response = model.generate("Explain quantum computing in simple terms");

String text = response.content();
TokenUsage tokenUsage = response.tokenUsage();
FinishReason finishReason = response.finishReason();

System.out.println("Generated text: " + text);
System.out.println("Input tokens: " + tokenUsage.inputTokenCount());
System.out.println("Output tokens: " + tokenUsage.outputTokenCount());
System.out.println("Total tokens: " + tokenUsage.totalTokenCount());
System.out.println("Finish reason: " + finishReason);

List<String> prompts = List.of(
    "What is machine learning?",
    "What is deep learning?",
    "What is a neural network?"
);

for (String prompt : prompts) {
    Response<String> response = model.generate(prompt);
    System.out.println("Prompt: " + prompt);
    System.out.println("Response: " + response.content());
    System.out.println();
}

String codePrompt = """
    Write a Java function that calculates the factorial of a number.
    Include error handling for negative numbers.
    """;

Response<String> response = model.generate(codePrompt);
System.out.println(response.content());

String longText = """
    [Long article or document text here...]
    """;

String summarizationPrompt = "Summarize the following text in 3 sentences:\n\n" + longText;

Response<String> response = model.generate(summarizationPrompt);
System.out.println("Summary: " + response.content());

String context = """
    The Solar System is the gravitationally bound system of the Sun and the objects
    that orbit it. It formed 4.6 billion years ago from the gravitational collapse of
    a giant interstellar molecular cloud.
    """;

String question = "When did the Solar System form?";
String prompt = "Context: " + context + "\n\nQuestion: " + question + "\n\nAnswer:";

Response<String> response = model.generate(prompt);
System.out.println("Answer: " + response.content());

VertexAiLanguageModel deterministicModel = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison@001")
    .temperature(0.0)  // Most deterministic
    .build();

// Will produce same output for same prompt
Response<String> response = deterministicModel.generate("List 3 programming languages");

VertexAiLanguageModel creativeModel = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison@001")
    .temperature(1.0)  // Most creative
    .build();

// Will produce varied outputs for same prompt
Response<String> response = creativeModel.generate("Write a creative story opener");

VertexAiLanguageModel model = new VertexAiLanguageModel(
    "https://us-central1-aiplatform.googleapis.com/v1/",  // endpoint
    "your-project-id",                                    // project
    "us-central1",                                        // location
    "google",                                             // publisher
    "text-bison@001",                                     // modelName
    0.7,                                                  // temperature
    500,                                                  // maxOutputTokens
    40,                                                   // topK
    0.95,                                                 // topP
    3                                                     // maxRetries
);

VertexAiLanguageModel longContextModel = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison-32k")  // 32k token context
    .build();

String veryLongPrompt = "..."; // Up to 32k tokens
Response<String> response = longContextModel.generate(veryLongPrompt);

VertexAiLanguageModel model = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison@001")
    .maxOutputTokens(50)  // Limit to ~50 tokens
    .build();

Response<String> response = model.generate("Write a long essay about climate change");
// Response will be truncated to ~50 tokens

try {
    Response<String> response = model.generate("Some prompt");
    System.out.println(response.content());
} catch (Exception e) {
    System.err.println("Generation failed: " + e.getMessage());
    // Model automatically retries transient errors up to maxRetries
    // This catches non-retryable errors only
}

String structuredPrompt = """
    Generate a JSON object with the following fields:
    - name: a person's name
    - age: an integer
    - hobbies: an array of strings

    Return only valid JSON, no other text.
    """;

Response<String> response = model.generate(structuredPrompt);
String jsonString = response.content();

// Parse JSON
// ObjectMapper mapper = new ObjectMapper();
// Person person = mapper.readValue(jsonString, Person.class);

import java.util.concurrent.CompletableFuture;
import java.util.List;
import java.util.stream.Collectors;

List<String> prompts = List.of(
    "What is AI?",
    "What is ML?",
    "What is DL?"
);

List<CompletableFuture<Response<String>>> futures = prompts.stream()
    .map(prompt -> CompletableFuture.supplyAsync(() -> model.generate(prompt)))
    .toList();

List<Response<String>> responses = futures.stream()
    .map(CompletableFuture::join)
    .toList();

for (int i = 0; i < responses.size(); i++) {
    System.out.println("Prompt: " + prompts.get(i));
    System.out.println("Response: " + responses.get(i).content());
    System.out.println();
}

// Use lower maxOutputTokens for cost efficiency
VertexAiLanguageModel economicalModel = VertexAiLanguageModel.builder()
    .endpoint("https://us-central1-aiplatform.googleapis.com/v1/")
    .project("your-project-id")
    .location("us-central1")
    .publisher("google")
    .modelName("text-bison@001")
    .maxOutputTokens(200)     // Limit output
    .temperature(0.0)          // Deterministic = fewer retries needed
    .maxRetries(1)             // Fast failure
    .build();