Add a factory able to select one of the supported AI providers

[cmoulliard]

Todo

Add a factory able to select one of the supported AI langchain4j providers: OpenAI, Anthropic, Vert-Anthropic, etc

[cmoulliard]

To achieve this goal, it will be needed to perform some changes like:

• Intercept the args of the users before to launch the Quarkus application

// Example: https://github.com/quarkiverse/quarkus-mcp-servers/blob/b11a6590cd9dd39c7b75742419c072f5819ad983/shared/src/main/java/io/quarkiverse/mcp/servers/shared/SharedApplication.java#L29
    public static void main(String[] args) {
        main(null, args, (remainingArgs) -> null);
    }

    public static void main(Class<? extends QuarkusApplication> app, String[] args) {
        main(app, args, (remainingArgs) -> null);
    }

    public static void main(String[] args, Function<List<String>, Void> onArgsProcessed) {
        main(null, args, onArgsProcessed);
    }

    /**
     *
     * @param app the QuarkusApplication class to run, can be null
     * @param args the arguments to pass to the application
     * @param onArgsProcessed a function that will be called with the remaining arguments after the config source has been
     *        processed
     */
    public static void main(Class<? extends QuarkusApplication> app, String[] args,
            Function<List<String>, Void> onArgsProcessed) {
        onArgsProcessed.apply(McpCliConfigSource.setupConfigSource(args));

        Quarkus.run(app,
                (exitCode, exception) -> {
                    if (exception != null) {
                        exception.printStackTrace();
                    }
                    // If we are not running in dev mode, exit the application
                    if (!LaunchMode.current().isDev())
                        System.exit(exitCode);
                },
                args);
    }

• According to the AI type selected, set the proper ChatModel to be used with the help of a Factory class and arg map to enum field

[cmoulliard]

There is something else to be investigated as today quarkus-langchain4j do not support to access Google Cloud Vertex Api - Anthropic but maybe using OpenAI - langchain4j/langchain4j#4735.

That works nevertheless using Langchain4j instead of quarkus-langchain4j

///usr/bin/env jbang "$0" "$@" ; exit $?
//DEPS dev.langchain4j:langchain4j:1.12.2
//DEPS dev.langchain4j:langchain4j-anthropic:1.12.2
//DEPS dev.langchain4j:langchain4j-vertex-ai-anthropic:1.12.2-beta22
//DEPS org.slf4j:slf4j-simple:2.0.17
import dev.langchain4j.memory.chat.MessageWindowChatMemory;
import dev.langchain4j.model.chat.ChatModel;
import dev.langchain4j.service.AiServices;
import dev.langchain4j.model.vertexai.anthropic.VertexAiAnthropicChatModel;

import java.util.Scanner;

public class VertexAIAnthropic {

    private static final String PROJECT_ID = "itpc-gcp-cp-pe-eng-claude";
    private static final String MODEL_NAME = "claude-opus-4-6";
    private static final String LOCATION = "europe-west1";

    interface Assistant {
        String chat(String message);
    }

    public static void main(String[] args) {

        ChatModel model = VertexAiAnthropicChatModel.builder()
                .project(PROJECT_ID)
                .location(LOCATION)
                .modelName(MODEL_NAME)
                .maxTokens(1000)
                .logRequests(true)
                .logResponses(true)
                .build();

        Assistant assistant = AiServices.builder(Assistant.class)
                .chatModel(model)
                .chatMemory(MessageWindowChatMemory.withMaxMessages(10))
                .build();

        Scanner scanner = new Scanner(System.in);
        System.out.println("Chat started! Type 'exit' to quit.");

        while (true) {
            System.out.print("User: ");
            String input = scanner.nextLine();
            if ("exit".equalsIgnoreCase(input)) break;

            String response = assistant.chat(input);
            System.out.println("AI: " + response);
        }
    }
}

So, do we wait till this ticket is implemented: quarkiverse/quarkus-langchain4j#2243 OR we adapt our code to use directly Langchain4j ?
WDYT: @aureamunoz

[aureamunoz]

I think adding a factory is premature at this point. The real blocker is that quarkus-langchain4j doesn't yet support VertexAiAnthropicChatModel. Given that we are still experimenting the AI side in mtool, I'd suggest we wait for #2243 to be resolved before adding anything on our side.

[cmoulliard]

Given that we are still experimenting the AI side in mtool, I'd suggest we wait for #2243 to be resolved before adding anything on our side.

The temporary solution (instead of having a factory) is to use as I did the code from langchain4j or the project used by langchain4j from Google => https://github.com/googleapis/java-genai without using quarkus-langchain.

WDYT ? @aureamunoz