Skip to content

giuseppe-trisciuoglio/developer-kit / plugins/developer-kit-java/skills/langchain4j-tool-function-calling-patterns

langchain4j-tool-function-calling-patterns

Unverified

Provides and generates LangChain4j tool and function calling patterns: annotates methods as tools with @Tool, configures tool executors, registers tools with AiServices, validates tool parameters, and handles tool execution errors. Use when building AI agents that call tools, define function specifications, manage tool responses, or integrate external APIs with LLM-driven applications.

LangChain4j Tool & Function Calling Patterns

Provides patterns for annotating methods as tools, configuring tool executors, registering tools with AI services, validating parameters, and handling tool execution errors in LangChain4j applications.

Overview

LangChain4j uses the @Tool annotation to expose Java methods as callable functions for AI agents. The AiServices builder registers tools with a chat model, enabling LLMs to perform actions beyond text generation: database queries, API calls, calculations, and business system integrations. Parameters use @P for descriptions that guide the LLM.

When to Use

  • Building AI agents that call external tools (weather, stocks, database queries)
  • Defining function specifications for LLM tool use (@Tool, @P annotations)
  • Registering and managing tool sets with AiServices.builder().tools()
  • Handling tool execution errors, timeouts, and hallucinated tool names
  • Implementing context-aware tools that inject user state via @ToolMemoryId
  • Configuring dynamic tool providers for large or conditional tool sets

Instructions

1. Annotate Methods with @Tool

Define a tool class with methods annotated @Tool. Provide a description as the first parameter. Use @P for each parameter description.

public class WeatherTools {
    private final WeatherService weatherService;

    public WeatherTools(WeatherService weatherService) {
        this.weatherService = weatherService;
    }

    @Tool("Get current weather for a city")
    public String getWeather(
            @P("City name") String city,
            @P("Temperature unit: celsius or fahrenheit") String unit) {
        return weatherService.getWeather(city, unit);
    }
}

Validate: Create an instance and confirm the class loads without errors.

2. Register Tools with AiServices

Use AiServices.builder() to register tool instances with the chat model.

MathAssistant assistant = AiServices.builder(MathAssistant.class)
    .chatModel(chatModel)
    .tools(new Calculator(), new WeatherTools(weatherService))
    .build();

Validate: Call assistant.chat("What is 2 + 2?") and verify the LLM responds without throwing.

3. Test Tool Invocation End-to-End

Send a prompt that triggers tool usage and verify the tool executes and its result is incorporated.

String response = assistant.chat("What is the weather in Rome?");
System.out.println(response);

Validate: Check logs for tool invocation and confirm the response uses the tool output.

4. Handle Tool Execution Errors

Add error handlers to gracefully manage failures without exposing stack traces.

AiServices.builder(Assistant.class)
    .chatModel(chatModel)
    .tools(new ExternalServiceTools())
    .toolExecutionErrorHandler((request, exception) -> {
        logger.error("Tool '{}' failed: {}", request.name(), exception.getMessage());
        return "An error occurred while processing your request";
    })
    .hallucinatedToolNameStrategy(request ->
        ToolExecutionResultMessage.from(request,
            "Error: tool '" + request.name() + "' does not exist"))
    .toolArgumentsErrorHandler((error, context) ->
        ToolErrorHandlerResult.text("Invalid arguments: " + error.getMessage()))
    .build();

Validate: Trigger an error condition and confirm the LLM receives a safe error message.

5. Optimize for Performance and Scale

Enable concurrent tool execution and set timeouts for long-running tools.

AiServices.builder(Assistant.class)
    .chatModel(chatModel)
    .tools(new DbTools(), new HttpTools())
    .executeToolsConcurrently(Executors.newFixedThreadPool(5))
    .toolExecutionTimeout(Duration.ofSeconds(30))
    .build();

Validate: Run concurrent requests and confirm no thread contention or deadlocks.

Examples

Calculator Tool with Full Class

public class Calculator {
    @Tool("Perform basic arithmetic")
    public double calculate(
            @P("Expression like 2+2 or 10*5") String expression) {
        // Parse and evaluate expression
        return eval(expression);
    }
}

Assistant assistant = AiServices.builder(Assistant.class)
    .chatModel(ChatModel.builder()
        .apiKey(System.getenv("API_KEY"))
        .model("gpt-4o")
        .build())
    .tools(new Calculator())
    .build();

Immediate Return Tool (No LLM Response)

@Tool(value = "Send email notification", returnBehavior = ReturnBehavior.IMMEDIATELY)
public void sendEmail(@P("Recipient email address") String to,
                     @P("Email subject") String subject,
                     @P("Email body") String body) {
    emailService.send(to, subject, body);
}

Dynamic Tool Provider

ToolProvider provider = request -> {
    if (request.userContext().contains("admin")) {
        return List.of(new AdminTools());
    }
    return List.of(new UserTools());
};

AiServices.builder(Assistant.class)
    .chatModel(chatModel)
    .toolProvider(provider)
    .build();

Best Practices

  • Descriptive @Tool names: Use imperative verbs ("Get", "Send", "Calculate") with clear scope
  • Precise @P descriptions: Include format, constraints, and valid values — vague descriptions cause incorrect LLM calls
  • Safe error handling: Never expose stack traces; return user-friendly error strings
  • Timeout configuration: Always set .toolExecutionTimeout() for external service calls
  • Concurrent execution: Enable .executeToolsConcurrently() when tools are independent
  • Input validation: Validate parameters inside the tool method; return descriptive errors
  • Permission checks: Perform authorization inside the tool, not at the AI service level
  • Audit logging: Log tool name, parameters, and execution result for debugging and compliance

Common Issues and Solutions

IssueSolution
LLM calls non-existent toolAdd .hallucinatedToolNameStrategy() returning a safe error message
Tools receive wrong parametersRefine @P descriptions; add .toolArgumentsErrorHandler()
Tool execution hangsSet .toolExecutionTimeout(Duration.ofSeconds(N))
Rate limit errors from external APIAdd retry logic or rate limiter inside the tool method
LLM ignores tool outputEnsure the tool returns a string the LLM can interpret

See references/error-handling.md for resilience patterns and references/core-patterns.md for parameter and return type details.

Quick Reference

Annotation / APIPurpose
@ToolMarks a method as a callable tool
@PDescribes a tool parameter for the LLM
@ToolMemoryIdInjects conversation/user ID into the tool
AiServices.builder()Creates AI service with registered tools
ReturnBehavior.IMMEDIATELYExecute tool without waiting for LLM response
ToolProviderDynamic tool provisioning based on context
executeToolsConcurrently()Run independent tool calls in parallel
toolExecutionTimeout()Timeout for individual tool calls

Constraints and Warnings

  • Sensitive data: Never pass API keys, passwords, or credentials in @Tool or @P descriptions
  • Side effects: Tools that modify data should warn in their description; AI models may call them multiple times
  • Large tool sets: Excessive tools confuse LLM models — use ToolProvider for conditional registration
  • Blocking operations: Tools should not perform long synchronous I/O without timeout configuration
  • Stack trace exposure: Always route exceptions through error handlers that return safe strings
  • Parameter precision: Vague @P descriptions directly cause incorrect tool calls — be specific about formats and constraints
  • Concurrent safety: Ensure tool classes are stateless or thread-safe when using executeToolsConcurrently()

Related Skills

  • langchain4j-ai-services-patterns — High-level AI service configuration
  • langchain4j-rag-implementation-patterns — RAG retrieval with tool integration
  • langchain4j-spring-boot-integration — Tool registration in Spring Boot applications

References

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 463kUpdated Claude CodeCodexVerifiedInstall
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 463kUpdated Claude CodeCodexVerifiedInstall
discernment-nudgeanthropics/skillsAfter you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format.Stars: 195kUpdated Claude CodeCodexVerifiedInstall
internal-commsanthropics/skillsA set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Claude should use this skill whenever asked to write some sort of internal communications (status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc.).Stars: 195kUpdated Claude CodeCodexVerifiedInstall