CAMEL-22851: Implement native tool-search-tool for langchain4j-tools component

[gnodet]

Implements CAMEL-22851.

Overview

This PR implements a native tool-search-tool feature for the camel-langchain4j-tools component that allows LLMs to discover and access tools dynamically without consuming the entire context window with tool definitions.

Problem Statement

When working with LLMs that support function calling, exposing all available tools in every request can:

• Consume significant context window space, reducing space for actual conversation
• Overwhelm the LLM with too many options, potentially degrading performance
• Limit scalability when dealing with hundreds or thousands of tools

Solution

This PR introduces an exposed parameter (default: true) that allows tools to be marked as "searchable" rather than immediately exposed to the LLM. A native toolSearchTool is automatically provided when searchable tools exist, enabling the LLM to discover tools on-demand based on tags.

Key Features

1. Exposed Parameter

• New exposed boolean parameter on LangChain4jToolsEndpoint (default: true)
• Tools with exposed=false are added to a searchable registry
• Maintains backward compatibility - existing code works unchanged

2. Native Tool Search Tool

• Automatically exposed when searchable tools exist
• Searches by tags with support for comma-separated lists
• Returns formatted tool descriptions for LLM consumption
• Prevents duplicate results when tools have multiple matching tags

3. Proper Resource Management

• Endpoints only remove their own tools on shutdown (fixes potential memory leak)
• Separate caches for exposed and searchable tools
• Clean separation of concerns

Usage Example

// Exposed tool - immediately available to LLM
from("langchain4j-tools:queryById?tags=users&description=Query user by ID&parameter.userId=integer")
    .to("sql:SELECT name FROM users WHERE id = :#userId");

// Searchable tool - discoverable via toolSearchTool
from("langchain4j-tools:queryBySSN?tags=users&description=Query user by SSN&parameter.ssn=string&exposed=false")
    .to("sql:SELECT name FROM users WHERE ssn = :#ssn");

// Another searchable tool with different tags
from("langchain4j-tools:sendEmail?tags=users,email&description=Send email&parameter.email=string&parameter.message=string&exposed=false")
    .to("smtp://mailserver");

Benefits

• Reduced Context Usage: Only expose commonly used tools initially
• Scalability: Support hundreds or thousands of tools without overwhelming the LLM
• Dynamic Discovery: LLM discovers tools as needed based on conversation context
• Better Organization: Tag-based tool grouping for easier discovery

[github-actions]

🌟 Thank you for your contribution to the Apache Camel project! 🌟

🤖 CI automation will test this PR automatically.

🐫 Apache Camel Committers, please review the following items:

• First-time contributors require MANUAL approval for the GitHub Actions to run

• You can use the command /component-test (camel-)component-name1 (camel-)component-name2.. to request a test from the test bot.

• You can label PRs using build-all, build-dependents, skip-tests and test-dependents to fine-tune the checks executed by this PR.

• Build and test logs are available in the Summary page. Only Apache Camel committers have access to the summary.

• ⚠️ Be careful when sharing logs. Review their contents before sharing them publicly.

[Croway]

@gnodet For the integration test could you use https://github.com/apache/camel/blob/main/components/camel-ai/camel-langchain4j-tools/src/test/java/org/apache/camel/component/langchain4j/tools/integration/LangChain4jToolIT.java#L30 ?

And the disable the test on the ci.

with the ollama test-infra extension you can run the test following https://github.com/apache/camel/blob/main/components/camel-ai/camel-langchain4j-tools/test-execution.md

in your case mvn verify -Dollama.endpoint=http://localhost:11434/ -Dollama.model=llama3.1:latest -Dollama.instance.type=remote

From time to time I do execute the langchain4j tests locally using mvn verify -Dollama.endpoint=http://localhost:11434/ -Dollama.model=granite4:3b -Dollama.instance.type=remote

[orpiske]

Nice! Looking forward to having this one on the code