langchain4j-testing-strategies

The langchain4j-testing-strategies skill provides unit, integration, and end-to-end testing patterns for LangChain4j-based Java AI applications, solving the problem of validating non-deterministic LLM behavior in a reliable, repeatable way. It covers mocking AI models for fast isolated tests, testing retrieval chains and RAG workflows, and running Testcontainers-based integration tests against real services.

The skill demonstrates unit testing by mocking ChatModel with Mockito to return deterministic responses, wiring the mock into an AiServices builder, and asserting on outputs. It lays out the required testing dependencies (langchain4j-test for guardrail assertions, testcontainers, mockito, and assertj), and shows integration testing with Testcontainers, including a containerized Ollama example with health checks that wait for the API server to be ready before generating responses. It advocates a testing pyramid of roughly 70% unit tests, 20% integration tests, and 10% end-to-end tests, and includes troubleshooting guidance for container startup and model responsiveness. It bundles six files including references for unit testing, integration testing, advanced testing, workflow patterns, and testing dependencies, and declares safe allowed-tools (Read, Write, Edit, Bash, Glob, Grep).

It targets Java developers and QA engineers building LLM-powered applications who need deterministic, well-structured test suites for AI services, RAG pipelines, and tool execution. Typical use cases include mocking models to avoid external API calls, validating retrieval accuracy, and standing up ephemeral vector stores or model servers for integration tests. The content is testing guidance and example code with no risky behavior.