langchain4j-testing-strategies_skill

Overview

This skill provides practical testing strategies for LangChain4J-powered applications, focusing on reliable unit, integration, and end-to-end tests. It covers mocking LLMs, Testcontainers-based integration, RAG validation, and patterns for streaming and async workflows. Use it to make AI-driven features testable, repeatable, and fast.

How this skill works

The skill prescribes a layered approach: fast unit tests with mocked ChatModel and EmbeddingModel, integration tests that run real models inside Testcontainers, and targeted end-to-end tests for RAG and tool execution. It includes code patterns, dependency guidance, assertion helpers, and advice on timeouts, memory cleanup, and test profiles to ensure isolation and repeatability.

When to use it

• When writing unit tests for AI services and guardrails
• When validating retrieval-augmented generation (RAG) workflows
• When setting up integration tests using Testcontainers and real models
• When mocking LLM responses for fast CI runs
• When testing streaming responses, async flows, or tool execution

Best practices

• Favor mocks for unit tests; never call real APIs in unit scope
• Follow the test pyramid: ~70% unit, 20% integration, 10% end-to-end
• Isolate tests with @BeforeEach/@AfterEach and test-specific profiles
• Use container reuse and timeouts for integration tests to reduce flakiness
• Cover both success and failure cases, including edge inputs and injection attempts

Example use cases

• Mock ChatModel to assert business logic and guardrail enforcement in service layer tests
• Run an Ollama container via Testcontainers to validate a RAG assistant end-to-end
• Simulate streaming model output to verify partial-result handling and cancellation
• Seed an in-memory embedding store to test retriever relevance and ranking
• Assert timeout and retry behavior for slow external model calls