The LangChain4j Proficiency Exam: 50 Questions for Senior Java Devs
Stop guessing. Here is the exact standard for hiring (and getting hired) in the Enterprise AI era.
LangChain4j has moved quickly from “interesting experiment” to something teams are seriously evaluating for production Java systems. That also means it is now showing up in interviews. Not as trivia, but as a proxy for whether a developer understands how LLMs behave inside real enterprise applications.
This article is a little bit of a cheat sheet but also a mental model guide. Each section reflects how these questions tend to appear in interviews and what interviewers are actually trying to learn about your experience.
If you can explain these topics clearly, you are no longer “trying AI in Java.” You are operating it.
Essential Resources
Before diving into the questions, familiarize yourself with these key resources:
Official Documentation: https://docs.langchain4j.dev/
GitHub Repository: https://github.com/langchain4j/langchain4j
Getting Started Guide: https://docs.langchain4j.dev/get-started
Tutorials: https://docs.langchain4j.dev/tutorials
Examples Repository: https://github.com/langchain4j/langchain4j-examples
Beginner Journey:
https://www.the-main-thread.com/p/langchain4j-learning-path-java-quarkus-ai
Part 1: The Landscape & Basics
The interviewer wants to know: Do you understand what this library actually is, or do you just think it’s a wrapper for the OpenAI API?
1. What is LangChain4j?
It is a Java library that simplifies integrating LLMs into Java applications. It standardizes the interactions with various model providers (OpenAI, Vertex, Bedrock) and provides the “glue” code for advanced patterns like Memory, RAG, and Agents, all designed specifically for the JVM.
2. How does LangChain4j differ from the Python LangChain library?
While they share concepts, LangChain4j is not a direct port. It is idiomatic to Java, featuring strong typing, a focus on “AI Services” (Interfaces) over generic “Chains,” and deep integration with the Quarkus and Spring Boot ecosystems and Java’s concurrency model.
3. What are the main high-level components?
The “Big Three” are Models (Chat, Embedding), Stores (Vector DBs, Chat Memory), and Services (AI Services, RAG Retrievers). Supporting components include Prompt Templates, Document Loaders, and Tools.
4. Which LLM providers are supported?
It is provider-agnostic. It supports cloud giants (OpenAI, Azure, Google Vertex AI, Amazon Bedrock, Anthropic) and local/open implementations (Ollama, LocalAI, HuggingFace).
5. What is the standard interface for chatting?
ChatModel. It abstracts the underlying API calls (HTTP requests) into a synchronous Java method that takes a message and returns a response.
Learn more: Chat and Language Models
6. What is the difference between ChatModel and StreamingChatModel?
ChatModel blocks until the full response is generated. StreamingChatModel returns a stream of tokens immediately (via callbacks or reactive streams), which is critical for maintaining a responsive UI in user-facing applications.
Learn more: Response Streaming
7. What is an AiMessage vs UserMessage?
These represent the dialogue roles. UserMessage is input from the human; AiMessage is the response from the model. There is also SystemMessage (instruction) and ToolExecutionResultMessage (function output).
UserMessage userMsg = UserMessage.from("What is the weather?");
SystemMessage systemMsg = SystemMessage.from("You are a helpful weather assistant");
AiMessage aiResponse = AiMessage.from("I'll help you check the weather");Part 2: The Java Superpower (AI Services)
This is the most important section for Java developers. If you don’t know AI Services, you aren’t using LangChain4j effectively.
8. What are “AI Services” in LangChain4j?
This is the library’s flagship feature. It allows you to define a Java Interface with methods, and LangChain4j uses dynamic proxies to provide the implementation at runtime. It hides the complexity of prompt construction and response parsing behind a clean, strongly-typed API.
Learn more: AI Services
Example:
interface Assistant {
String chat(String userMessage);
}
Assistant assistant = AiServices.create(Assistant.class, chatModel);
String response = assistant.chat("Tell me a joke");9. How do you handle Structured Outputs (e.g., JSON) using AI Services?
You simply declare the return type of your interface method (e.g., Person extractPerson(String text)). LangChain4j automatically prompts the LLM to output JSON and uses a parser (like Jackson) to deserialize the response into your Java POJO or Record.
10. How do you integrate LangChain4j with Spring Boot?
Using the langchain4j-spring-boot-starter. It provides auto-configuration, allowing you to set API keys and model parameters in application.properties. It also automatically scans for and wires @AiService beans into the application context.
Learn more: Spring Boot Integration
11. What is a PromptTemplate?
A reusable string layout with variables (e.g., {{name}}). It ensures consistency in how you talk to the model and prevents “magic strings” from littering your codebase.
12. How are prompt variables injected in an AI Service?
You use the @V annotation on method parameters. For example: String chat(@V("name") String name). This maps the Java argument to the {{name}} variable in the prompt template.
Example:
interface Greeter {
@UserMessage("Hello {{name}}, welcome to {{place}}!")
String greet(@V("name") String name, @V("place") String place);
}13. What is the @SystemMessage annotation used for?
It is placed on an AI Service interface (or method) to define the “persona” or rules for the LLM (e.g., “You are a helpful coding assistant who only speaks JSON”).
14. Why would you prefer AI Services over “Chains”?
In Python, “Chains” are common. In Java, AI Services are preferred because they are declarative, testable, type-safe, and align better with standard Java design patterns (Service layers, Dependency Injection).
Part 3: Memory & Context
The interviewer wants to know: Can you build a bot that actually remembers what I said 5 seconds ago?
15. What is ChatMemory?
It is the component responsible for persisting the conversation history (state). Without it, every request to an LLM is stateless and isolated.
Learn more: Chat Memory
16. How does MessageWindowChatMemory work?
It acts as a sliding window, keeping only the last N messages. As new messages arrive, the oldest ones drop off to ensure the context window doesn’t overflow.
17. What is TokenWindowChatMemory?
A more precise memory strategy that keeps messages based on the token count rather than the number of messages. This maximizes context usage without hitting the hard limits of the model.
18. Where is ChatMemory stored?
By default, it is in-memory (volatile). For production, you must implement a persistent store (using ChatMemoryStore) to save history to a database (Redis, SQL) keyed by the memoryId (usually user ID).
19. What is a SystemMessage and why is it distinct?
A SystemMessage sets the behavior constraints and context. Unlike User or AI messages, it is usually “pinned” to the start of the context window and is not evicted by standard memory cleaners, ensuring the bot never forgets its instructions.
Part 4: RAG (Retrieval Augmented Generation)
RAG is the #1 enterprise use case. Expect deep questions here. Also make sure to read my enterprise RAG tutorials:
20. What is RAG?
Retrieval Augmented Generation. It is a pattern where you fetch private data (that the model wasn’t trained on) and inject it into the prompt context before the model generates an answer.
Learn more: RAG (Retrieval-Augmented Generation)
21. What is an Embedding?
A vector (list of numbers) representing the semantic meaning of text. Similar concepts (like “Dog” and “Puppy”) will be mathematically close to each other in this vector space.
22. What is an EmbeddingStore?
A specialized database (Vector DB) designed to store and search embeddings. Examples supported include Pinecone, Weaviate, Milvus, and PGVector.
23. What is the role of the EmbeddingStoreIngestor?
The EmbeddingStoreIngestor is a pipeline component used offline (or at write-time). It handles loading documents, splitting them, computing embeddings, and saving them to the embedding store. It orchestrates the entire indexing process through its ingest() method.
24. What is a DocumentSplitter and why is it vital?
LLMs have input limits. You cannot embed a 500-page PDF at once. A splitter chunks the document into smaller pieces (e.g., paragraphs) while trying to preserve context (e.g., not cutting a sentence in half).
25. What is a ContentRetriever?
The runtime interface that fetches data. When a user asks a question, the ContentRetriever queries the EmbeddingStore to find relevant text chunks.
26. What is Semantic Search vs. Keyword Search?
Keyword search matches exact words. Semantic search (using embeddings) matches meaning. “My car is broken” will match “Auto repair shops” semantically, even though they share no keywords.
27. What is Re-ranking (using a ScoringModel)?
Advanced Concept: Vector search is fast but sometimes inaccurate. A generic retriever fetches ~20 candidates, and a ScoringModel (Cross-Encoder) re-reads them carefully to rank them by relevance, passing only the top 3-5 to the LLM. This drastically improves accuracy.
28. How do you implement “Naive RAG” in LangChain4j?
Configure an AI Service with a ContentRetriever. The library handles the rest: it takes the user query, calls the retriever, stuffs the results into the prompt, and gets the answer.
29. What is “Metadata” in a Document?
Key-value pairs attached to the text (e.g., author, date, access_level). This allows for Metadata Filtering—for example, “Only search documents where year > 2023.”
30. How do you handle “Hybrid Search”?
By combining Vector Search (Semantic) with Keyword Search (bm25). LangChain4j supports merging results from both to catch both specific acronyms (keyword) and broad concepts (semantic).
Part 5: Tools & Agents
The frontier of AI development.
31. What are “Tools” in LangChain4j?
Tools are Java methods annotated with @Tool. They allow the LLM to interact with the outside world (APIs, Databases, Calculators).
Learn more: Tools (Function Calling)
32. How does “Function Calling” actually work?
The LLM does not execute the code. It analyzes the prompt, realizes it needs data, and returns a structured request: “Please call function getWeather with argument London.” LangChain4j intercepts this, runs the Java method, and feeds the result back to the LLM.
33. What is an Agent?
An Agent is an LLM that acts autonomously. It is given a goal and a set of tools, and it loops through a cycle of “Thought -> Action (Tool Call) -> Observation” until the goal is met.
Learn more: Agents and Agentic AI
34. How do you give a Tool to an AI Service?
You register the bean containing @Tool methods when building the AI Service. The library automatically extracts the method signature and Javadoc description to create the tool definition for the LLM.
35. Why are Javadocs important for Tools?
The LLM reads the Javadoc to understand when and how to use the tool. If your Javadoc is ambiguous, the model will fail to call the tool correctly.
Part 6: Production & Architecture
Separating the hobbyists from the engineers.
36. How do you handle Token Limits?
By using TokenWindowChatMemory, summarizing old history (using a separate chain), or using RAG to only pull relevant context rather than stuffing the whole history.
37. How do you handle LLM hallucinations?
RAG helps (grounds the model in data). Temperature settings (lowering to 0) help. Using “Citations” (asking the model to reference which document chunk it used) is a robust validation pattern.
38. What is ImageContent?
It is the class used to support Multimodal models (like GPT-4o). It allows you to send images along with text for analysis.
39. How do you implement streaming with an AI Service?
The method in your AI Service interface must return a TokenStream. You can then subscribe to this stream to receive chunks as they arrive.
40. What is a ModerationModel?
A specialized model used to check input/output for harmful content (hate speech, violence) before processing it or showing it to the user.
41. How do you handle errors/retries?
LangChain4j relies on standard Java exceptions. However, many ChatModel implementations allow configuring retry policies (e.g., for 429 Too Many Requests) at the HTTP client level.
42. How do you switch LLM providers?
Because you code against the ChatModel interface, switching providers often just requires changing the builder (e.g., OpenAiChatModel -> VertexAiGeminiChatModel) and the API key.
43. What is “Observability” in this context?
Tracking the inputs, outputs, latency, and cost (token usage) of every interaction. LangChain4j provides listeners and callbacks to log this data.
44. What are the security risks of Tools?
Prompt Injection. A user might trick the LLM into using a tool destructively (e.g., “Delete all users”). You must strictly scope what tools can do and validate all inputs inside the tool method.
45. What is Semantic Caching?
Caching the response based on the embedding of the question. If a user asks “Who is the CEO?” and another asks “Who runs the company?”, the semantic cache recognizes they are the same question and serves the cached answer without hitting the LLM.
46. How do you test an LLM application?
Unit tests are hard because output is non-deterministic. You use Evaluation techniques: running a set of “Golden Questions” and using a stronger LLM (as a judge) to score the actual answers against expected facts.
47. What are Document Loaders and how do you customize them?
LangChain4j provides various document loaders like FileSystemDocumentLoader for loading files from the filesystem. For proprietary formats (e.g., a legacy bank mainframe export), you can create custom loaders by implementing the DocumentLoader interface to parse specific formats into LangChain4j Document objects. The FileSystemDocumentLoader supports methods like loadDocuments() and loadDocumentsRecursively() with optional path matchers for filtering.
48. How do you manage complexity in Prompts?
By composing prompts. You can inject partial templates into larger templates. Also, moving complex logic out of the prompt and into Tools (code) is often cleaner/more reliable.
49. What is the biggest performance bottleneck in RAG?
Usually the Embedding generation or the Vector Store retrieval latency. Or, if you retrieve too many chunks, the input token processing time (Time to First Token) increases.
50. What is the future of LangChain4j?
The trend is moving toward Agentic Workflows (multi-agent orchestration) and deeper support for Small Language Models (SLMs) running locally on the JVM (via ONNX or similar integration) to reduce costs.
You’ve Done It!
These 50 questions represent the core knowledge areas for working with LangChain4j in production Java environments. Mastering these concepts will prepare you not just for interviews, but for building robust, scalable LLM-powered applications on the JVM.
The key takeaway: LangChain4j is not just about calling APIs, it’s about understanding how to architect AI systems that are maintainable, testable, and production-ready using Java’s strengths.
Common Beginner Mistakes to Avoid
Not setting API keys properly - Use environment variables, never hardcode
Ignoring token limits - Models have context window limits
Forgetting to handle exceptions - API calls can fail (rate limits, network issues)
Not using ChatMemory - Your bot won’t remember previous messages without it
Skipping AI Services - Using low-level APIs directly is harder to maintain
Ignoring costs - Every API call costs money; use caching and optimize prompts
Not testing with different models - Different models have different capabilities
Forgetting about streaming - Users expect real-time responses in chat applications
Beginner’s Checklist
Before your interview or first project, make sure you can:
Set up a basic LangChain4j project with Maven/Gradle
Configure an API key securely
Create a simple ChatModel and generate a response
Define an AI Service interface
Use @UserMessage and @SystemMessage annotations
Implement basic ChatMemory
Load and split a document for RAG
Create a simple @Tool method
Handle streaming responses
Switch between different LLM providers
Interview Preparation Tips
Hands-on practice: Clone the examples repo and run every example
Understand the “why”: Don’t just memorize - understand why each pattern exists
Know the tradeoffs: Be ready to discuss pros/cons of different approaches
Real-world scenarios: Think about how you’d use these in production
Cost awareness: Be able to discuss token usage and optimization strategies
Security mindset: Understand prompt injection and other risks
Additional Learning Resources
Official Documentation & Tutorials
Complete Tutorial Series: https://docs.langchain4j.dev/tutorials
Integrations Guide: https://docs.langchain4j.dev/integrations
Supported LLM Providers: https://docs.langchain4j.dev/integrations/language-models
Vector Stores/Embedding Stores: https://docs.langchain4j.dev/integrations/embedding-stores
Hands-On Examples
LangChain4j Examples Repository: https://github.com/langchain4j/langchain4j-examples
Contains working examples for all major features
Spring Boot integration examples
RAG implementation examples
Agent and tool usage examples
Framework Integrations
Quarkus Integration: https://docs.langchain4j.dev/integrations/frameworks/quarkus
Spring Boot Integration: https://docs.langchain4j.dev/tutorials/spring-boot-integration
Community & Support
GitHub Discussions: https://github.com/langchain4j/langchain4j/discussions
Discord Community: Join the LangChain4j Discord for real-time help
Advanced Topics
Structured Outputs: https://docs.langchain4j.dev/tutorials/structured-outputs
Guardrails: https://docs.langchain4j.dev/tutorials/guardrails
Model Parameters: https://docs.langchain4j.dev/tutorials/model-parameters
Pro Tip: Clone the examples repository and run the examples locally. Hands-on experimentation is the fastest way to internalize these concepts for interviews and real-world projects.