Building a Knowledge Assistant with Spring AI
Build a Spring AI knowledge assistant with document ingestion, vector search, team filtering, source evidence and focused tests for missing or outdated content.
Administrator
Lokesh Gupta
Website: https://howtodoinjava.com
Github: https://github.com/lokeshgupta1981/
Twitter: @HowToDoInJAVA
My Articles:
Getting Consistent LLM Responses with Spring AI
Configure Spring AI sampling and structured output, validate classifications, and measure repeatability separately from correctness with focused Java tests.
Connect Existing REST APIs to LLM using Spring AI
Connect Spring AI tools to an existing REST API, pass trusted customer context, handle lookup failures and test the adapter without a language model.
Spring AI Error Handling with Retries, Timeouts and Fallbacks
Handle Spring AI request failures, timeouts, invalid structured output and interrupted streams with configuration, code examples and focused tests.
Testing LLM Responses with Spring AI : LLM-as-a-judge
During unit testing of an e-commerce application, we can easily test a method that calculates an order count by comparing the test result with the expected number. But what if an LLM explains that count? “Three orders went to Pune” and “Pune received three orders” are both correct, even though …
Create a Custom Spring Boot Starter
Build a custom Spring Boot 4 starter with conditional auto-configuration, configuration properties, a consuming app, and tests for defaults and bean overrides.
How to Extract Data From Invoices With AI in Python
Learn how to extract data from invoices with AI in Python using a local Ollama model, Pydantic, OCR and checks, with accuracy measured on 24 invoices.
How to Screen Resumes Against a Job Description With AI
Screen resumes with AI in Python. Redact names, check each requirement with a local Ollama model, verify every quote, and leave the decision to a human.
How to Find Root Cause of Errors in Logs With AI
Find root cause of errors in logs with AI. Group logs with Drain3, mask secrets, ask a local Ollama model for ranked causes, and check every cited line.
Multi-Agent AI in Java with LangChain4j Example
In multi-agent AI system, each agent has its own instructions and a specific part of the task to complete. The agents can use the same language model.
Spring AI vs LangChain4j vs Embabel vs AgentScope Java
Compare Spring AI, LangChain4j, Embabel, and AgentScope Java with the same ticket-lookup example. Choose a starting framework for Spring Boot, plain Java, or agents that need planning.
How to Reduce Claude Code Costs Without Losing Quality
Reduce Claude Code costs with measured changes to CLAUDE.md, prompts, subagents and sessions, tested on a Spring Boot project with a passing test suite.
How to Debug a Java Memory Leak Step by Step
Debug a Java memory leak step by step with a runnable example, jcmd commands, heap dumps, Eclipse MAT screenshots, and measured results before and after the fix.
Top 40 DSA Problems to Practice for Coding Interviews
Practice 40 DSA problems by pattern, with direct problem links, approach hints, edge-case traces and a flexible study plan.
Spring AI Interview Questions and Answers
Spring AI interview questions with answers for Spring AI 2.0: ChatClient, advisors, chat memory, tool calling, MCP, RAG, structured output and testing.
Spring AI 1.x to 2.0 Migration Guide
Spring AI 2.0 migration guide: upgrade a Spring AI 1.x app to 2.0.1 step by step, with the build errors you will see and the fix for each one.
Spring AI with Docker Model Runner: Local LLM Example
Spring AI Docker Model Runner example: run LLMs locally with the OpenAI starter and a base URL, then keep a second model in Ollama in the same app.
Spring AI with Amazon Bedrock: Converse API Example
Spring AI Bedrock Converse in Spring Boot: the starter, AWS credentials, model access, a local model beside it, and switching model families by id.
Spring AI with DeepSeek: Setup and Thinking Mode
Spring AI DeepSeek setup in Spring Boot: the starter, the API key, two ChatClients for DeepSeek and a local model, and reasoning mode in Spring AI 2.0.
Spring AI with Google Gemini: Setup and Thinking Budget
Spring AI Gemini setup in Spring Boot: the starter, the API key, two ChatClients for Gemini and a local model, and the thinking budget in Spring AI 2.0.
Spring AI with Claude: Setup and Extended Thinking
Spring AI Anthropic Claude setup in Spring Boot: the starter, the API key, two ChatClients for Claude and a local model, and extended thinking.
Spring AI Moderation with Ollama and OpenAI
Spring AI moderation with Llama Guard on Ollama and OpenAI’s ModerationModel: check customer messages in an advisor and block the unsafe ones.
Spring AI Observability: Track Tokens, Latency and Cost
Spring AI observability with Micrometer: read token and latency metrics in Actuator, trace one RAG request in Zipkin, and turn tokens into cost.
Testing Spring AI: Unit Tests, Testcontainers and Evaluators
Spring AI testing at three levels: unit tests with a fake ChatModel, integration tests with Ollama in Testcontainers, and evaluators as judges.
Building AI Agents with Spring AI: 5 Workflow Patterns
Build Spring AI agents with five workflow patterns: chain, routing, parallelization, orchestrator and workers, and evaluator and optimizer, with real runs.
Spring AI MCP: Server and Client Example
Build a Spring AI MCP server with @McpTool, @McpResource and @McpPrompt, test it with MCP Inspector, and call its tools from a Spring AI MCP client.
Spring AI Advisors Explained: Built-in and Custom Advisors
Spring AI advisors explained: how the advisor chain works, the built-in advisors, why their order matters, and how to write your own call advisor.
Spring AI RAG Advisors: QuestionAnswer and Modular RAG
Spring AI RAG with Ollama: add QuestionAnswerAdvisor and RetrievalAugmentationAdvisor, compare them on real questions and fix wrong or missing answers.
Spring AI Vector Stores Compared: pgvector, Redis and More
Spring AI vector store comparison: the same app on SimpleVectorStore, Chroma, pgvector, Redis and Qdrant, with real timings, setup, traps and a rule.
Spring AI Redis Vector Store Example with Redis Stack
Spring AI Redis vector store example: run Redis Stack, configure RedisVectorStore, declare metadata fields for filters, and fix the common Redis errors.
Spring AI Multimodality: Send Images to a Chat Model
Spring AI multimodality with Ollama: send a parcel photo and a question to a vision model, accept an upload, and map the answer to a Java record.
Spring AI Chat Memory: Keep Conversation History
Spring AI chat memory with Ollama: add MessageChatMemoryAdvisor, keep one conversation per customer, save it in HSQLDB with JDBC and pick a window size.
Spring AI Streaming Responses with Server-Sent Events
Spring AI streaming with Server-Sent Events in Spring MVC: return a Flux, test it with curl, read it in the browser with EventSource and keep the spaces.
Spring AI ChatClient API: Fluent Chat Examples
Spring AI ChatClient examples: default system prompt, template parameters, Java records with entity(), token usage, per-request options and two clients.
Prompt Injection in LLM Apps: Risks and Defenses
Prompt injection in LLM apps explained: direct and indirect attacks run against a Python assistant, five defenses, and the attack counts after each.
Fine-Tuning vs RAG vs Prompt Engineering for LLM Apps
Fine tuning vs RAG vs prompt engineering for LLM apps: what each changes, cost and data needed, a Java prompt example, and a flow to pick one.
Run LLMs Locally: Ollama, LM Studio and Docker Model Runner
Run LLMs locally from Java with Ollama, Docker Model Runner or LM Studio: compare the three, then call any of them with one plain Java HttpClient class.
RAG Explained: Retrieval-Augmented Generation in Java
Retrieval augmented generation (RAG) explained for Java developers: index, retrieve, generate, the key terms, a runnable offline example and the trap.
Spring Boot and Vaadin CRUD Example
Build a Spring Boot and Vaadin CRUD app with a Grid, a validated form and Spring Data JPA. Complete code for every file, run steps and real screenshots.
TypeScript 7 Native Go Compiler: tsc, tsgo and Migration
TypeScript 7 ports the compiler to Go, and tsc is now a native binary. We measure it against TypeScript 6 on 1,001 modules and migrate a tsconfig step by step.
MCP Tool Poisoning: Python Demo and Practical Defenses
MCP tool poisoning hides instructions in a tool description. See it demonstrated in Python, with hash pinning, scanning and sandboxing as defenses.
Free-Threaded Python (No-GIL): Writing Thread-Safe Code
Free-threaded Python (python3.14t) runs threads in parallel without the GIL. We install the build, reproduce race conditions and write thread-safe code with locks and queues.
Python JIT Compiler: Enable, Benchmark and When to Use
CPython’s experimental JIT compiler (PEP 744) explained: how to enable it, a real from-source benchmark, and when it helps today.
Python Lazy Imports: The lazy Keyword and Alternatives
Python 3.15 adds the lazy keyword (PEP 810) for imports that load on first use. This guide measures start-up time with -X importtime, shows a CLI tool that starts in 57 ms instead of 141 ms, and compares the older alternatives.
Context Bloat in AI Agents: Causes, Cost and Fixes
Context bloat is the growth of an LLM request with tokens the model does not need. Measure a real agent request and cut it with six techniques.
Python UTF-8 Default Encoding: What Changes in Python 3.15
Python 3.15 makes UTF-8 the default encoding of open() on every platform. What breaks with cp1252 files on Windows, and the six steps to migrate safely.
Stateless MCP: Changes in the 2026-07-28 Specification
Stateless MCP means every request carries what the server needs. What revision 2026-07-28 removed and added, shown with a stateful and a stateless Python server behind a two-instance proxy.
Upgrade Python Version: A Safe Migration Guide
Upgrade the Python version of an existing project in seven steps with pip and venv: fix the deprecation warnings on the old version, check the wheels, test on both versions, then switch. Shown on a real project moving from 3.12 to 3.15.
SKILL.md Guide: Write and Test Skills for Coding Agents
A SKILL.md file holds on-demand instructions for a coding agent. This guide covers the format, the description that triggers it, a validator and two tests.
Dropwizard Could Not Resolve Type Id ‘http’ Error: Fix
Fix the Dropwizard error Failed to parse configuration, could not resolve type id ‘http’ as a subtype of ConnectorFactory: add the Shade services merge.