Point one Spring AI 2.0.2 OpenAiChatModel at Docker Model Runner and at Ollama by changing only base-url and the model name. Same GGUF model, same prompts: decode speed, time to first token, container memory and tool-call success, with transcripts. Measured on a CPU-only shared sandbox with the real docker/model-runner image; GPU and Docker Desktop paths were not measured and the post says so.
Export Excel from Spring Boot without running out of memory: XSSF vs SXSSF measured under a fixed heap (XSSF dies between 50,000 and 100,000 rows in 512 MB), 1,000,000 rows streamed with StreamingResponseBody under 64 MB, sheet rolling at 1,048,576 rows, temp-file cleanup, OpenCSV traps and a file LibreOffice will not open.
A Spring Batch 6 chunk step calls a Spring AI ChatClient for 100,000 generated support tickets. Resilience4j rate limiting and retry, skip-and-record for malformed JSON, a restart after kill -9 in a second JVM that ends byte-identical, and a per-chunk cost ledger, all run against a deterministic simulated model with every number taken from a committed transcript.
Hibernate 7.4 multi-tenancy on Spring Boot 4.1 and PostgreSQL 17, run end to end: @TenantId discriminator vs schema-per-tenant, tenant taken from a signed JWT claim, the search_path leak, one shared pool vs a pool per tenant (measured connection counts), and what happens when no tenant is set.
Three ways to build an LRU cache in Java: LinkedHashMap with accessOrder, HashMap plus a hand-rolled doubly linked list with sentinels, and thread-safe variants (synchronized, ReentrantLock, striping). Includes an O(1) dry-run table, a demo of why ReadWriteLock corrupts the list, a stress test with real assertions, an indicative JMH comparison and Caffeine as the production answer.
Instancio 6.1 fills whole Java object graphs with random data and replays any failure from a seed. Run on JUnit 6.1.3: selectors, Settings, records and sealed types, TypeToken, the extension, and the real seed message from a failing test.
A hands-on Spring Data R2DBC guide on Spring Boot 4.1 and JDK 25: ReactiveCrudRepository, custom queries, DatabaseClient, reactive transactions that roll back, Testcontainers Postgres with StepVerifier, an honest look at having no relations, and when virtual threads with JPA are the better default. Every transcript comes from a real run.
java.lang.classfile has been final since JDK 24 (JEP 484). Parse a class, generate one at runtime and load it as a normal or hidden class, and insert timing around every method with ClassTransform, MethodTransform and CodeTransform, all run on JDK 25. The same three jobs are repeated in ASM 9.11.1 and compared, including the traps in both, the class-file versions each reads, a JDK 24/25/26 API diff and what spring-core 7.0.10 actually ships.
How to generate images with Spring AI's ImageModel and keep them safely: what the real OpenAiImageModel sends (including an options object that silently resets the model), a moderation gate and what Spring AI does not moderate, verified storage in MinIO/S3 with sniffed content types, size and pixel caps and server-chosen keys, and a per-provider cost table dated 2026-10-28. No image provider was called; the model and moderation are scripted fakes and the bucket is a real MinIO.
Build rate limiting into a Spring Boot 4.1 API with Bucket4j 8.21 and Redis: per-user, per-API-key and global token buckets in a servlet filter, 429 with Retry-After and X-RateLimit headers, two instances sharing one bucket, and the failures worth reproducing (limit changes that do nothing, clock skew, Redis down), all with tests that assert it.