KotlinLLM tackles a common pain for Kotlin/JVM teams: turning messy, semi‑structured data from third‑party APIs into reliable, typed values without writing boilerplate parsers or constantly updating hand‑crafted test doubles. Developers often spend time writing fragile parsing logic, keeping mocks in sync with evolving interfaces, and paying repeated LLM inference costs for every call site. The result is slower development, higher maintenance overhead, and unpredictable latency when AI is invoked at runtime.
The KotlinLLM plugin introduces Smart macros – simple Kotlin function calls like asLlm(input, hint) or mockLlm() – whose bodies are generated by an LLM the first time a new scenario is encountered. When the project runs under the KotlinLLM configuration, the plugin scans for these calls, creates or updates bootstrap/provider/parser/mock files, launches the program under JDI, and sets breakpoints on generated regenerate hooks. If a scenario has not been seen before, execution hits a hook, the plugin captures runtime values and types, asks the LLM for a code update, compiles it, redefines the loaded class, and retries the original call. Once the logic is generated, subsequent runs execute the produced Kotlin directly – no further LLM request, no added latency, and virtually zero extra cost.
Evaluation on an adapted Spring Petclinic project showed 24 out of 24 scenarios covered, a 100% hot‑reload success rate, and only about 1% runtime overhead from compilation and redefinition. A synthetic GitHub Beginner Issue Radar achieved roughly 0.89 recall on beginner‑label classification across 20 repositories and 30k+ issues.
To use KotlinLLM you need IntelliJ IDEA 2025.2.x, JDK 21, and an OpenAI API key stored in a .kotlinllm file via the plugin settings. The plugin is released under Apache 2.0 and is marked as a research prototype, but the code it generates is plain Kotlin that can be committed, reviewed, tested, and shipped without any model dependency. This makes it suitable for R&D groups, platform teams at mid‑size to large Kotlin/JVM shops, and startups comfortable with experimental tooling, especially in fintech, developer tooling, e‑commerce, logistics, or any domain that must normalize unstable API payloads or evolve test doubles.
#AI #Product #Kotlin #LLM #DeveloperTools #JavaJVM