Prebuilt llama.cpp AAR for Android — on-device LLM inference, no NDK required. Chat completion, embeddings, GGUF models. Free on Maven Central.