Only llama.rn's own native code lives here. llama.cpp, codec.cpp and the other dependencies are vendored under vendor/ in their upstream layout (see vendor/README.md), and the builds compile them from there.

• rn-llama.*: context wrapper and lifecycle
• rn-completion.*: legacy completion flow
• rn-slot.*, rn-slot-manager.*: parallel decoding/queueing
• rn-mtmd.hpp: multimodal (vision/audio) helpers
• rn-tts.*: TTS/vocoder integration
• rn-common.hpp: shared helpers (tokenization, rerank formatting, etc.)
• anyascii.*: ASCII transliteration used by TTS
• jsi/: the JSI bindings that expose the above to JavaScript. Platform glue (ios/RNLlama.mm, android/src/main/RNLlamaJSI.cpp) installs them on app startup.

The source and include lists every CMake build uses are in cmake/rnllama-sources.cmake; the CocoaPods equivalent is in llama-rn.podspec.

To change llama.cpp behavior, edit the vendored file and regenerate its patch with scripts/update-patch.sh (see vendor/README.md).