@runanywhere/llamacpp 0.17.4 → 0.17.7

package/android/CMakeLists.txt

@@ -18,17 +18,17 @@ set(JNILIB_DIR ${CMAKE_SOURCE_DIR}/src/main/jniLibs/${ANDROID_ABI})
 # RABackendLlamaCPP - Llama LLM backend (REQUIRED)
 # Downloaded via Gradle downloadNativeLibs task
 # =============================================================================
-if(NOT EXISTS "${JNILIB_DIR}/librunanywhere_llamacpp.so")
-    message(FATAL_ERROR "[RunAnywhereLlama] RABackendLlamaCPP not found at ${JNILIB_DIR}/librunanywhere_llamacpp.so\n"
+if(NOT EXISTS "${JNILIB_DIR}/librac_backend_llamacpp.so")
+    message(FATAL_ERROR "[RunAnywhereLlama] RABackendLlamaCPP not found at ${JNILIB_DIR}/librac_backend_llamacpp.so\n"
                         "Run: ./gradlew :runanywhere_llamacpp:downloadNativeLibs")
 endif()
 
-add_library(runanywhere_llamacpp SHARED IMPORTED)
-set_target_properties(runanywhere_llamacpp PROPERTIES
-    IMPORTED_LOCATION "${JNILIB_DIR}/librunanywhere_llamacpp.so"
+add_library(rac_backend_llamacpp SHARED IMPORTED)
+set_target_properties(rac_backend_llamacpp PROPERTIES
+    IMPORTED_LOCATION "${JNILIB_DIR}/librac_backend_llamacpp.so"
     IMPORTED_NO_SONAME TRUE
 )
-message(STATUS "[RunAnywhereLlama] Found RABackendLlamaCPP at ${JNILIB_DIR}/librunanywhere_llamacpp.so")
+message(STATUS "[RunAnywhereLlama] Found RABackendLlamaCPP at ${JNILIB_DIR}/librac_backend_llamacpp.so")
 
 # =============================================================================
 # Source files - Llama bridges
@@ -121,7 +121,7 @@ target_link_libraries(
     ${LOG_LIB}
     android
     rac_commons
-    runanywhere_llamacpp
+    rac_backend_llamacpp
 )
 
 # HAS_LLAMACPP and HAS_RACOMMONS are always defined since backends are required

package/ios/Frameworks/RABackendLLAMACPP.xcframework/Info.plist

@@ -8,32 +8,32 @@
 			<key>BinaryPath</key>
 			<string>RABackendLLAMACPP.framework/RABackendLLAMACPP</string>
 			<key>LibraryIdentifier</key>
-			<string>ios-arm64</string>
+			<string>ios-arm64_x86_64-simulator</string>
 			<key>LibraryPath</key>
 			<string>RABackendLLAMACPP.framework</string>
 			<key>SupportedArchitectures</key>
 			<array>
 				<string>arm64</string>
+				<string>x86_64</string>
 			</array>
 			<key>SupportedPlatform</key>
 			<string>ios</string>
+			<key>SupportedPlatformVariant</key>
+			<string>simulator</string>
 		</dict>
 		<dict>
 			<key>BinaryPath</key>
 			<string>RABackendLLAMACPP.framework/RABackendLLAMACPP</string>
 			<key>LibraryIdentifier</key>
-			<string>ios-arm64_x86_64-simulator</string>
+			<string>ios-arm64</string>
 			<key>LibraryPath</key>
 			<string>RABackendLLAMACPP.framework</string>
 			<key>SupportedArchitectures</key>
 			<array>
 				<string>arm64</string>
-				<string>x86_64</string>
 			</array>
 			<key>SupportedPlatform</key>
 			<string>ios</string>
-			<key>SupportedPlatformVariant</key>
-			<string>simulator</string>
 		</dict>
 	</array>
 	<key>CFBundlePackageType</key>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@runanywhere/llamacpp",
-  "version": "0.17.4",
+  "version": "0.17.7",
   "description": "LlamaCpp backend for RunAnywhere React Native SDK - GGUF model support for on-device LLM",
   "main": "src/index.ts",
   "types": "src/index.ts",
@@ -50,8 +50,8 @@
     "react-native-nitro-modules": ">=0.31.3"
   },
   "devDependencies": {
-    "nitrogen": "^0.31.10",
-    "react-native-nitro-modules": "^0.31.10",
+    "nitrogen": "^0.33.2",
+    "react-native-nitro-modules": "0.33.2",
     "typescript": "~5.9.2"
   },
   "create-react-native-library": {

package/android/src/main/include/rac/backends/rac_llm_llamacpp.h

@@ -1,218 +0,0 @@
-/**
- * @file rac_llm_llamacpp.h
- * @brief RunAnywhere Core - LlamaCPP Backend RAC API
- *
- * Direct RAC API export from runanywhere-core's LlamaCPP backend.
- * This header defines the public C API for LLM inference using llama.cpp.
- *
- * Mirrors Swift's LlamaCPPService implementation pattern.
- */
-
-#ifndef RAC_LLM_LLAMACPP_H
-#define RAC_LLM_LLAMACPP_H
-
-#include "rac/core/rac_error.h"
-#include "rac/core/rac_types.h"
-#include "rac/features/llm/rac_llm.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// =============================================================================
-// EXPORT MACRO
-// =============================================================================
-
-#if defined(RAC_LLAMACPP_BUILDING)
-#if defined(_WIN32)
-#define RAC_LLAMACPP_API __declspec(dllexport)
-#elif defined(__GNUC__) || defined(__clang__)
-#define RAC_LLAMACPP_API __attribute__((visibility("default")))
-#else
-#define RAC_LLAMACPP_API
-#endif
-#else
-#define RAC_LLAMACPP_API
-#endif
-
-// =============================================================================
-// CONFIGURATION - Mirrors Swift's LlamaCPPGenerationConfig
-// =============================================================================
-
-/**
- * LlamaCPP-specific configuration.
- *
- * Mirrors Swift's LlamaCPPGenerationConfig.
- */
-typedef struct rac_llm_llamacpp_config {
-    /** Context size (0 = auto-detect from model) */
-    int32_t context_size;
-
-    /** Number of threads (0 = auto-detect) */
-    int32_t num_threads;
-
-    /** Number of layers to offload to GPU (Metal on iOS/macOS) */
-    int32_t gpu_layers;
-
-    /** Batch size for prompt processing */
-    int32_t batch_size;
-} rac_llm_llamacpp_config_t;
-
-/**
- * Default LlamaCPP configuration.
- */
-static const rac_llm_llamacpp_config_t RAC_LLM_LLAMACPP_CONFIG_DEFAULT = {
-    .context_size = 0,  // Auto-detect
-    .num_threads = 0,   // Auto-detect
-    .gpu_layers = -1,   // All layers on GPU
-    .batch_size = 512};
-
-// =============================================================================
-// LLAMACPP-SPECIFIC API
-// =============================================================================
-
-/**
- * Creates a LlamaCPP LLM service.
- *
- * Mirrors Swift's LlamaCPPService.initialize(modelPath:)
- *
- * @param model_path Path to the GGUF model file
- * @param config LlamaCPP-specific configuration (can be NULL for defaults)
- * @param out_handle Output: Handle to the created service
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_llm_llamacpp_create(const char* model_path,
-                                                      const rac_llm_llamacpp_config_t* config,
-                                                      rac_handle_t* out_handle);
-
-/**
- * Loads a GGUF model into an existing service.
- *
- * Mirrors Swift's LlamaCPPService.loadModel(path:config:)
- *
- * @param handle Service handle
- * @param model_path Path to the GGUF model file
- * @param config LlamaCPP configuration (can be NULL)
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_llm_llamacpp_load_model(rac_handle_t handle,
-                                                          const char* model_path,
-                                                          const rac_llm_llamacpp_config_t* config);
-
-/**
- * Unloads the current model.
- *
- * Mirrors Swift's LlamaCPPService.unloadModel()
- *
- * @param handle Service handle
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_llm_llamacpp_unload_model(rac_handle_t handle);
-
-/**
- * Checks if a model is loaded.
- *
- * Mirrors Swift's LlamaCPPService.isModelLoaded
- *
- * @param handle Service handle
- * @return RAC_TRUE if model is loaded, RAC_FALSE otherwise
- */
-RAC_LLAMACPP_API rac_bool_t rac_llm_llamacpp_is_model_loaded(rac_handle_t handle);
-
-/**
- * Generates text completion.
- *
- * Mirrors Swift's LlamaCPPService.generate(prompt:config:)
- *
- * @param handle Service handle
- * @param prompt Input prompt text
- * @param options Generation options (can be NULL for defaults)
- * @param out_result Output: Generation result (caller must free text with rac_free)
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_llm_llamacpp_generate(rac_handle_t handle, const char* prompt,
-                                                        const rac_llm_options_t* options,
-                                                        rac_llm_result_t* out_result);
-
-/**
- * Streaming text generation callback.
- *
- * Mirrors Swift's streaming callback pattern.
- *
- * @param token Generated token string
- * @param is_final Whether this is the final token
- * @param user_data User-provided context
- * @return RAC_TRUE to continue, RAC_FALSE to stop
- */
-typedef rac_bool_t (*rac_llm_llamacpp_stream_callback_fn)(const char* token, rac_bool_t is_final,
-                                                          void* user_data);
-
-/**
- * Generates text with streaming callback.
- *
- * Mirrors Swift's LlamaCPPService.generateStream(prompt:config:)
- *
- * @param handle Service handle
- * @param prompt Input prompt text
- * @param options Generation options
- * @param callback Callback for each token
- * @param user_data User context passed to callback
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_llm_llamacpp_generate_stream(
-    rac_handle_t handle, const char* prompt, const rac_llm_options_t* options,
-    rac_llm_llamacpp_stream_callback_fn callback, void* user_data);
-
-/**
- * Cancels ongoing generation.
- *
- * Mirrors Swift's LlamaCPPService.cancel()
- *
- * @param handle Service handle
- */
-RAC_LLAMACPP_API void rac_llm_llamacpp_cancel(rac_handle_t handle);
-
-/**
- * Gets model information as JSON.
- *
- * @param handle Service handle
- * @param out_json Output: JSON string (caller must free with rac_free)
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_llm_llamacpp_get_model_info(rac_handle_t handle, char** out_json);
-
-/**
- * Destroys a LlamaCPP LLM service.
- *
- * @param handle Service handle to destroy
- */
-RAC_LLAMACPP_API void rac_llm_llamacpp_destroy(rac_handle_t handle);
-
-// =============================================================================
-// BACKEND REGISTRATION
-// =============================================================================
-
-/**
- * Registers the LlamaCPP backend with the commons module and service registries.
- *
- * Should be called once during SDK initialization.
- * This registers:
- * - Module: "llamacpp" with TEXT_GENERATION capability
- * - Service provider: LlamaCPP LLM provider
- *
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_backend_llamacpp_register(void);
-
-/**
- * Unregisters the LlamaCPP backend.
- *
- * @return RAC_SUCCESS or error code
- */
-RAC_LLAMACPP_API rac_result_t rac_backend_llamacpp_unregister(void);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* RAC_LLM_LLAMACPP_H */

package/android/src/main/include/rac/backends/rac_stt_onnx.h

@@ -1,99 +0,0 @@
-/**
- * @file rac_stt_onnx.h
- * @brief RunAnywhere Core - ONNX Backend RAC API for STT
- *
- * Direct RAC API export from runanywhere-core's ONNX STT backend.
- * Mirrors Swift's ONNXSTTService implementation pattern.
- */
-
-#ifndef RAC_STT_ONNX_H
-#define RAC_STT_ONNX_H
-
-#include "rac/core/rac_error.h"
-#include "rac/core/rac_types.h"
-#include "rac/features/stt/rac_stt.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// =============================================================================
-// EXPORT MACRO
-// =============================================================================
-
-#if defined(RAC_ONNX_BUILDING)
-#if defined(_WIN32)
-#define RAC_ONNX_API __declspec(dllexport)
-#elif defined(__GNUC__) || defined(__clang__)
-#define RAC_ONNX_API __attribute__((visibility("default")))
-#else
-#define RAC_ONNX_API
-#endif
-#else
-#define RAC_ONNX_API
-#endif
-
-// =============================================================================
-// CONFIGURATION
-// =============================================================================
-
-/**
- * ONNX STT model types.
- */
-typedef enum rac_stt_onnx_model_type {
-    RAC_STT_ONNX_MODEL_WHISPER = 0,
-    RAC_STT_ONNX_MODEL_ZIPFORMER = 1,
-    RAC_STT_ONNX_MODEL_PARAFORMER = 2,
-    RAC_STT_ONNX_MODEL_AUTO = 99
-} rac_stt_onnx_model_type_t;
-
-/**
- * ONNX STT configuration.
- */
-typedef struct rac_stt_onnx_config {
-    rac_stt_onnx_model_type_t model_type;
-    int32_t num_threads;
-    rac_bool_t use_coreml;
-} rac_stt_onnx_config_t;
-
-static const rac_stt_onnx_config_t RAC_STT_ONNX_CONFIG_DEFAULT = {
-    .model_type = RAC_STT_ONNX_MODEL_AUTO, .num_threads = 0, .use_coreml = RAC_TRUE};
-
-// =============================================================================
-// ONNX STT API
-// =============================================================================
-
-RAC_ONNX_API rac_result_t rac_stt_onnx_create(const char* model_path,
-                                              const rac_stt_onnx_config_t* config,
-                                              rac_handle_t* out_handle);
-
-RAC_ONNX_API rac_result_t rac_stt_onnx_transcribe(rac_handle_t handle, const float* audio_samples,
-                                                  size_t num_samples,
-                                                  const rac_stt_options_t* options,
-                                                  rac_stt_result_t* out_result);
-
-RAC_ONNX_API rac_bool_t rac_stt_onnx_supports_streaming(rac_handle_t handle);
-
-RAC_ONNX_API rac_result_t rac_stt_onnx_create_stream(rac_handle_t handle, rac_handle_t* out_stream);
-
-RAC_ONNX_API rac_result_t rac_stt_onnx_feed_audio(rac_handle_t handle, rac_handle_t stream,
-                                                  const float* audio_samples, size_t num_samples);
-
-RAC_ONNX_API rac_bool_t rac_stt_onnx_stream_is_ready(rac_handle_t handle, rac_handle_t stream);
-
-RAC_ONNX_API rac_result_t rac_stt_onnx_decode_stream(rac_handle_t handle, rac_handle_t stream,
-                                                     char** out_text);
-
-RAC_ONNX_API void rac_stt_onnx_input_finished(rac_handle_t handle, rac_handle_t stream);
-
-RAC_ONNX_API rac_bool_t rac_stt_onnx_is_endpoint(rac_handle_t handle, rac_handle_t stream);
-
-RAC_ONNX_API void rac_stt_onnx_destroy_stream(rac_handle_t handle, rac_handle_t stream);
-
-RAC_ONNX_API void rac_stt_onnx_destroy(rac_handle_t handle);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* RAC_STT_ONNX_H */

package/android/src/main/include/rac/backends/rac_stt_whispercpp.h

@@ -1,153 +0,0 @@
-/**
- * @file rac_stt_whispercpp.h
- * @brief RunAnywhere Core - WhisperCPP Backend for STT
- *
- * RAC API for WhisperCPP-based speech-to-text.
- * Provides high-quality transcription using whisper.cpp.
- *
- * NOTE: WhisperCPP and LlamaCPP both use GGML, which can cause symbol
- * conflicts if linked together. Use ONNX Whisper for STT when also
- * using LlamaCPP for LLM, or build with symbol prefixing.
- */
-
-#ifndef RAC_STT_WHISPERCPP_H
-#define RAC_STT_WHISPERCPP_H
-
-#include "rac/core/rac_error.h"
-#include "rac/core/rac_types.h"
-#include "rac/features/stt/rac_stt.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// =============================================================================
-// EXPORT MACRO
-// =============================================================================
-
-#if defined(RAC_WHISPERCPP_BUILDING)
-#if defined(_WIN32)
-#define RAC_WHISPERCPP_API __declspec(dllexport)
-#elif defined(__GNUC__) || defined(__clang__)
-#define RAC_WHISPERCPP_API __attribute__((visibility("default")))
-#else
-#define RAC_WHISPERCPP_API
-#endif
-#else
-#define RAC_WHISPERCPP_API
-#endif
-
-// =============================================================================
-// CONFIGURATION
-// =============================================================================
-
-/**
- * WhisperCPP-specific configuration.
- */
-typedef struct rac_stt_whispercpp_config {
-    /** Number of threads (0 = auto) */
-    int32_t num_threads;
-
-    /** Enable GPU acceleration (Metal on Apple) */
-    rac_bool_t use_gpu;
-
-    /** Enable CoreML acceleration (Apple only) */
-    rac_bool_t use_coreml;
-
-    /** Language code for transcription (NULL = auto-detect) */
-    const char* language;
-
-    /** Translate to English (when source is non-English) */
-    rac_bool_t translate;
-} rac_stt_whispercpp_config_t;
-
-/**
- * Default WhisperCPP configuration.
- */
-static const rac_stt_whispercpp_config_t RAC_STT_WHISPERCPP_CONFIG_DEFAULT = {
-    .num_threads = 0,
-    .use_gpu = RAC_TRUE,
-    .use_coreml = RAC_TRUE,
-    .language = NULL,
-    .translate = RAC_FALSE};
-
-// =============================================================================
-// WHISPERCPP STT API
-// =============================================================================
-
-/**
- * Creates a WhisperCPP STT service.
- *
- * @param model_path Path to the Whisper GGML model file (.bin)
- * @param config WhisperCPP-specific configuration (can be NULL for defaults)
- * @param out_handle Output: Handle to the created service
- * @return RAC_SUCCESS or error code
- */
-RAC_WHISPERCPP_API rac_result_t rac_stt_whispercpp_create(const char* model_path,
-                                                          const rac_stt_whispercpp_config_t* config,
-                                                          rac_handle_t* out_handle);
-
-/**
- * Transcribes audio data.
- *
- * @param handle Service handle
- * @param audio_samples Float32 PCM samples (16kHz mono)
- * @param num_samples Number of samples
- * @param options STT options (can be NULL for defaults)
- * @param out_result Output: Transcription result
- * @return RAC_SUCCESS or error code
- */
-RAC_WHISPERCPP_API rac_result_t rac_stt_whispercpp_transcribe(rac_handle_t handle,
-                                                              const float* audio_samples,
-                                                              size_t num_samples,
-                                                              const rac_stt_options_t* options,
-                                                              rac_stt_result_t* out_result);
-
-/**
- * Gets detected language after transcription.
- *
- * @param handle Service handle
- * @param out_language Output: Language code (caller must free)
- * @return RAC_SUCCESS or error code
- */
-RAC_WHISPERCPP_API rac_result_t rac_stt_whispercpp_get_language(rac_handle_t handle,
-                                                                char** out_language);
-
-/**
- * Checks if model is loaded and ready.
- *
- * @param handle Service handle
- * @return RAC_TRUE if ready
- */
-RAC_WHISPERCPP_API rac_bool_t rac_stt_whispercpp_is_ready(rac_handle_t handle);
-
-/**
- * Destroys a WhisperCPP STT service.
- *
- * @param handle Service handle to destroy
- */
-RAC_WHISPERCPP_API void rac_stt_whispercpp_destroy(rac_handle_t handle);
-
-// =============================================================================
-// BACKEND REGISTRATION
-// =============================================================================
-
-/**
- * Registers the WhisperCPP backend with the commons module and service registries.
- *
- * @return RAC_SUCCESS or error code
- */
-RAC_WHISPERCPP_API rac_result_t rac_backend_whispercpp_register(void);
-
-/**
- * Unregisters the WhisperCPP backend.
- *
- * @return RAC_SUCCESS or error code
- */
-RAC_WHISPERCPP_API rac_result_t rac_backend_whispercpp_unregister(void);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* RAC_STT_WHISPERCPP_H */

package/android/src/main/include/rac/backends/rac_tts_onnx.h

@@ -1,71 +0,0 @@
-/**
- * @file rac_tts_onnx.h
- * @brief RunAnywhere Core - ONNX Backend RAC API for TTS
- *
- * Direct RAC API export from runanywhere-core's ONNX TTS backend.
- */
-
-#ifndef RAC_TTS_ONNX_H
-#define RAC_TTS_ONNX_H
-
-#include "rac/core/rac_error.h"
-#include "rac/core/rac_types.h"
-#include "rac/features/tts/rac_tts.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// =============================================================================
-// EXPORT MACRO
-// =============================================================================
-
-#if defined(RAC_ONNX_BUILDING)
-#if defined(_WIN32)
-#define RAC_ONNX_API __declspec(dllexport)
-#elif defined(__GNUC__) || defined(__clang__)
-#define RAC_ONNX_API __attribute__((visibility("default")))
-#else
-#define RAC_ONNX_API
-#endif
-#else
-#define RAC_ONNX_API
-#endif
-
-// =============================================================================
-// CONFIGURATION
-// =============================================================================
-
-typedef struct rac_tts_onnx_config {
-    int32_t num_threads;
-    rac_bool_t use_coreml;
-    int32_t sample_rate;
-} rac_tts_onnx_config_t;
-
-static const rac_tts_onnx_config_t RAC_TTS_ONNX_CONFIG_DEFAULT = {
-    .num_threads = 0, .use_coreml = RAC_TRUE, .sample_rate = 22050};
-
-// =============================================================================
-// ONNX TTS API
-// =============================================================================
-
-RAC_ONNX_API rac_result_t rac_tts_onnx_create(const char* model_path,
-                                              const rac_tts_onnx_config_t* config,
-                                              rac_handle_t* out_handle);
-
-RAC_ONNX_API rac_result_t rac_tts_onnx_synthesize(rac_handle_t handle, const char* text,
-                                                  const rac_tts_options_t* options,
-                                                  rac_tts_result_t* out_result);
-
-RAC_ONNX_API rac_result_t rac_tts_onnx_get_voices(rac_handle_t handle, char*** out_voices,
-                                                  size_t* out_count);
-
-RAC_ONNX_API void rac_tts_onnx_stop(rac_handle_t handle);
-
-RAC_ONNX_API void rac_tts_onnx_destroy(rac_handle_t handle);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* RAC_TTS_ONNX_H */