@runanywhere/llamacpp 0.18.0 → 0.19.13

package/android/src/main/include/rac/infrastructure/telemetry/rac_telemetry_types.h

@@ -1,234 +0,0 @@
-/**
- * @file rac_telemetry_types.h
- * @brief Telemetry data structures - canonical source of truth
- *
- * All telemetry payloads are defined here. Platform SDKs (Swift, Kotlin, Flutter)
- * use these types directly or create thin wrappers.
- *
- * Mirrors Swift's TelemetryEventPayload.swift structure.
- */
-
-#ifndef RAC_TELEMETRY_TYPES_H
-#define RAC_TELEMETRY_TYPES_H
-
-#include "rac/core/rac_types.h"
-#include "rac/infrastructure/network/rac_environment.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// =============================================================================
-// TELEMETRY EVENT PAYLOAD
-// =============================================================================
-
-/**
- * @brief Complete telemetry event payload
- *
- * Maps to backend SDKTelemetryEvent schema with all fields for:
- * - LLM events (tokens, generation times, etc.)
- * - STT events (audio duration, word count, etc.)
- * - TTS events (character count, audio size, etc.)
- * - VAD events (speech duration)
- * - Model lifecycle events (size, archive type)
- * - SDK lifecycle events (count)
- * - Storage events (freed bytes)
- * - Network events (online status)
- */
-typedef struct rac_telemetry_payload {
-    // Required fields
-    const char* id;          // Unique event ID (UUID)
-    const char* event_type;  // Event type string
-    int64_t timestamp_ms;    // Unix timestamp in milliseconds
-    int64_t created_at_ms;   // When payload was created
-
-    // Event classification
-    const char* modality;  // "llm", "stt", "tts", "model", "system"
-
-    // Device identification
-    const char* device_id;   // Persistent device UUID
-    const char* session_id;  // Optional session ID
-
-    // Model info
-    const char* model_id;
-    const char* model_name;
-    const char* framework;  // "llamacpp", "onnx", "mlx", etc.
-
-    // Device info
-    const char* device;       // Device model (e.g., "iPhone15,2")
-    const char* os_version;   // OS version (e.g., "17.0")
-    const char* platform;     // "ios", "android", "flutter"
-    const char* sdk_version;  // SDK version string
-
-    // Common performance metrics
-    double processing_time_ms;
-    rac_bool_t success;
-    rac_bool_t has_success;  // Whether success field is set
-    const char* error_message;
-    const char* error_code;
-
-    // LLM-specific fields
-    int32_t input_tokens;
-    int32_t output_tokens;
-    int32_t total_tokens;
-    double tokens_per_second;
-    double time_to_first_token_ms;
-    double prompt_eval_time_ms;
-    double generation_time_ms;
-    int32_t context_length;
-    double temperature;
-    int32_t max_tokens;
-
-    // STT-specific fields
-    double audio_duration_ms;
-    double real_time_factor;
-    int32_t word_count;
-    double confidence;
-    const char* language;
-    rac_bool_t is_streaming;
-    rac_bool_t has_is_streaming;
-    int32_t segment_index;
-
-    // TTS-specific fields
-    int32_t character_count;
-    double characters_per_second;
-    int32_t audio_size_bytes;
-    int32_t sample_rate;
-    const char* voice;
-    double output_duration_ms;
-
-    // Model lifecycle fields
-    int64_t model_size_bytes;
-    const char* archive_type;
-
-    // VAD fields
-    double speech_duration_ms;
-
-    // SDK lifecycle fields
-    int32_t count;
-
-    // Storage fields
-    int64_t freed_bytes;
-
-    // Network fields
-    rac_bool_t is_online;
-    rac_bool_t has_is_online;
-} rac_telemetry_payload_t;
-
-/**
- * @brief Default/empty telemetry payload
- */
-RAC_API rac_telemetry_payload_t rac_telemetry_payload_default(void);
-
-/**
- * @brief Free any allocated strings in a telemetry payload
- */
-RAC_API void rac_telemetry_payload_free(rac_telemetry_payload_t* payload);
-
-// =============================================================================
-// TELEMETRY BATCH REQUEST
-// =============================================================================
-
-/**
- * @brief Batch telemetry request for API
- *
- * Supports both V1 and V2 storage paths:
- * - V1 (legacy): modality = NULL → stores in sdk_telemetry_events table
- * - V2 (normalized): modality = "llm"/"stt"/"tts"/"model" → normalized tables
- */
-typedef struct rac_telemetry_batch_request {
-    rac_telemetry_payload_t* events;
-    size_t events_count;
-    const char* device_id;
-    int64_t timestamp_ms;
-    const char* modality;  // NULL for V1, "llm"/"stt"/"tts"/"model" for V2
-} rac_telemetry_batch_request_t;
-
-/**
- * @brief Batch telemetry response from API
- */
-typedef struct rac_telemetry_batch_response {
-    rac_bool_t success;
-    int32_t events_received;
-    int32_t events_stored;
-    int32_t events_skipped;  // Duplicates skipped
-    const char** errors;     // Array of error messages
-    size_t errors_count;
-    const char* storage_version;  // "V1" or "V2"
-} rac_telemetry_batch_response_t;
-
-/**
- * @brief Free batch response
- */
-RAC_API void rac_telemetry_batch_response_free(rac_telemetry_batch_response_t* response);
-
-// =============================================================================
-// DEVICE REGISTRATION TYPES
-// =============================================================================
-
-/**
- * @brief Device information for registration (telemetry-specific)
- *
- * Platform-specific values are passed in from Swift/Kotlin.
- * Matches backend schemas/device.py DeviceInfo schema.
- * Note: Named differently from rac_device_info_t to avoid conflict.
- */
-typedef struct rac_device_registration_info {
-    // Required fields (backend schema)
-    const char* device_id;           // Persistent UUID from Keychain/secure storage
-    const char* device_model;        // "iPhone 16 Pro Max", "Pixel 7", etc.
-    const char* device_name;         // User-assigned device name
-    const char* platform;            // "ios", "android"
-    const char* os_version;          // "17.0", "14"
-    const char* form_factor;         // "phone", "tablet", "laptop", etc.
-    const char* architecture;        // "arm64", "x86_64", etc.
-    const char* chip_name;           // "A18 Pro", "Snapdragon 888", etc.
-    int64_t total_memory;            // Total RAM in bytes
-    int64_t available_memory;        // Available RAM in bytes
-    rac_bool_t has_neural_engine;    // true if device has Neural Engine / NPU
-    int32_t neural_engine_cores;     // Number of Neural Engine cores (0 if none)
-    const char* gpu_family;          // "apple", "adreno", etc.
-    double battery_level;            // 0.0-1.0, negative if unavailable
-    const char* battery_state;       // "charging", "full", "unplugged", NULL if unavailable
-    rac_bool_t is_low_power_mode;    // Low power mode enabled
-    int32_t core_count;              // Total CPU cores
-    int32_t performance_cores;       // Performance (P) cores
-    int32_t efficiency_cores;        // Efficiency (E) cores
-    const char* device_fingerprint;  // Unique device fingerprint (may be same as device_id)
-
-    // Legacy fields (for backward compatibility)
-    const char* device_type;  // "smartphone", "tablet", etc. (deprecated - use form_factor)
-    const char* os_name;      // "iOS", "Android" (deprecated - use platform)
-    int64_t total_disk_bytes;
-    int64_t available_disk_bytes;
-    const char* processor_info;
-    int32_t processor_count;  // Deprecated - use core_count
-    rac_bool_t is_simulator;
-    const char* locale;
-    const char* timezone;
-} rac_device_registration_info_t;
-
-/**
- * @brief Device registration request
- */
-typedef struct rac_device_registration_request {
-    rac_device_registration_info_t device_info;
-    const char* sdk_version;
-    const char* build_token;  // For development mode
-    int64_t last_seen_at_ms;
-} rac_device_registration_request_t;
-
-/**
- * @brief Device registration response
- */
-typedef struct rac_device_registration_response {
-    const char* device_id;
-    const char* status;       // "registered" or "updated"
-    const char* sync_status;  // "synced" or "pending"
-} rac_device_registration_response_t;
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif  // RAC_TELEMETRY_TYPES_H

package/android/src/main/java/com/margelo/nitro/runanywhere/llama/RunAnywhereLlamaPackage.kt

@@ -1,35 +0,0 @@
-package com.margelo.nitro.runanywhere.llama
-
-import com.facebook.react.BaseReactPackage
-import com.facebook.react.bridge.NativeModule
-import com.facebook.react.bridge.ReactApplicationContext
-import com.facebook.react.module.model.ReactModuleInfoProvider
-import com.margelo.nitro.runanywhere.SDKLogger
-
-/**
- * React Native package for RunAnywhere LlamaCPP backend.
- * This class is required for React Native autolinking.
- */
-class RunAnywhereLlamaPackage : BaseReactPackage() {
-    override fun getModule(name: String, reactContext: ReactApplicationContext): NativeModule? {
-        return null
-    }
-
-    override fun getReactModuleInfoProvider(): ReactModuleInfoProvider {
-        return ReactModuleInfoProvider { HashMap() }
-    }
-
-    companion object {
-        private val log = SDKLogger("LLM.LlamaCpp")
-
-        init {
-            // Load the native library which registers the HybridObject factory
-            try {
-                System.loadLibrary("runanywherellama")
-            } catch (e: UnsatisfiedLinkError) {
-                // Native library may already be loaded or bundled differently
-                log.error("Failed to load runanywherellama: ${e.message}")
-            }
-        }
-    }
-}

package/cpp/bridges/LLMBridge.cpp

@@ -1,209 +0,0 @@
-/**
- * @file LLMBridge.cpp
- * @brief LLM capability bridge implementation
- *
- * NOTE: RACommons and LlamaCPP backend are REQUIRED and always linked via the build system.
- */
-
-#include "LLMBridge.hpp"
-#include <stdexcept>
-
-namespace runanywhere {
-namespace bridges {
-
-LLMBridge& LLMBridge::shared() {
-    static LLMBridge instance;
-    return instance;
-}
-
-LLMBridge::LLMBridge() = default;
-
-LLMBridge::~LLMBridge() {
-    destroy();
-}
-
-bool LLMBridge::isLoaded() const {
-    if (handle_) {
-        return rac_llm_component_is_loaded(handle_) == RAC_TRUE;
-    }
-    return false;
-}
-
-std::string LLMBridge::currentModelId() const {
-    return loadedModelId_;
-}
-
-rac_result_t LLMBridge::loadModel(const std::string& modelPath,
-                                  const std::string& modelId,
-                                  const std::string& modelName) {
-    // Create component if needed
-    if (!handle_) {
-        rac_result_t result = rac_llm_component_create(&handle_);
-        if (result != RAC_SUCCESS) {
-            throw std::runtime_error("LLMBridge: Failed to create LLM component. Error: " + std::to_string(result));
-        }
-    }
-
-    // Use modelPath as modelId if not provided
-    std::string effectiveModelId = modelId.empty() ? modelPath : modelId;
-    std::string effectiveModelName = modelName.empty() ? effectiveModelId : modelName;
-
-    // Unload existing model if different
-    if (isLoaded() && loadedModelId_ != effectiveModelId) {
-        rac_llm_component_unload(handle_);
-    }
-
-    // Load new model with correct 4-arg signature
-    // rac_llm_component_load_model(handle, model_path, model_id, model_name)
-    rac_result_t result = rac_llm_component_load_model(
-        handle_,
-        modelPath.c_str(),
-        effectiveModelId.c_str(),
-        effectiveModelName.c_str()
-    );
-    if (result == RAC_SUCCESS) {
-        loadedModelId_ = effectiveModelId;
-    } else {
-        throw std::runtime_error("LLMBridge: Failed to load LLM model '" + effectiveModelId + "'. Error: " + std::to_string(result));
-    }
-    return result;
-}
-
-rac_result_t LLMBridge::unload() {
-    if (handle_) {
-        rac_result_t result = rac_llm_component_unload(handle_);
-        if (result == RAC_SUCCESS) {
-            loadedModelId_.clear();
-        } else {
-            throw std::runtime_error("LLMBridge: Failed to unload LLM model. Error: " + std::to_string(result));
-        }
-        return result;
-    }
-    loadedModelId_.clear();
-    return RAC_SUCCESS;
-}
-
-void LLMBridge::cleanup() {
-    if (handle_) {
-        rac_llm_component_cleanup(handle_);
-    }
-    loadedModelId_.clear();
-}
-
-void LLMBridge::cancel() {
-    cancellationRequested_ = true;
-    if (handle_) {
-        rac_llm_component_cancel(handle_);
-    }
-}
-
-void LLMBridge::destroy() {
-    if (handle_) {
-        rac_llm_component_destroy(handle_);
-        handle_ = nullptr;
-    }
-    loadedModelId_.clear();
-}
-
-LLMResult LLMBridge::generate(const std::string& prompt, const LLMOptions& options) {
-    LLMResult result;
-    cancellationRequested_ = false;
-
-    if (!handle_ || !isLoaded()) {
-        throw std::runtime_error("LLMBridge: LLM model not loaded. Call loadModel() first.");
-    }
-
-    rac_llm_options_t racOptions = {};
-    racOptions.max_tokens = options.maxTokens;
-    racOptions.temperature = static_cast<float>(options.temperature);
-    racOptions.top_p = static_cast<float>(options.topP);
-    // NOTE: top_k is not available in rac_llm_options_t, only top_p
-
-    rac_llm_result_t racResult = {};
-    rac_result_t status = rac_llm_component_generate(handle_, prompt.c_str(),
-                                                      &racOptions, &racResult);
-
-    if (status == RAC_SUCCESS) {
-        if (racResult.text) {
-            result.text = racResult.text;
-        }
-        result.tokenCount = racResult.completion_tokens;
-        result.durationMs = static_cast<double>(racResult.total_time_ms);
-    } else {
-        throw std::runtime_error("LLMBridge: Text generation failed with error code: " + std::to_string(status));
-    }
-
-    result.cancelled = cancellationRequested_;
-    return result;
-}
-
-void LLMBridge::generateStream(const std::string& prompt, const LLMOptions& options,
-                               const LLMStreamCallbacks& callbacks) {
-    cancellationRequested_ = false;
-
-    if (!handle_ || !isLoaded()) {
-        if (callbacks.onError) {
-            callbacks.onError(-4, "LLM model not loaded. Call loadModel() first.");
-        }
-        return;
-    }
-
-    rac_llm_options_t racOptions = {};
-    racOptions.max_tokens = options.maxTokens;
-    racOptions.temperature = static_cast<float>(options.temperature);
-    racOptions.top_p = static_cast<float>(options.topP);
-    // NOTE: top_k is not available in rac_llm_options_t, only top_p
-
-    // Stream context for callbacks
-    struct StreamContext {
-        const LLMStreamCallbacks* callbacks;
-        bool* cancellationRequested;
-        std::string accumulatedText;
-    };
-
-    StreamContext ctx = { &callbacks, &cancellationRequested_, "" };
-
-    auto tokenCallback = [](const char* token, void* user_data) -> rac_bool_t {
-        auto* ctx = static_cast<StreamContext*>(user_data);
-        if (*ctx->cancellationRequested) {
-            return RAC_FALSE;
-        }
-        if (ctx->callbacks->onToken && token) {
-            ctx->accumulatedText += token;
-            return ctx->callbacks->onToken(token) ? RAC_TRUE : RAC_FALSE;
-        }
-        return RAC_TRUE;
-    };
-
-    auto completeCallback = [](const rac_llm_result_t* result, void* user_data) {
-        auto* ctx = static_cast<StreamContext*>(user_data);
-        if (ctx->callbacks->onComplete) {
-            ctx->callbacks->onComplete(
-                ctx->accumulatedText,
-                result ? result->completion_tokens : 0,
-                result ? static_cast<double>(result->total_time_ms) : 0.0
-            );
-        }
-    };
-
-    auto errorCallback = [](rac_result_t error_code, const char* error_message,
-                           void* user_data) {
-        auto* ctx = static_cast<StreamContext*>(user_data);
-        if (ctx->callbacks->onError) {
-            ctx->callbacks->onError(error_code, error_message ? error_message : "Unknown error");
-        }
-    };
-
-    rac_llm_component_generate_stream(handle_, prompt.c_str(), &racOptions,
-                                      tokenCallback, completeCallback, errorCallback, &ctx);
-}
-
-rac_lifecycle_state_t LLMBridge::getState() const {
-    if (handle_) {
-        return rac_llm_component_get_state(handle_);
-    }
-    return RAC_LIFECYCLE_STATE_IDLE;
-}
-
-} // namespace bridges
-} // namespace runanywhere

package/cpp/bridges/LLMBridge.hpp

@@ -1,109 +0,0 @@
-/**
- * @file LLMBridge.hpp
- * @brief LLM capability bridge for React Native
- *
- * Matches Swift's CppBridge+LLM.swift pattern, providing:
- * - Model lifecycle (load/unload)
- * - Text generation (sync and streaming)
- * - Cancellation support
- *
- * Aligned with rac_llm_component.h and rac_llm_types.h API.
- * RACommons is REQUIRED - no stub implementations.
- */
-
-#pragma once
-
-#include <functional>
-#include <memory>
-#include <string>
-
-// RACommons LLM headers - REQUIRED (flat include paths)
-#include "rac_llm_component.h"
-#include "rac_llm_types.h"
-
-namespace runanywhere {
-namespace bridges {
-
-/**
- * @brief LLM streaming callbacks
- */
-struct LLMStreamCallbacks {
-    std::function<bool(const std::string&)> onToken;
-    std::function<void(const std::string&, int, double)> onComplete;
-    std::function<void(int, const std::string&)> onError;
-};
-
-/**
- * @brief LLM generation options
- */
-struct LLMOptions {
-    int maxTokens = 512;
-    double temperature = 0.7;
-    double topP = 0.9;
-    int topK = 40;
-    std::string systemPrompt;
-    std::string stopSequence;
-};
-
-/**
- * @brief LLM generation result
- */
-struct LLMResult {
-    std::string text;
-    int tokenCount = 0;
-    double durationMs = 0.0;
-    bool cancelled = false;
-};
-
-/**
- * @brief LLM capability bridge singleton
- *
- * Matches CppBridge+LLM.swift API.
- * NOTE: RACommons is REQUIRED. All methods will throw std::runtime_error if
- * the underlying C API calls fail.
- */
-class LLMBridge {
-public:
-    static LLMBridge& shared();
-
-    // Lifecycle
-    bool isLoaded() const;
-    std::string currentModelId() const;
-    /**
-     * Load an LLM model
-     * @param modelPath Path to the model file (.gguf)
-     * @param modelId Model identifier for telemetry (e.g., "smollm2-360m-q8_0")
-     * @param modelName Human-readable model name (e.g., "SmolLM2 360M Q8_0")
-     * @return RAC_SUCCESS or error code
-     */
-    rac_result_t loadModel(const std::string& modelPath,
-                           const std::string& modelId = "",
-                           const std::string& modelName = "");
-    rac_result_t unload();
-    void cleanup();
-    void cancel();
-    void destroy();
-
-    // Generation
-    LLMResult generate(const std::string& prompt, const LLMOptions& options);
-    void generateStream(const std::string& prompt, const LLMOptions& options,
-                       const LLMStreamCallbacks& callbacks);
-
-    // State
-    rac_lifecycle_state_t getState() const;
-
-private:
-    LLMBridge();
-    ~LLMBridge();
-
-    // Disable copy/move
-    LLMBridge(const LLMBridge&) = delete;
-    LLMBridge& operator=(const LLMBridge&) = delete;
-
-    rac_handle_t handle_ = nullptr;
-    std::string loadedModelId_;
-    bool cancellationRequested_ = false;
-};
-
-} // namespace bridges
-} // namespace runanywhere