* Create directories recursively (like mkdir -p). */

* Called when an error occurs. * * @param serviceType The service type * @param errorCode The error code * @param

* Get the next available speech segment (if any). * Returns null if no complete segments are available. * * After calling processSamples(), c

* Save a string value */

* Emit an LLM generation event. * Maps to rac_analytics_llm_generation_t struct in C++. */

* Emit an LLM model event. * Maps to rac_analytics_llm_model_t struct in C++. */

* Convert a UTF-8 std::string to std::wstring (UTF-16) for Windows wide-char APIs. * Uses MultiByteToWideChar so non-ASCII paths (Chinese, Japanese,

* Remove all listeners.

* Save model data to OPFS from a ReadableStream. * Streams data directly to disk without buffering the entire file in memory. * * @param key

* Start the voice session

* Start capturing audio and emit audio chunks as a Flow * Returns PCM audio data at 16kHz, mono, 16-bit */

* Stop recording and return recorded audio file path

* @brief Case-insensitive string comparison */

* Convert a C++ raw error code to a Kotlin ErrorCode enum. * * @return The corresponding Kotlin ErrorCode enum value */

* Convert to JSON string for C++ interop. */

* Convert to JSON string for C++ interop. */

* Unload the current TTS model

* Update threshold_sq whenever energy_threshold changes. * This pre-computes the squared threshold so the hot-path * comparison can use mean-square

* Log a warning-level message. */

* @brief Write a little-endian uint16_t to a buffer */

* Model Loaded Toast - matches iOS ModelLoadedToast.swift * * Shows a brief notification when a model finishes loading. * - Green checkmark + "Mode

* Model Setup Card * * ModelSetupCard * * A card showing model selection and loading state */

* Apply rule-based action selection. Public so executor can call directly.