* Collect device info on standard JVM. */

* Configure HTTP service with base URL and API key

* Configure the sampler chain with the given generation parameters. * Only rebuilds the sampler when parameters actually change, avoiding * unnecess

* Convert LogLevel to Sentry severity level

* Convert VLMImage discriminated union to native parameters * @internal

* Convert Int16 PCM audio to Float32 normalized to [-1.0, 1.0]. * SDKs may send Int16 audio but Sherpa-ONNX expects Float32. */

* Convert Int16 PCM audio to Float32 normalized to [-1.0, 1.0]. */

* Create secure storage instance for Android */

* Create a new conversation */

* Create an initial plan for a new task

* Create a ZIP archive containing test files. * Returns path to the created archive, or empty string on failure. */

* Delete data from the specified namespace. * * @param namespace The storage namespace * @param key The key to delete * @param sto

* Delete a file. */

* Destroy the SDK and clean up resources

* Attempt to dismiss overlays/modals that might block interaction

* Dismiss the restart dialog */

Download a LoRA adapter GGUF file via the SDK. */

* Download a file and stream it directly to persistent storage (OPFS or local FS) * without buffering the entire payload in memory. * * Retur

* Download a file from a URL with optional progress callback. * Exposed so ModelManager can use it for on-demand file downloads during load. *

* Emit download progress event via C++. */

* Ensure a conversation is in the in-memory list and persisted. * If not present (by id), adds it at the front so it appears in history. */

* Ensure the LlamaCPP JNI library is loaded. * * Loads librac_backend_llamacpp_jni.so and its dependencies: * - librac_backend_llamacpp

* Ensure the ONNX JNI library is loaded. * * Loads librac_backend_onnx_jni.so and its dependencies: * - librac_backend_onnx.so (ONNX C+

* Ensure the RAG JNI library is loaded. * * Loads librac_backend_rag_jni.so which links against librac_commons.so * (the main SDK's lib

Private helper

* Execute a platform STT request. * * @param request The STT request * @return The STT response */

* Execute a platform TTS request. * * @param request The TTS request * @return The TTS response */

* Execute a task from start to finish * @param task - Natural language task description * @param getDOMState - Function to get current DOM state

* Type text into an input element

* Wait for an element to appear or a fixed time

Check if data exists in actual OPFS storage (NOT memory cache).

Check existence in local storage, OPFS, or in-memory cache.

* Extract an archive (tar.bz2, tar.gz, zip) * @param archivePath Path to the archive * @param destPath Destination directory

* Extract search query for YouTube tasks

Fast hash via LUT – returns 0..1. */

* Fetch model assignments from the backend. * * This fetches models assigned to this device based on device type and platform. * Result

Extract filename from path

Helper: Populate rac_vlm_image_t from JNI parameters

* Find clickable element by visible text

* Find a matching pattern in the page text

* @brief Find matching closing brace for JSON object * * Tracks string boundaries to ignore braces inside strings. * * @param str String to search

* Find a single model file in a directory, searching recursively up to max_depth levels. * Ported from Swift's ExtractionService.findSingleModelFile(

* @brief Find substring in string */

* Get example prompts for a loaded adapter by its file path. * Extracts the filename from the path and looks up prompts. */

* Format prompt using model's built-in chat template via llama_chat_apply_template. * Falls back to manual formatting if template application fails.

Convert TypeScript framework string to C++ enum

Create from a RACommons rac_result_t error code