Remove a registered provider.

Populate suggestion chips (general or tool-specific depending on toggle).

* Retrieve a registered provider or throw a descriptive error. * Use in code that requires a specific backend to be registered.

* Save model metadata map to OPFS as a small JSON file. * Used for LRU tracking (lastUsedAt timestamps).

* Save model data to OPFS. * * Supports nested paths: `saveModel('org/model/file.gguf', data)` creates * `models/org/model/` directories and

* Set the base URL for telemetry HTTP requests. * Should be called before SDK initialization if using a custom URL. */

Set the local file storage backend for persistent model storage.

* Store a string value securely * @param key Unique identifier for the value * @param value String value to store * @throws SDKError.Se

* @brief Skip whitespace in string */

Begin capturing audio and monitoring with SDK VAD

* Start recording audio * @param onAudioData Callback for audio data chunks

* Stop audio capture */

* Stop audio playback * iOS Reference: stopPlayback() using AVAudioPlayer */

* Store a string in the specified namespace. * * @param namespace The storage namespace * @param key The key to store under * @par

* Synthesize speech from text * @param text Text to synthesize * @param voiceId Voice ID to use * @param speedRate Speech speed rate (1.0 = n

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

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

* Transcribe audio data * @param audioBase64 Base64 encoded audio data * @param sampleRate Audio sample rate * @param language Language code

Remove a specific loaded adapter by path. */

Unload the currently loaded model for a specific category

* Unload the current STT model

* Unmount a directory (and remove it). * @param mountDir - The directory path (e.g. /mnt-123)

* Unregister the ONNX backend from C++ registry. * @returns Promise<boolean> true if unregistered successfully

Show/hide the empty state based on whether there are messages.

* @brief Validate that a component is ready for use * * Performs defensive checks: * 1. Handle is non-null * 2. Component is in LOADED state * *

* Wait for content script to become available with timeout * Returns true if content script is ready, false otherwise

* Screens with a fully custom top bar (Chat, VLM) call this. * The root Scaffold renders [content] instead of the default TopAppBar. */

* Storage Management Button - iOS StorageManagementButton with icon, title, subtitle */

* Voice Pipeline Setup View * * VoicePipelineSetupView * * A setup view specifically for Voice Assistant which requires 3 models: * - STT (Speech

Always outputs a boolean value

* Analyze a page screenshot for navigation actions

Check if all required models are available

Check if wake word models are available

Build VLMImageInput from JS bridge parameters

Helper: Build a VLM result JSON string matching what Kotlin expects

* @brief Build a WAV header for PCM audio * * @param header Buffer to write header to (must be at least 44 bytes) * @param sample_rate Sample rate

* Shared implementation for byte array checksum calculation. * Platform-specific: Uses java.security.MessageDigest. */

* Shared implementation for byte array checksum calculation. * Platform-specific: Uses java.security.MessageDigest. */

* Shared implementation for file-based checksum calculation. * Platform-specific: Uses java.io.File for file I/O. */

* Shared implementation for file-based checksum calculation. * Platform-specific: Uses java.io.File for file I/O. */

* Calculate RMS (Root Mean Square) for audio level visualization */

* Cancel ongoing text generation

* Capture a screenshot of the current tab

* Send a chat completion request to the LLM

* Check if a LoRA adapter is compatible with the currently loaded model. * * @param loraPath Path to the LoRA adapter file * @return nu

* Check network connectivity and return detailed status. * * @return Pair of (isAvailable, statusMessage) */

* Check if a service is available. * * @param serviceType The service type to check * @return The availability status

* Clean temporary files */

Clean up the VAD resources.

* Cleanup all registered backends in reverse registration order. * Called during SDK shutdown.

Clear all loaded adapters. */

* Clear error state */

* Remove all LoRA adapters. * * @return 0 on success, error code on failure */