speech_recognition.mojom | Explore in Territory

// Copyright 2023 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

module ax.mojom;

import "services/accessibility/public/mojom/assistive_technology_type.mojom";

// Possible types of speech recognition.
enum SpeechRecognitionType {
  kOnDevice,
  kNetwork,
};

// An event that is sent when speech recognition results are available.
struct SpeechRecognitionResultEvent {
  // The recognized phrase.
  string transcript;
  // Whether the result is a final or an interim result.
  bool is_final;
};

// An event that is sent when speech recognition encounters an error.
struct SpeechRecognitionErrorEvent {
  // The human-readable error message.
  string message;
};

// Options used when starting speech recognition.
struct StartOptions {
  // The type of assistive technology requesting speech recognition start.
  // Note: Currently the only supported assistive technology type is Dictation.
  AssistiveTechnologyType type;
  // The locale, in BCP-47 format to use for speech recognition.
  string? locale;
  // Whether interim speech results should be returned.
  bool? interim_results;
};

// Options used when stopping speech recognition.
struct StopOptions {
  // The type of assistive technology requesting speech recognition stop.
  // This must match the AssistiveTechnologyType used when starting speech
  // recognition to work as intended.
  // Note: Currently the only supported assistive technology type is Dictation.
  AssistiveTechnologyType type;
};

// This interface receives speech recognition events. The remote is in the OS
// browser process and the receiver is in the Accessibility Service.
interface SpeechRecognitionEventObserver {
  // Called when speech recognition stops. Note that this differs from the
  // callback of the Stop() method on SpeechRecognition, which gets called when
  // speech recognition is stopped because of an explicit request from the
  // caller. Destroys the remote in the browser process so that no other events
  // will be fired after this gets called.
  OnStop();
  // Called when a speech recognition result is returned from the speech
  // recognition service to the browser process.
  OnResult(SpeechRecognitionResultEvent event);
  // Called when a speech recognition error is returned from the speech
  // recognition service to the browser process. Destroys the remote in the
  // browser process so that no other events will be fired after this gets
  // called.
  OnError(SpeechRecognitionErrorEvent event);
};

// A union that either holds a SpeechRecognitionEventObserver, in the case
// that speech recognition started successfully, or an error message, in the
// case the speech recognition failed to start.
union ObserverOrError {
  // An observer that can receive speech recognition events.
  pending_receiver<SpeechRecognitionEventObserver> observer;
  // An error message if speech recognition failed to start.
  string error;
};

// Information that is returned once speech recognition has started.
struct SpeechRecognitionStartInfo {
  // The type of speech recognition being used.
  SpeechRecognitionType type;
  ObserverOrError observer_or_error;
};

// Provides speech recognition services. Implemented in the main OS browser
// process and called from Accessibility Service Javascript.
interface SpeechRecognition {
  // Starts listening to audio from the user. The callback is invoked when
  // speech recognition has started or failed to start.
  Start(StartOptions options) => (SpeechRecognitionStartInfo info);

  // Stops listening to audio from the user. The callback is invoked when
  // speech recognition has stopped. Note that this differs from the OnStop()
  // method on SpeechRecognitionEventObserver, which gets called when speech
  // recognition is stopped in the background without an explicit request from
  // the caller.
  Stop(StopOptions options) => (string? error);
};
chromium/services/accessibility/public/mojom/speech_recognition.mojom
