[voice] Add dialog group and location (#3798)

Signed-off-by: Miguel Álvarez <[email protected]>

bundles/org.openhab.core.voice/src/main/java/org/openhab/core/voice/DialogContext.java

@@ -29,78 +29,10 @@
  * @author Miguel Álvarez - Initial contribution
  */
 @NonNullByDefault
-public class DialogContext {
-    private final @Nullable KSService ks;
-    private final @Nullable String keyword;
-    private final STTService stt;
-    private final TTSService tts;
-    private final @Nullable Voice voice;
-    private final List<HumanLanguageInterpreter> hlis;
-    private final AudioSource source;
-    private final AudioSink sink;
-    private final Locale locale;
-    private final @Nullable String listeningItem;
-    private final @Nullable String listeningMelody;
-
-    public DialogContext(@Nullable KSService ks, @Nullable String keyword, STTService stt, TTSService tts,
-            @Nullable Voice voice, List<HumanLanguageInterpreter> hlis, AudioSource source, AudioSink sink,
-            Locale locale, @Nullable String listeningItem, @Nullable String listeningMelody) {
-        this.ks = ks;
-        this.keyword = keyword;
-        this.stt = stt;
-        this.tts = tts;
-        this.voice = voice;
-        this.hlis = hlis;
-        this.source = source;
-        this.sink = sink;
-        this.locale = locale;
-        this.listeningItem = listeningItem;
-        this.listeningMelody = listeningMelody;
-    }
-
-    public @Nullable KSService ks() {
-        return ks;
-    }
-
-    public @Nullable String keyword() {
-        return keyword;
-    }
-
-    public STTService stt() {
-        return stt;
-    }
-
-    public TTSService tts() {
-        return tts;
-    }
-
-    public @Nullable Voice voice() {
-        return voice;
-    }
-
-    public List<HumanLanguageInterpreter> hlis() {
-        return hlis;
-    }
-
-    public AudioSource source() {
-        return source;
-    }
-
-    public AudioSink sink() {
-        return sink;
-    }
-
-    public Locale locale() {
-        return locale;
-    }
-
-    public @Nullable String listeningItem() {
-        return listeningItem;
-    }
-
-    public @Nullable String listeningMelody() {
-        return listeningMelody;
-    }
+public record DialogContext(@Nullable KSService ks, @Nullable String keyword, STTService stt, TTSService tts,
+        @Nullable Voice voice, List<HumanLanguageInterpreter> hlis, AudioSource source, AudioSink sink, Locale locale,
+        String dialogGroup, @Nullable String locationItem, @Nullable String listeningItem,
+        @Nullable String listeningMelody) {
 
     /**
      * Builder for {@link DialogContext}
@@ -116,6 +48,8 @@ public static class Builder {
         private @Nullable Voice voice;
         private List<HumanLanguageInterpreter> hlis = List.of();
         // options
+        private String dialogGroup = "default";
+        private @Nullable String locationItem;
         private @Nullable String listeningItem;
         private @Nullable String listeningMelody;
         private String keyword;
@@ -189,6 +123,20 @@ public Builder withVoice(@Nullable Voice voice) {
             return this;
         }
 
+        public Builder withDialogGroup(@Nullable String dialogGroup) {
+            if (dialogGroup != null) {
+                this.dialogGroup = dialogGroup;
+            }
+            return this;
+        }
+
+        public Builder withLocationItem(@Nullable String locationItem) {
+            if (locationItem != null) {
+                this.locationItem = locationItem;
+            }
+            return this;
+        }
+
         public Builder withListeningItem(@Nullable String listeningItem) {
             if (listeningItem != null) {
                 this.listeningItem = listeningItem;
@@ -244,7 +192,7 @@ public DialogContext build() throws IllegalStateException {
                 throw new IllegalStateException("Cannot build dialog context: " + String.join(", ", errors) + ".");
             } else {
                 return new DialogContext(ksService, keyword, sttService, ttsService, voice, hliServices, audioSource,
-                        audioSink, locale, listeningItem, listeningMelody);
+                        audioSink, locale, dialogGroup, locationItem, listeningItem, listeningMelody);
             }
         }
     }

bundles/org.openhab.core.voice/src/main/java/org/openhab/core/voice/DialogRegistration.java

@@ -65,6 +65,14 @@ public class DialogRegistration {
      * Linked listening item
      */
     public @Nullable String listeningItem;
+    /**
+     * Linked location item
+     */
+    public @Nullable String locationItem;
+    /**
+     * Dialog group name
+     */
+    public @Nullable String dialogGroup;
     /**
      * Custom listening melody
      */

bundles/org.openhab.core.voice/src/main/java/org/openhab/core/voice/internal/DialogProcessor.java

@@ -17,6 +17,7 @@
 import java.util.HashSet;
 import java.util.List;
 import java.util.Objects;
+import java.util.WeakHashMap;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
 
@@ -69,13 +70,13 @@
  * @author Miguel Álvarez - Close audio streams + use RecognitionStartEvent
  * @author Miguel Álvarez - Use dialog context
  * @author Miguel Álvarez - Add sounds
+ * @author Miguel Álvarez - Add dialog groups
  *
  */
 @NonNullByDefault
 public class DialogProcessor implements KSListener, STTListener {
-
     private final Logger logger = LoggerFactory.getLogger(DialogProcessor.class);
-
+    private final WeakHashMap<String, DialogContext> activeDialogGroups;
     public final DialogContext dialogContext;
     private @Nullable List<ToneSynthesizer.Tone> listeningMelody;
     private final EventPublisher eventPublisher;
@@ -105,11 +106,12 @@ public class DialogProcessor implements KSListener, STTListener {
     private @Nullable ToneSynthesizer toneSynthesizer;
 
     public DialogProcessor(DialogContext context, DialogEventListener eventListener, EventPublisher eventPublisher,
-            TranslationProvider i18nProvider, Bundle bundle) {
+            WeakHashMap<String, DialogContext> activeDialogGroups, TranslationProvider i18nProvider, Bundle bundle) {
         this.dialogContext = context;
         this.eventListener = eventListener;
         this.eventPublisher = eventPublisher;
         this.i18nProvider = i18nProvider;
+        this.activeDialogGroups = activeDialogGroups;
         this.bundle = bundle;
         var ks = context.ks();
         this.ksFormat = ks != null
@@ -182,7 +184,15 @@ public void start() throws IllegalStateException {
      * Starts a single dialog
      */
     public void startSimpleDialog() {
-        abortSTT();
+        synchronized (activeDialogGroups) {
+            if (!activeDialogGroups.containsKey(dialogContext.dialogGroup())) {
+                logger.debug("Acquiring dialog group '{}'", dialogContext.dialogGroup());
+                activeDialogGroups.put(dialogContext.dialogGroup(), dialogContext);
+            } else {
+                logger.warn("Ignoring keyword spotting event, dialog group '{}' running", dialogContext.dialogGroup());
+                return;
+            }
+        }
         closeStreamSTT();
         isSTTServerAborting = false;
         AudioFormat fmt = sttFormat;
@@ -196,6 +206,7 @@ public void startSimpleDialog() {
             AudioStream stream = dialogContext.source().getInputStream(fmt);
             streamSTT = stream;
             sttServiceHandle = dialogContext.stt().recognize(this, stream, dialogContext.locale(), new HashSet<>());
+            return;
         } catch (AudioException e) {
             logger.warn("Error creating the audio stream: {}", e.getMessage());
         } catch (STTException e) {
@@ -208,6 +219,11 @@ public void startSimpleDialog() {
                 say(text.replace("{0}", ""));
             }
         }
+        // In case of error release dialog group
+        synchronized (activeDialogGroups) {
+            logger.debug("Releasing dialog group '{}' due to errors", dialogContext.dialogGroup());
+            activeDialogGroups.remove(dialogContext.dialogGroup());
+        }
     }
 
     /**
@@ -264,6 +280,10 @@ private void abortSTT() {
             sttServiceHandle = null;
         }
         isSTTServerAborting = true;
+        synchronized (activeDialogGroups) {
+            logger.debug("Releasing dialog group '{}'", dialogContext.dialogGroup());
+            activeDialogGroups.remove(dialogContext.dialogGroup());
+        }
     }
 
     private void closeStreamSTT() {
@@ -292,20 +312,18 @@ private void toggleProcessing(boolean value) {
 
     @Override
     public void ksEventReceived(KSEvent ksEvent) {
-        if (!processing) {
-            isSTTServerAborting = false;
-            if (ksEvent instanceof KSpottedEvent) {
-                logger.debug("KSpottedEvent event received");
-                try {
-                    startSimpleDialog();
-                } catch (IllegalStateException e) {
-                    logger.warn("{}", e.getMessage());
-                }
-            } else if (ksEvent instanceof KSErrorEvent kse) {
-                logger.debug("KSErrorEvent event received");
-                String text = i18nProvider.getText(bundle, "error.ks-error", null, dialogContext.locale());
-                say(text == null ? kse.getMessage() : text.replace("{0}", kse.getMessage()));
+        isSTTServerAborting = false;
+        if (ksEvent instanceof KSpottedEvent) {
+            logger.debug("KSpottedEvent event received");
+            try {
+                startSimpleDialog();
+            } catch (IllegalStateException e) {
+                logger.warn("{}", e.getMessage());
             }
+        } else if (ksEvent instanceof KSErrorEvent kse) {
+            logger.debug("KSErrorEvent event received");
+            String text = i18nProvider.getText(bundle, "error.ks-error", null, dialogContext.locale());
+            say(text == null ? kse.getMessage() : text.replace("{0}", kse.getMessage()));
         }
     }
 
@@ -322,7 +340,7 @@ public synchronized void sttEventReceived(STTEvent sttEvent) {
                 String error = null;
                 for (HumanLanguageInterpreter interpreter : dialogContext.hlis()) {
                     try {
-                        answer = interpreter.interpret(dialogContext.locale(), question);
+                        answer = interpreter.interpret(dialogContext.locale(), question, dialogContext);
                         logger.debug("Interpretation result: {}", answer);
                         error = null;
                         break;

bundles/org.openhab.core.voice/src/main/java/org/openhab/core/voice/internal/VoiceConsoleCommandExtension.java

@@ -97,17 +97,17 @@ public List<String> getUsages() {
                 buildCommandUsage(SUBCMD_DIALOG_REGS,
                         "lists the existing dialog registrations and their selected audio/voice services"),
                 buildCommandUsage(SUBCMD_REGISTER_DIALOG
-                        + " [--source <source>] [--sink <sink>] [--hlis <comma,separated,interpreters>] [--tts <tts> [--voice <voice>]] [--stt <stt>] [--ks ks [--keyword <ks>]] [--listening-item <listeningItem>]",
+                        + " [--source <source>] [--sink <sink>] [--hlis <comma,separated,interpreters>] [--tts <tts> [--voice <voice>]] [--stt <stt>] [--ks ks [--keyword <ks>]] [--listening-item <listeningItem>] [--location-item <locationItem>] [--dialog-group <dialogGroup>]",
                         "register a new dialog processing using the default services or the services identified with provided arguments, it will be persisted and keep running whenever is possible."),
                 buildCommandUsage(SUBCMD_UNREGISTER_DIALOG + " [source]",
                         "unregister the dialog processing for the default audio source or the audio source identified with provided argument, stopping it if started"),
                 buildCommandUsage(SUBCMD_START_DIALOG
-                        + " [--source <source>] [--sink <sink>] [--hlis <comma,separated,interpreters>] [--tts <tts> [--voice <voice>]] [--stt <stt>] [--ks ks [--keyword <ks>]] [--listening-item <listeningItem>]",
+                        + " [--source <source>] [--sink <sink>] [--hlis <comma,separated,interpreters>] [--tts <tts> [--voice <voice>]] [--stt <stt>] [--ks ks [--keyword <ks>]] [--listening-item <listeningItem>] [--location-item <locationItem>] [--dialog-group <dialogGroup>]",
                         "start a new dialog processing using the default services or the services identified with provided arguments"),
                 buildCommandUsage(SUBCMD_STOP_DIALOG + " [<source>]",
                         "stop the dialog processing for the default audio source or the audio source identified with provided argument"),
                 buildCommandUsage(SUBCMD_LISTEN_ANSWER
-                        + " [--source <source>] [--sink <sink>] [--hlis <comma,separated,interpreters>] [--tts <tts> [--voice <voice>]] [--stt <stt>] [--listening-item <listeningItem>]",
+                        + " [--source <source>] [--sink <sink>] [--hlis <comma,separated,interpreters>] [--tts <tts> [--voice <voice>]] [--stt <stt>] [--listening-item <listeningItem>] [--location-item <locationItem>] [--dialog-group <dialogGroup>]",
                         "Execute a simple dialog sequence without keyword spotting using the default services or the services identified with provided arguments"),
                 buildCommandUsage(SUBCMD_INTERPRETERS, "lists the interpreters"),
                 buildCommandUsage(SUBCMD_KEYWORD_SPOTTERS, "lists the keyword spotters"),
@@ -309,11 +309,12 @@ private void listDialogRegistrations(Console console) {
         Collection<DialogRegistration> registrations = voiceManager.getDialogRegistrations();
         if (!registrations.isEmpty()) {
             registrations.stream().sorted(comparing(dr -> dr.sourceId)).forEach(dr -> {
-                console.println(
-                        String.format(" Source: %s - Sink: %s (STT: %s, TTS: %s, HLIs: %s, KS: %s, Keyword: %s)",
-                                dr.sourceId, dr.sinkId, getOrDefault(dr.sttId), getOrDefault(dr.ttsId),
-                                dr.hliIds.isEmpty() ? getOrDefault(null) : String.join("->", dr.hliIds),
-                                getOrDefault(dr.ksId), getOrDefault(dr.keyword)));
+                String locationText = dr.locationItem != null ? String.format(" Location: %s", dr.locationItem) : "";
+                console.println(String.format(
+                        " Source: %s - Sink: %s (STT: %s, TTS: %s, HLIs: %s, KS: %s, Keyword: %s, Dialog Group: %s)%s",
+                        dr.sourceId, dr.sinkId, getOrDefault(dr.sttId), getOrDefault(dr.ttsId),
+                        dr.hliIds.isEmpty() ? getOrDefault(null) : String.join("->", dr.hliIds), getOrDefault(dr.ksId),
+                        getOrDefault(dr.keyword), getOrDefault(dr.dialogGroup), locationText));
             });
         } else {
             console.println("No dialog registrations.");
@@ -330,11 +331,12 @@ private void listDialogs(Console console) {
             dialogContexts.stream().sorted(comparing(s -> s.source().getId())).forEach(c -> {
                 var ks = c.ks();
                 String ksText = ks != null ? String.format(", KS: %s, Keyword: %s", ks.getId(), c.keyword()) : "";
-                console.println(
-                        String.format(" Source: %s - Sink: %s (STT: %s, TTS: %s, HLIs: %s%s)", c.source().getId(),
-                                c.sink().getId(), c.stt().getId(), c.tts().getId(), c.hlis().stream()
-                                        .map(HumanLanguageInterpreter::getId).collect(Collectors.joining("->")),
-                                ksText));
+                String locationText = c.locationItem() != null ? String.format(" Location: %s", c.locationItem()) : "";
+                console.println(String.format(
+                        " Source: %s - Sink: %s (STT: %s, TTS: %s, HLIs: %s%s, Dialog Group: %s)%s", c.source().getId(),
+                        c.sink().getId(), c.stt().getId(), c.tts().getId(),
+                        c.hlis().stream().map(HumanLanguageInterpreter::getId).collect(Collectors.joining("->")),
+                        ksText, c.dialogGroup(), locationText));
             });
         } else {
             console.println("No running dialogs.");
@@ -450,6 +452,8 @@ private DialogContext.Builder parseDialogContext(String[] args) {
                 .withHLIs(voiceManager.getHLIsByIds(parameters.remove("hlis"))) //
                 .withKS(voiceManager.getKS(parameters.remove("ks"))) //
                 .withListeningItem(parameters.remove("listening-item")) //
+                .withLocationItem(parameters.remove("location-item")) //
+                .withDialogGroup(parameters.remove("dialog-group")) //
                 .withKeyword(parameters.remove("keyword"));
         if (!parameters.isEmpty()) {
             throw new IllegalStateException(
@@ -483,6 +487,9 @@ private DialogRegistration parseDialogRegistration(String[] args) {
         dr.ttsId = parameters.remove("tts");
         dr.voiceId = parameters.remove("voice");
         dr.listeningItem = parameters.remove("listening-item");
+        dr.locationItem = parameters.remove("location-item");
+        dr.dialogGroup = parameters.remove("dialog-group");
+
         String hliIds = parameters.remove("hlis");
         if (hliIds != null) {
             dr.hliIds = Arrays.stream(hliIds.split(",")).map(String::trim).collect(Collectors.toList());

bundles/org.openhab.core.voice/src/main/java/org/openhab/core/voice/internal/VoiceManagerImpl.java

@@ -26,6 +26,7 @@
 import java.util.Map;
 import java.util.Objects;
 import java.util.Set;
+import java.util.WeakHashMap;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ScheduledExecutorService;
 import java.util.concurrent.ScheduledFuture;
@@ -116,6 +117,8 @@ public class VoiceManagerImpl implements VoiceManager, ConfigOptionProvider, Dia
     private final Map<String, TTSService> ttsServices = new HashMap<>();
     private final Map<String, HumanLanguageInterpreter> humanLanguageInterpreters = new HashMap<>();
 
+    private final WeakHashMap<String, DialogContext> activeDialogGroups = new WeakHashMap<>();
+
     private final LocaleProvider localeProvider;
     private final AudioManager audioManager;
     private final EventPublisher eventPublisher;
@@ -526,7 +529,8 @@ public void startDialog(DialogContext context) throws IllegalStateException {
             if (processor == null) {
                 logger.debug("Starting a new dialog for source {} ({})", context.source().getLabel(null),
                         context.source().getId());
-                processor = new DialogProcessor(context, this, this.eventPublisher, this.i18nProvider, b);
+                processor = new DialogProcessor(context, this, this.eventPublisher, this.activeDialogGroups,
+                        this.i18nProvider, b);
                 dialogProcessors.put(context.source().getId(), processor);
                 processor.start();
             } else {
@@ -582,7 +586,8 @@ public void listenAndAnswer(DialogContext context) throws IllegalStateException
                 isSingleDialog = true;
                 activeProcessor = singleDialogProcessors.get(audioSource.getId());
             }
-            var processor = new DialogProcessor(context, this, this.eventPublisher, this.i18nProvider, b);
+            var processor = new DialogProcessor(context, this, this.eventPublisher, this.activeDialogGroups,
+                    this.i18nProvider, b);
             if (activeProcessor == null) {
                 logger.debug("Executing a simple dialog for source {} ({})", audioSource.getLabel(null),
                         audioSource.getId());
@@ -970,6 +975,8 @@ private void buildDialogRegistrations() {
                                 .withVoice(getVoice(dr.voiceId)) //
                                 .withHLIs(getHLIsByIds(dr.hliIds)) //
                                 .withLocale(dr.locale) //
+                                .withDialogGroup(dr.dialogGroup) //
+                                .withLocationItem(dr.locationItem) //
                                 .withListeningItem(dr.listeningItem) //
                                 .withMelody(dr.listeningMelody) //
                                 .build());