@absolutejs/voice 0.0.22-beta.483 → 0.0.22-beta.484

package/dist/assistantMode.d.ts

@@ -0,0 +1,22 @@
+import type { CreateVoiceSessionOptions, VoiceSessionRecord } from "./types";
+export type VoiceAssistantMode = "cascade" | "s2s";
+export type VoiceSemanticVADConfig = {
+    createResponseAutomatically?: boolean;
+    eagerness?: "auto" | "high" | "low" | "medium";
+    silenceDurationMs?: number;
+};
+export type VoiceAssistantModality = "audio" | "text";
+export declare const resolveVoiceAssistantMode: <TContext = unknown, TSession extends VoiceSessionRecord = VoiceSessionRecord, TResult = unknown>(options: Pick<CreateVoiceSessionOptions<TContext, TSession, TResult>, "realtime" | "stt" | "tts"> & {
+    assistantMode?: VoiceAssistantMode;
+}) => VoiceAssistantMode;
+export type VoiceAssistantModeDescriptor = {
+    hasRealtime: boolean;
+    hasSTT: boolean;
+    hasTTS: boolean;
+    mode: VoiceAssistantMode;
+    modalities: VoiceAssistantModality[];
+};
+export declare const describeVoiceAssistantMode: <TContext = unknown, TSession extends VoiceSessionRecord = VoiceSessionRecord, TResult = unknown>(options: Pick<CreateVoiceSessionOptions<TContext, TSession, TResult>, "realtime" | "stt" | "tts"> & {
+    assistantMode?: VoiceAssistantMode;
+    modalities?: ReadonlyArray<VoiceAssistantModality>;
+}) => VoiceAssistantModeDescriptor;

package/dist/index.d.ts

@@ -77,6 +77,8 @@ export { DEFAULT_VOICE_REDACTION_PATTERNS, createVoiceTranscriptRedactor, redact
 export type { CreateVoiceTranscriptRedactorOptions, VoiceRedactionPattern, VoiceTranscriptRedactor, } from "./redaction";
 export { DEFAULT_VOICE_PRICE_BOOK, createVoiceCostAccountant, } from "./costAccounting";
 export type { CreateVoiceCostAccountantOptions, VoiceCostAccountant, VoiceCostBreakdown, VoiceCostLLMRecord, VoiceCostSTTRecord, VoiceCostTTSRecord, VoiceCostTelephonyRecord, VoicePriceBook, VoiceProviderRates, } from "./costAccounting";
+export { describeVoiceAssistantMode, resolveVoiceAssistantMode, } from "./assistantMode";
+export type { VoiceAssistantMode, VoiceAssistantModality, VoiceAssistantModeDescriptor, VoiceSemanticVADConfig, } from "./assistantMode";
 export { createPunctuationSemanticTurnDetector, createRegexSemanticTurnDetector, } from "./semanticTurn";
 export type { CreatePunctuationSemanticTurnDetectorOptions, CreateRegexSemanticTurnDetectorOptions, VoiceSemanticTurnDetector, VoiceSemanticTurnInput, VoiceSemanticTurnVerdict, } from "./semanticTurn";
 export { createMonologueAMDDetector } from "./amdDetector";

package/dist/index.js

@@ -3440,6 +3440,28 @@ var createVoiceMemoryRecordingStore = () => {
   };
 };
 
+// src/assistantMode.ts
+var resolveVoiceAssistantMode = (options) => {
+  if (options.assistantMode) {
+    return options.assistantMode;
+  }
+  if (options.realtime) {
+    return "s2s";
+  }
+  return "cascade";
+};
+var describeVoiceAssistantMode = (options) => {
+  const mode = resolveVoiceAssistantMode(options);
+  const modalities = options.modalities ? Array.from(new Set(options.modalities)) : ["audio"];
+  return {
+    hasRealtime: Boolean(options.realtime),
+    hasSTT: Boolean(options.stt),
+    hasTTS: Boolean(options.tts),
+    modalities,
+    mode
+  };
+};
+
 // src/session.ts
 var DEFAULT_RECONNECT_TIMEOUT = 30000;
 var DEFAULT_MAX_RECONNECT_ATTEMPTS = 10;
@@ -4822,13 +4844,20 @@ var createVoiceSession = (options) => {
     if (!inputAdapter) {
       throw new Error("Voice session requires either an stt or realtime adapter.");
     }
-    const openedSession = await inputAdapter.open({
-      format: options.realtime ? options.realtimeInputFormat ?? DEFAULT_REALTIME_FORMAT : DEFAULT_FORMAT,
+    const openedSession = await (options.realtime ? options.realtime.open({
+      format: options.realtimeInputFormat ?? DEFAULT_REALTIME_FORMAT,
       languageStrategy: options.languageStrategy,
       lexicon,
+      modalities: options.modalities,
       phraseHints,
       sessionId: options.id
-    });
+    }) : inputAdapter.open({
+      format: DEFAULT_FORMAT,
+      languageStrategy: options.languageStrategy,
+      lexicon,
+      phraseHints,
+      sessionId: options.id
+    }));
     const generation = ++adapterGenerationCounter;
     sttSession = openedSession;
     activeAdapterGeneration = generation;
@@ -4998,9 +5027,10 @@ var createVoiceSession = (options) => {
       });
       await appendTrace({
         payload: {
+          assistantMode: resolveVoiceAssistantMode(options),
+          realtimeConfigured: Boolean(options.realtime),
           text: output.assistantText,
-          ttsConfigured: Boolean(options.tts),
-          realtimeConfigured: Boolean(options.realtime)
+          ttsConfigured: Boolean(options.tts)
         },
         session,
         turnId: turn.id,
@@ -46034,6 +46064,7 @@ export {
   resolveVoiceDiagnosticsTraceFilter,
   resolveVoiceAuditTrailFilter,
   resolveVoiceAuditDeliveryFilter,
+  resolveVoiceAssistantMode,
   resolveVoiceAssistantMemoryNamespace,
   resolveTurnDetectionConfig,
   resolveLatestVoiceCallDebuggerSessionId,
@@ -46228,6 +46259,7 @@ export {
   evaluateVoiceAgentSquadContractEvidence,
   encodeTwilioMulawBase64,
   encodePcmAsWav,
+  describeVoiceAssistantMode,
   deliverVoiceTraceEventsToSinks,
   deliverVoiceObservabilityExport,
   deliverVoiceMonitorIssueNotifications,

package/dist/testing/index.js

@@ -5408,6 +5408,28 @@ var createVoiceMemoryRecordingStore = () => {
   };
 };
 
+// src/assistantMode.ts
+var resolveVoiceAssistantMode = (options) => {
+  if (options.assistantMode) {
+    return options.assistantMode;
+  }
+  if (options.realtime) {
+    return "s2s";
+  }
+  return "cascade";
+};
+var describeVoiceAssistantMode = (options) => {
+  const mode = resolveVoiceAssistantMode(options);
+  const modalities = options.modalities ? Array.from(new Set(options.modalities)) : ["audio"];
+  return {
+    hasRealtime: Boolean(options.realtime),
+    hasSTT: Boolean(options.stt),
+    hasTTS: Boolean(options.tts),
+    modalities,
+    mode
+  };
+};
+
 // src/session.ts
 var DEFAULT_RECONNECT_TIMEOUT = 30000;
 var DEFAULT_MAX_RECONNECT_ATTEMPTS2 = 10;
@@ -6790,13 +6812,20 @@ var createVoiceSession = (options) => {
     if (!inputAdapter) {
       throw new Error("Voice session requires either an stt or realtime adapter.");
     }
-    const openedSession = await inputAdapter.open({
-      format: options.realtime ? options.realtimeInputFormat ?? DEFAULT_REALTIME_FORMAT : DEFAULT_FORMAT,
+    const openedSession = await (options.realtime ? options.realtime.open({
+      format: options.realtimeInputFormat ?? DEFAULT_REALTIME_FORMAT,
       languageStrategy: options.languageStrategy,
       lexicon,
+      modalities: options.modalities,
       phraseHints,
       sessionId: options.id
-    });
+    }) : inputAdapter.open({
+      format: DEFAULT_FORMAT,
+      languageStrategy: options.languageStrategy,
+      lexicon,
+      phraseHints,
+      sessionId: options.id
+    }));
     const generation = ++adapterGenerationCounter;
     sttSession = openedSession;
     activeAdapterGeneration = generation;
@@ -6966,9 +6995,10 @@ var createVoiceSession = (options) => {
       });
       await appendTrace({
         payload: {
+          assistantMode: resolveVoiceAssistantMode(options),
+          realtimeConfigured: Boolean(options.realtime),
           text: output.assistantText,
-          ttsConfigured: Boolean(options.tts),
-          realtimeConfigured: Boolean(options.realtime)
+          ttsConfigured: Boolean(options.tts)
         },
         session,
         turnId: turn.id,

package/dist/types.d.ts

@@ -194,7 +194,10 @@ export type RealtimeAdapterOpenOptions = {
     format: AudioFormat;
     languageStrategy?: VoiceLanguageStrategy;
     lexicon?: VoiceLexiconEntry[];
+    modalities?: ReadonlyArray<"audio" | "text">;
     phraseHints?: VoicePhraseHint[];
+    promptCacheKey?: string;
+    semanticVAD?: import("./assistantMode").VoiceSemanticVADConfig;
     signal?: AbortSignal;
 };
 export type RealtimeAdapter<TOptions extends RealtimeAdapterOpenOptions = RealtimeAdapterOpenOptions> = {
@@ -732,6 +735,8 @@ export type CreateVoiceSessionOptions<TContext = unknown, TSession extends Voice
     };
     redact?: import("./redaction").VoiceTranscriptRedactor;
     semanticTurnDetector?: import("./semanticTurn").VoiceSemanticTurnDetector;
+    assistantMode?: import("./assistantMode").VoiceAssistantMode;
+    modalities?: ReadonlyArray<"audio" | "text">;
     reconnect: Required<VoiceReconnectConfig>;
     phraseHints?: VoicePhraseHint[];
     sessionMetadata?: Record<string, unknown>;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.483",
+  "version": "0.0.22-beta.484",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",