@simonpeacocks/peacockery-voice-client 0.1.4 → 0.2.0

package/dist/index.d.ts

@@ -1,7 +1,7 @@
 import { type PeacockeryVoiceClientOptions } from "./auth.js";
 import type { paths } from "./schema.js";
 export { type PeacockeryVoiceClientOptions, type PeacockeryVoiceDpopBindingProofInput, type PeacockeryVoiceDpopBindingProofProvider, type PeacockeryVoiceDpopProofInput, type PeacockeryVoiceDpopProofProvider, type PeacockeryVoiceEnvironment, peacockeryVoiceBaseUrls, } from "./auth.js";
-export { createPeacockeryVoiceRealtimeRequest, type PeacockeryVoiceRealtimeAudioReceivedEvent, type PeacockeryVoiceRealtimeEvent, type PeacockeryVoiceRealtimeOptions, type PeacockeryVoiceRealtimePongEvent, type PeacockeryVoiceRealtimeRequest, type PeacockeryVoiceRealtimeSessionStartedEvent, type PeacockeryVoiceRealtimeTranscriptEvent, parsePeacockeryVoiceRealtimeEvent, } from "./realtime.js";
+export { createPeacockeryVoiceRealtimeRequest, type PeacockeryVoiceRealtimeAudioReceivedEvent, type PeacockeryVoiceRealtimeEvent, type PeacockeryVoiceRealtimeOptions, type PeacockeryVoiceRealtimePongEvent, type PeacockeryVoiceRealtimeRequest, type PeacockeryVoiceRealtimeSessionStartedEvent, type PeacockeryVoiceRealtimeTranscriptEvent, type PeacockeryVoiceRealtimeTurnEvent, parsePeacockeryVoiceRealtimeEvent, } from "./realtime.js";
 export type { components, operations, paths } from "./schema.js";
 export declare const createPeacockeryVoiceClient: (options: PeacockeryVoiceClientOptions) => import("openapi-fetch").Client<paths, `${string}/${string}`>;
 export type PeacockeryVoiceClient = ReturnType<typeof createPeacockeryVoiceClient>;

package/dist/realtime.d.ts

@@ -1,19 +1,43 @@
 import { type PeacockeryVoiceClientOptions } from "./auth.js";
 import type { components } from "./schema.js";
 type RealtimeTerminalEvent = components["schemas"]["RealtimeSessionResult"];
-type TranscriptionArtifact = components["schemas"]["TranscriptionArtifact"];
-type TranscriptSegment = TranscriptionArtifact["content"]["segments"]["items"][number];
-type TranscriptSpeakerTurn = TranscriptionArtifact["content"]["speaker_turns"]["items"][number];
-type TranscriptWord = TranscriptionArtifact["content"]["words"]["items"][number];
+interface TranscriptSegment {
+    endSeconds: number | null;
+    scores?: {
+        confidence?: number;
+        logProbability?: number;
+        probability?: number;
+        score?: number;
+        speakerConfidence?: number;
+    };
+    speaker?: string | number;
+    startSeconds: number | null;
+    text: string;
+}
+type TranscriptSpeakerTurn = Omit<TranscriptSegment, "scores">;
+type TranscriptWord = TranscriptSegment & {
+    startSeconds: number;
+    endSeconds: number;
+};
 export interface PeacockeryVoiceRealtimeOptions extends PeacockeryVoiceClientOptions {
     channels?: number;
+    contextBias?: string[];
+    diarize?: boolean;
     encoding?: string;
     endpointing?: string;
     interimResults?: boolean;
     language?: string;
     model: string;
     sampleRate?: number;
+    speakerReferences?: Record<string, string>;
     targetStreamingDelayMs?: number;
+    turnDetection?: {
+        start_threshold?: number;
+        eager_end_threshold?: number;
+        end_threshold?: number;
+        min_silence_ms?: number;
+        end_timeout_ms?: number;
+    };
     utteranceEndMs?: number;
     vadEvents?: boolean;
 }
@@ -51,7 +75,12 @@ export interface PeacockeryVoiceRealtimePongEvent {
     session_id: string;
     type: "pong";
 }
-export type PeacockeryVoiceRealtimeEvent = PeacockeryVoiceRealtimeAudioReceivedEvent | PeacockeryVoiceRealtimePongEvent | PeacockeryVoiceRealtimeSessionStartedEvent | PeacockeryVoiceRealtimeTranscriptEvent | RealtimeTerminalEvent;
+export interface PeacockeryVoiceRealtimeTurnEvent extends RealtimeEventBase {
+    end_reason?: "threshold" | "timeout";
+    time_seconds: number;
+    type: "turn.start" | "turn.eager_end" | "turn.resume" | "turn.end";
+}
+export type PeacockeryVoiceRealtimeEvent = PeacockeryVoiceRealtimeTurnEvent | PeacockeryVoiceRealtimeAudioReceivedEvent | PeacockeryVoiceRealtimePongEvent | PeacockeryVoiceRealtimeSessionStartedEvent | PeacockeryVoiceRealtimeTranscriptEvent | RealtimeTerminalEvent;
 export declare const createPeacockeryVoiceRealtimeRequest: (options: PeacockeryVoiceRealtimeOptions) => Promise<PeacockeryVoiceRealtimeRequest>;
 export declare const parsePeacockeryVoiceRealtimeEvent: (value: string | unknown) => PeacockeryVoiceRealtimeEvent;
 export {};

package/dist/realtime.js

@@ -4,13 +4,20 @@ const realtimeHttpUrl = (options) => {
     const url = new URL("/v1/realtime", options.baseUrl ?? peacockeryVoiceBaseUrls[environment]);
     const query = {
         channels: options.channels,
+        diarize: options.diarize,
         encoding: options.encoding,
         endpointing: options.endpointing,
         interim_results: options.interimResults,
         language: options.language,
         model: options.model,
         sample_rate: options.sampleRate,
+        speaker_references: options.speakerReferences
+            ? JSON.stringify(options.speakerReferences)
+            : undefined,
         target_streaming_delay_ms: options.targetStreamingDelayMs,
+        turn_detection: options.turnDetection
+            ? JSON.stringify(options.turnDetection)
+            : undefined,
         utterance_end_ms: options.utteranceEndMs,
         vad_events: options.vadEvents,
     };
@@ -19,6 +26,9 @@ const realtimeHttpUrl = (options) => {
             url.searchParams.set(name, String(value));
         }
     }
+    for (const term of options.contextBias ?? []) {
+        url.searchParams.append("context_bias", term);
+    }
     return url;
 };
 const headersRecord = (headers) => {
@@ -79,6 +89,12 @@ export const createPeacockeryVoiceRealtimeRequest = async (options) => {
     return { headers: headersRecord(headers), url: url.href };
 };
 const isRecord = (value) => typeof value === "object" && value !== null;
+const isTurnEvent = (value) => typeof value.time_seconds === "number" &&
+    Number.isFinite(value.time_seconds) &&
+    value.time_seconds >= 0 &&
+    (value.end_reason === undefined ||
+        value.end_reason === "threshold" ||
+        value.end_reason === "timeout");
 const isRealtimeEvent = (value) => {
     if (!isRecord(value) || typeof value.type !== "string") {
         return false;
@@ -99,6 +115,9 @@ const isRealtimeEvent = (value) => {
     if (!hasProtocolBase) {
         return false;
     }
+    if (["turn.start", "turn.eager_end", "turn.resume", "turn.end"].includes(value.type)) {
+        return isTurnEvent(value);
+    }
     if (value.type === "session.started") {
         return ((typeof value.language === "string" || value.language === null) &&
             typeof value.model === "string");

package/dist/schema.d.ts

@@ -1402,6 +1402,7 @@ export interface paths {
                 query?: {
                     audio_format?: string;
                     channels?: number;
+                    context_bias?: string;
                     detect_entities?: "true" | "false";
                     diarize?: "true" | "false";
                     diarize_model?: "latest" | "v1";
@@ -1425,8 +1426,12 @@ export interface paths {
                     sample_rate?: number;
                     search?: string;
                     smart_format?: "true" | "false";
+                    /** @description JSON object mapping speaker names to completed upload input_key values */
+                    speaker_references?: string;
                     tag?: string;
                     target_streaming_delay_ms?: number;
+                    /** @description JSON turn-detection configuration; {} uses provider defaults */
+                    turn_detection?: string;
                     utterance_end_ms?: number;
                     vad_events?: "true" | "false";
                     version?: string;
@@ -2114,7 +2119,7 @@ export interface components {
                     /** @enum {string|null} */
                     source: "derived" | "provider" | null;
                     items: {
-                        end_seconds: number;
+                        end_seconds: number | null;
                         scores: {
                             confidence: number | null;
                             log_probability: number | null;
@@ -2123,7 +2128,7 @@ export interface components {
                             speaker_confidence: number | null;
                         } | null;
                         speaker: string | number | null;
-                        start_seconds: number;
+                        start_seconds: number | null;
                         text: string;
                     }[];
                 };
@@ -2133,9 +2138,9 @@ export interface components {
                     /** @enum {string|null} */
                     source: "derived" | "provider" | null;
                     items: {
-                        end_seconds: number;
+                        end_seconds: number | null;
                         speaker: string | number | null;
-                        start_seconds: number;
+                        start_seconds: number | null;
                         text: string;
                         scores: null;
                     }[];

package/package.json

@@ -1,14 +1,14 @@
 {
   "name": "@simonpeacocks/peacockery-voice-client",
-  "version": "0.1.4",
+  "version": "0.2.0",
   "description": "Generated TypeScript client for the Peacockery Voice API",
-  "type": "module",
-  "license": "MIT",
   "repository": {
     "type": "git",
     "url": "git+https://github.com/chikingsley/peacockery-voice.git",
     "directory": "clients/typescript"
   },
+  "license": "MIT",
+  "type": "module",
   "exports": {
     ".": {
       "types": "./dist/index.d.ts",
@@ -18,10 +18,6 @@
   "files": [
     "dist"
   ],
-  "publishConfig": {
-    "access": "public",
-    "provenance": false
-  },
   "scripts": {
     "build": "tsc",
     "check": "tsc --noEmit",
@@ -32,5 +28,9 @@
   },
   "devDependencies": {
     "typescript": "^7.0.2"
+  },
+  "publishConfig": {
+    "access": "public",
+    "provenance": false
   }
 }