@simonpeacocks/peacockery-voice-client 0.1.4 → 0.3.0

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

@@ -1288,6 +1288,93 @@ export interface paths {
         patch?: never;
         trace?: never;
     };
+    "/v1/diarizations": {
+        parameters: {
+            query?: never;
+            header?: never;
+            path?: never;
+            cookie?: never;
+        };
+        get?: never;
+        put?: never;
+        /** Create a standalone speaker-track job */
+        post: {
+            parameters: {
+                query?: never;
+                header?: {
+                    "idempotency-key"?: string;
+                };
+                path?: never;
+                cookie?: never;
+            };
+            requestBody: {
+                content: {
+                    "application/json": components["schemas"]["DiarizationRequest"];
+                };
+            };
+            responses: {
+                /** @description Existing idempotent job */
+                200: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["JobView"];
+                    };
+                };
+                /** @description Queued diarization */
+                202: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["JobView"];
+                    };
+                };
+                /** @description Invalid request */
+                400: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["ErrorResponse"];
+                    };
+                };
+                /** @description Unauthorized */
+                401: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["ErrorResponse"];
+                    };
+                };
+                /** @description Owned input or linked transcript unavailable */
+                404: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["ErrorResponse"];
+                    };
+                };
+                /** @description Idempotency conflict */
+                409: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["ErrorResponse"];
+                    };
+                };
+            };
+        };
+        delete?: never;
+        options?: never;
+        head?: never;
+        patch?: never;
+        trace?: never;
+    };
     "/v1/jobs/{job_id}": {
         parameters: {
             query?: never;
@@ -1402,6 +1489,7 @@ 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 +1513,12 @@ 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;
@@ -1544,6 +1636,64 @@ export interface paths {
         patch?: never;
         trace?: never;
     };
+    "/v1/realtime/sessions/{session_id}/diarizations": {
+        parameters: {
+            query?: never;
+            header?: never;
+            path?: never;
+            cookie?: never;
+        };
+        /** List speaker-track jobs linked to an owned realtime session */
+        get: {
+            parameters: {
+                query?: never;
+                header?: never;
+                path: {
+                    session_id: string;
+                };
+                cookie?: never;
+            };
+            requestBody?: never;
+            responses: {
+                /** @description Linked jobs with separate speaker artifacts */
+                200: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": {
+                            jobs: components["schemas"]["JobView"][];
+                        };
+                    };
+                };
+                /** @description Unauthorized */
+                401: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["ErrorResponse"];
+                    };
+                };
+                /** @description Session unavailable */
+                404: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["ErrorResponse"];
+                    };
+                };
+            };
+        };
+        put?: never;
+        post?: never;
+        delete?: never;
+        options?: never;
+        head?: never;
+        patch?: never;
+        trace?: never;
+    };
     "/v1/text": {
         parameters: {
             query?: never;
@@ -2075,10 +2225,10 @@ export interface components {
             error_details: components["schemas"]["ProviderFailureDiagnostic"];
             job_id: string;
             /** @enum {string} */
-            kind: "transcription";
+            kind: "transcription" | "diarization";
             progress: number;
             queued_at: string | null;
-            result: components["schemas"]["TranscriptionArtifact"] | null;
+            result: components["schemas"]["TranscriptionArtifact"] | components["schemas"]["DiarizationArtifact"] | null;
             started_at: string | null;
             /** @enum {string} */
             status: "pending" | "queued" | "running" | "succeeded" | "failed";
@@ -2114,7 +2264,7 @@ 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 +2273,7 @@ export interface components {
                             speaker_confidence: number | null;
                         } | null;
                         speaker: string | number | null;
-                        start_seconds: number;
+                        start_seconds: number | null;
                         text: string;
                     }[];
                 };
@@ -2133,9 +2283,9 @@ 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;
                     }[];
@@ -2242,6 +2392,58 @@ export interface components {
                 message: string;
             }[];
         };
+        DiarizationArtifact: {
+            /** @enum {string} */
+            artifact_type: "speaker_tracks";
+            /** @enum {string} */
+            attribution: "separate_timeline_requires_alignment";
+            duration_seconds: number;
+            input: {
+                key: string;
+                r2_etag: string;
+                r2_version: string;
+                sha256: string;
+                size_bytes: number;
+            };
+            processing_seconds: number;
+            provenance: {
+                completed_at: string;
+                device: string;
+                job_id: string;
+                /** @enum {string} */
+                model: "nvidia/Nemotron-3-Diarization";
+                model_revision: string;
+                /** @enum {string} */
+                provider: "gmk";
+                provider_job_id: string;
+                /** @enum {string} */
+                revision_source: "deployment_manifest";
+                runtime: string;
+                /** @enum {string} */
+                speaker_origin: "model_activity";
+            };
+            /** @enum {number} */
+            schema_version: 1;
+            segmentation: {
+                /** @enum {number} */
+                minimum_duration_seconds: 0;
+                /** @enum {number} */
+                offset: 0.5;
+                /** @enum {number} */
+                onset: 0.5;
+                /** @enum {number} */
+                padding_seconds: 0;
+                /** @enum {string} */
+                preset: "neutral-0.5";
+            };
+            /** @enum {string} */
+            speaker_identity: "anonymous_recording_local";
+            tracks: {
+                end_seconds: number;
+                speaker: string;
+                start_seconds: number;
+            }[];
+        };
         TranscriptionRequest: {
             asr_model: string;
             diarize?: boolean;
@@ -2258,6 +2460,11 @@ export interface components {
              */
             sync?: boolean;
         };
+        DiarizationRequest: {
+            input_key: string;
+            session_id?: string;
+            transcription_job_id?: string;
+        };
         TextResponse: {
             cache: components["schemas"]["TextCacheStatuses"];
             context: components["schemas"]["TextExecutionContext"];

package/package.json

@@ -1,14 +1,14 @@
 {
   "name": "@simonpeacocks/peacockery-voice-client",
-  "version": "0.1.4",
+  "version": "0.3.0",
   "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
   }
 }