voice-router-dev 0.2.1 → 0.2.3

package/dist/index.d.mts

@@ -4978,9 +4978,6 @@ declare const StreamingSupportedBitDepthEnum: {
 /**
  * ListenV1EncodingParameter type definition
  */
-/**
- * ListenV1EncodingParameter type definition
- */
 type ListenV1EncodingParameter = typeof ListenV1EncodingParameter[keyof typeof ListenV1EncodingParameter];
 declare const ListenV1EncodingParameter: {
     readonly linear16: "linear16";
@@ -5412,9 +5409,6 @@ declare function createVoiceRouter(config: VoiceRouterConfig, adapters?: Transcr
 /**
  * SpeakV1EncodingParameter type definition
  */
-/**
- * SpeakV1EncodingParameter type definition
- */
 type SpeakV1EncodingParameter = typeof SpeakV1EncodingParameter[keyof typeof SpeakV1EncodingParameter];
 declare const SpeakV1EncodingParameter: {
     readonly linear16: "linear16";
@@ -5452,9 +5446,6 @@ declare const SpeakV1EncodingParameter: {
 /**
  * SpeakV1ContainerParameter type definition
  */
-/**
- * SpeakV1ContainerParameter type definition
- */
 type SpeakV1ContainerParameter = typeof SpeakV1ContainerParameter[keyof typeof SpeakV1ContainerParameter];
 declare const SpeakV1ContainerParameter: {
     readonly none: "none";
@@ -5491,9 +5482,6 @@ declare const SpeakV1ContainerParameter: {
 /**
  * SpeakV1SampleRateParameter type definition
  */
-/**
- * SpeakV1SampleRateParameter type definition
- */
 type SpeakV1SampleRateParameter = typeof SpeakV1SampleRateParameter[keyof typeof SpeakV1SampleRateParameter];
 declare const SpeakV1SampleRateParameter: {
     readonly NUMBER_16000: 16000;
@@ -6740,8 +6728,7 @@ declare abstract class BaseWebhookHandler {
  * Gladia webhook handler
  *
  * Handles webhook callbacks from Gladia API:
- * - transcription.created - Job created and queued
- * - transcription.success - Job completed successfully
+ * - transcription.success - Job completed successfully (includes full transcript)
  * - transcription.error - Job failed with error
  *
  * @example
@@ -6763,13 +6750,26 @@ declare abstract class BaseWebhookHandler {
  *
  * if (event.eventType === 'transcription.completed') {
  *   console.log('Transcript:', event.data?.text);
+ *   console.log('Utterances:', event.data?.utterances);
  * }
  * ```
  */
 declare class GladiaWebhookHandler extends BaseWebhookHandler {
     readonly provider: TranscriptionProvider;
+    /**
+     * Convert Gladia WordDTO to unified Word type
+     */
+    private mapWord;
+    /**
+     * Convert Gladia UtteranceDTO to unified Utterance type
+     */
+    private mapUtterance;
     /**
      * Check if payload matches Gladia webhook format
+     *
+     * Gladia callbacks have the structure:
+     * - { id, event: "transcription.success", payload: TranscriptionResultDTO, custom_metadata? }
+     * - { id, event: "transcription.error", error: ErrorDTO, custom_metadata? }
      */
     matches(payload: unknown, _options?: {
         queryParams?: Record<string, string>;

package/dist/index.d.ts

@@ -4978,9 +4978,6 @@ declare const StreamingSupportedBitDepthEnum: {
 /**
  * ListenV1EncodingParameter type definition
  */
-/**
- * ListenV1EncodingParameter type definition
- */
 type ListenV1EncodingParameter = typeof ListenV1EncodingParameter[keyof typeof ListenV1EncodingParameter];
 declare const ListenV1EncodingParameter: {
     readonly linear16: "linear16";
@@ -5412,9 +5409,6 @@ declare function createVoiceRouter(config: VoiceRouterConfig, adapters?: Transcr
 /**
  * SpeakV1EncodingParameter type definition
  */
-/**
- * SpeakV1EncodingParameter type definition
- */
 type SpeakV1EncodingParameter = typeof SpeakV1EncodingParameter[keyof typeof SpeakV1EncodingParameter];
 declare const SpeakV1EncodingParameter: {
     readonly linear16: "linear16";
@@ -5452,9 +5446,6 @@ declare const SpeakV1EncodingParameter: {
 /**
  * SpeakV1ContainerParameter type definition
  */
-/**
- * SpeakV1ContainerParameter type definition
- */
 type SpeakV1ContainerParameter = typeof SpeakV1ContainerParameter[keyof typeof SpeakV1ContainerParameter];
 declare const SpeakV1ContainerParameter: {
     readonly none: "none";
@@ -5491,9 +5482,6 @@ declare const SpeakV1ContainerParameter: {
 /**
  * SpeakV1SampleRateParameter type definition
  */
-/**
- * SpeakV1SampleRateParameter type definition
- */
 type SpeakV1SampleRateParameter = typeof SpeakV1SampleRateParameter[keyof typeof SpeakV1SampleRateParameter];
 declare const SpeakV1SampleRateParameter: {
     readonly NUMBER_16000: 16000;
@@ -6740,8 +6728,7 @@ declare abstract class BaseWebhookHandler {
  * Gladia webhook handler
  *
  * Handles webhook callbacks from Gladia API:
- * - transcription.created - Job created and queued
- * - transcription.success - Job completed successfully
+ * - transcription.success - Job completed successfully (includes full transcript)
  * - transcription.error - Job failed with error
  *
  * @example
@@ -6763,13 +6750,26 @@ declare abstract class BaseWebhookHandler {
  *
  * if (event.eventType === 'transcription.completed') {
  *   console.log('Transcript:', event.data?.text);
+ *   console.log('Utterances:', event.data?.utterances);
  * }
  * ```
  */
 declare class GladiaWebhookHandler extends BaseWebhookHandler {
     readonly provider: TranscriptionProvider;
+    /**
+     * Convert Gladia WordDTO to unified Word type
+     */
+    private mapWord;
+    /**
+     * Convert Gladia UtteranceDTO to unified Utterance type
+     */
+    private mapUtterance;
     /**
      * Check if payload matches Gladia webhook format
+     *
+     * Gladia callbacks have the structure:
+     * - { id, event: "transcription.success", payload: TranscriptionResultDTO, custom_metadata? }
+     * - { id, event: "transcription.error", error: ErrorDTO, custom_metadata? }
      */
     matches(payload: unknown, _options?: {
         queryParams?: Record<string, string>;

package/dist/index.js

@@ -2896,6 +2896,15 @@ var AssemblyAIAdapter = class extends BaseAdapter {
     });
     let sessionStatus = "connecting";
     const sessionId = `assemblyai-${Date.now()}-${Math.random().toString(36).substring(7)}`;
+    let audioBuffer = Buffer.alloc(0);
+    const MIN_CHUNK_SIZE = 1600;
+    const MAX_CHUNK_SIZE = 32e3;
+    const flushAudioBuffer = () => {
+      if (audioBuffer.length > 0 && ws.readyState === import_ws2.default.OPEN) {
+        ws.send(audioBuffer);
+        audioBuffer = Buffer.alloc(0);
+      }
+    };
     ws.on("open", () => {
       sessionStatus = "open";
       callbacks?.onOpen?.();
@@ -2984,8 +2993,13 @@ var AssemblyAIAdapter = class extends BaseAdapter {
         if (ws.readyState !== import_ws2.default.OPEN) {
           throw new Error("WebSocket is not open");
         }
-        ws.send(chunk.data);
+        audioBuffer = Buffer.concat([audioBuffer, chunk.data]);
+        if (audioBuffer.length >= MIN_CHUNK_SIZE || audioBuffer.length >= MAX_CHUNK_SIZE) {
+          ws.send(audioBuffer);
+          audioBuffer = Buffer.alloc(0);
+        }
         if (chunk.isLast) {
+          flushAudioBuffer();
           ws.send(
             JSON.stringify({
               terminate_session: true
@@ -2998,6 +3012,7 @@ var AssemblyAIAdapter = class extends BaseAdapter {
           return;
         }
         sessionStatus = "closing";
+        flushAudioBuffer();
         if (ws.readyState === import_ws2.default.OPEN) {
           ws.send(
             JSON.stringify({
@@ -4338,15 +4353,46 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
     super(...arguments);
     this.provider = "gladia";
   }
+  /**
+   * Convert Gladia WordDTO to unified Word type
+   */
+  mapWord(word) {
+    return {
+      text: word.word,
+      start: word.start,
+      end: word.end,
+      confidence: word.confidence
+    };
+  }
+  /**
+   * Convert Gladia UtteranceDTO to unified Utterance type
+   */
+  mapUtterance(utterance) {
+    return {
+      text: utterance.text,
+      start: utterance.start,
+      end: utterance.end,
+      confidence: utterance.confidence,
+      speaker: utterance.speaker !== void 0 ? String(utterance.speaker) : void 0,
+      words: utterance.words?.map((w) => this.mapWord(w))
+    };
+  }
   /**
    * Check if payload matches Gladia webhook format
+   *
+   * Gladia callbacks have the structure:
+   * - { id, event: "transcription.success", payload: TranscriptionResultDTO, custom_metadata? }
+   * - { id, event: "transcription.error", error: ErrorDTO, custom_metadata? }
    */
   matches(payload, _options) {
     if (!payload || typeof payload !== "object") {
       return false;
     }
     const obj = payload;
-    if (!("event" in obj) || !("payload" in obj)) {
+    if (!("id" in obj) || !("event" in obj)) {
+      return false;
+    }
+    if (typeof obj.id !== "string") {
       return false;
     }
     if (typeof obj.event !== "string") {
@@ -4355,11 +4401,13 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
     if (!obj.event.startsWith("transcription.")) {
       return false;
     }
-    if (!obj.payload || typeof obj.payload !== "object") {
+    if (obj.event === "transcription.success" && !("payload" in obj)) {
+      return false;
+    }
+    if (obj.event === "transcription.error" && !("error" in obj)) {
       return false;
     }
-    const payloadObj = obj.payload;
-    return typeof payloadObj.id === "string";
+    return true;
   }
   /**
    * Parse Gladia webhook payload to unified format
@@ -4368,38 +4416,57 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
     if (!this.matches(payload)) {
       return this.createErrorEvent(payload, "Invalid Gladia webhook payload");
     }
-    const webhookPayload = payload;
-    const jobId = webhookPayload.payload.id;
-    const event = webhookPayload.event;
-    if (event === "transcription.created") {
-      return {
-        success: true,
-        provider: this.provider,
-        eventType: "transcription.created",
-        data: {
-          id: jobId,
-          status: "queued"
-        },
-        timestamp: (/* @__PURE__ */ new Date()).toISOString(),
-        raw: payload
-      };
-    }
+    const obj = payload;
+    const jobId = obj.id;
+    const event = obj.event;
     if (event === "transcription.success") {
+      const successPayload = payload;
+      const result = successPayload.payload;
+      const transcription = result.transcription;
+      const metadata = result.metadata;
+      const utterances = transcription?.utterances?.map(
+        (u) => this.mapUtterance(u)
+      );
+      const words = transcription?.utterances?.flatMap(
+        (u) => u.words?.map((w) => this.mapWord(w)) ?? []
+      );
+      const speakerIds = /* @__PURE__ */ new Set();
+      transcription?.utterances?.forEach((u) => {
+        if (u.speaker !== void 0) {
+          speakerIds.add(u.speaker);
+        }
+      });
+      const speakers = speakerIds.size > 0 ? Array.from(speakerIds).map((id) => ({ id: String(id) })) : void 0;
+      const summary = result.summarization?.success && result.summarization.results ? result.summarization.results : void 0;
       return {
         success: true,
         provider: this.provider,
         eventType: "transcription.completed",
         data: {
           id: jobId,
-          status: "completed"
-          // Note: Full transcript data needs to be fetched via API
-          // using GladiaAdapter.getTranscript(jobId)
+          status: "completed",
+          text: transcription?.full_transcript,
+          duration: metadata?.audio_duration,
+          language: transcription?.languages?.[0],
+          speakers,
+          words,
+          utterances,
+          summary,
+          metadata: {
+            transcription_time: metadata?.transcription_time,
+            billing_time: metadata?.billing_time,
+            number_of_distinct_channels: metadata?.number_of_distinct_channels,
+            custom_metadata: successPayload.custom_metadata
+          },
+          completedAt: (/* @__PURE__ */ new Date()).toISOString()
         },
         timestamp: (/* @__PURE__ */ new Date()).toISOString(),
         raw: payload
       };
     }
     if (event === "transcription.error") {
+      const errorPayload = payload;
+      const error = errorPayload.error;
       return {
         success: false,
         provider: this.provider,
@@ -4407,7 +4474,11 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
         data: {
           id: jobId,
           status: "error",
-          error: "Transcription failed"
+          error: error?.message || "Transcription failed",
+          metadata: {
+            error_code: error?.code,
+            custom_metadata: errorPayload.custom_metadata
+          }
         },
         timestamp: (/* @__PURE__ */ new Date()).toISOString(),
         raw: payload