voice-router-dev 0.7.9 → 0.8.1

package/dist/webhooks.js

@@ -35,12 +35,14 @@ __export(webhooks_exports, {
   AzureWebhookHandler: () => AzureWebhookHandler,
   BaseWebhookHandler: () => BaseWebhookHandler,
   DeepgramWebhookHandler: () => DeepgramWebhookHandler,
+  ElevenLabsWebhookHandler: () => ElevenLabsWebhookHandler,
   GladiaWebhookHandler: () => GladiaWebhookHandler,
   SpeechmaticsWebhookHandler: () => SpeechmaticsWebhookHandler,
   WebhookRouter: () => WebhookRouter,
   createAssemblyAIWebhookHandler: () => createAssemblyAIWebhookHandler,
   createAzureWebhookHandler: () => createAzureWebhookHandler,
   createDeepgramWebhookHandler: () => createDeepgramWebhookHandler,
+  createElevenLabsWebhookHandler: () => createElevenLabsWebhookHandler,
   createGladiaWebhookHandler: () => createGladiaWebhookHandler,
   createWebhookRouter: () => createWebhookRouter
 });
@@ -134,7 +136,7 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
       end: utterance.end,
       confidence: utterance.confidence,
       speaker: utterance.speaker !== void 0 ? String(utterance.speaker) : void 0,
-      words: utterance.words?.map((w) => this.mapWord(w))
+      words: utterance.words?.map((w) => this.mapWord(w)) ?? []
     };
   }
   /**
@@ -196,7 +198,7 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
           speakerIds.add(u.speaker);
         }
       });
-      const speakers = speakerIds.size > 0 ? Array.from(speakerIds).map((id) => ({ id: String(id) })) : void 0;
+      const speakers = speakerIds.size > 0 ? Array.from(speakerIds).map((id) => ({ id: String(id), label: `Speaker ${id}` })) : void 0;
       const summary = result.summarization?.success && result.summarization.results ? result.summarization.results : void 0;
       return {
         success: true,
@@ -274,34 +276,68 @@ var AssemblyAIWebhookHandler = class extends BaseWebhookHandler {
   }
   /**
    * Check if payload matches AssemblyAI webhook format
+   *
+   * Supports two formats:
+   * - Notification format: `{ transcript_id, status }` (lightweight callback)
+   * - Full transcript format: `{ id, status, audio_url, text, words, ... }` (complete response)
    */
   matches(payload, _options) {
     if (!payload || typeof payload !== "object") {
       return false;
     }
     const obj = payload;
-    if (!("transcript_id" in obj) || !("status" in obj)) {
-      return false;
-    }
-    if (typeof obj.transcript_id !== "string") {
-      return false;
+    if ("transcript_id" in obj && "status" in obj) {
+      if (typeof obj.transcript_id !== "string") return false;
+      if (obj.status !== "completed" && obj.status !== "error") return false;
+      return true;
     }
-    if (obj.status !== "completed" && obj.status !== "error") {
-      return false;
+    if ("id" in obj && "status" in obj && "audio_url" in obj) {
+      if (typeof obj.id !== "string") return false;
+      if (obj.status !== "completed" && obj.status !== "error") return false;
+      return true;
     }
-    return true;
+    return false;
+  }
+  /**
+   * Determine if the payload is a full transcript (vs a lightweight notification)
+   */
+  isFullTranscript(payload) {
+    return "audio_url" in payload && "id" in payload;
   }
   /**
    * Parse AssemblyAI webhook payload to unified format
+   *
+   * Supports two payload formats:
+   * - Notification: `{ transcript_id, status }` — returns minimal event (ID + status only)
+   * - Full transcript: `{ id, status, text, words, utterances, ... }` — returns complete data
    */
   parse(payload, _options) {
     if (!this.matches(payload)) {
       return this.createErrorEvent(payload, "Invalid AssemblyAI webhook payload");
     }
-    const notification = payload;
-    const transcriptId = notification.transcript_id;
-    const status = notification.status;
+    const obj = payload;
+    const isFullFormat = this.isFullTranscript(obj);
+    const transcriptId = isFullFormat ? payload.id : payload.transcript_id;
+    const status = obj.status;
+    if (status === "error") {
+      const error = isFullFormat ? payload.error : void 0;
+      return {
+        success: false,
+        provider: this.provider,
+        eventType: "transcription.failed",
+        data: {
+          id: transcriptId,
+          status: "error",
+          error: error || "Transcription failed"
+        },
+        timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+        raw: payload
+      };
+    }
     if (status === "completed") {
+      if (isFullFormat) {
+        return this.parseFullTranscript(payload, payload);
+      }
       return {
         success: true,
         provider: this.provider,
@@ -309,28 +345,76 @@ var AssemblyAIWebhookHandler = class extends BaseWebhookHandler {
         data: {
           id: transcriptId,
           status: "completed"
-          // Note: Full transcript data needs to be fetched via API
-          // using AssemblyAIAdapter.getTranscript(transcriptId)
         },
         timestamp: (/* @__PURE__ */ new Date()).toISOString(),
         raw: payload
       };
     }
-    if (status === "error") {
+    return this.createErrorEvent(payload, `Unknown AssemblyAI status: ${status}`);
+  }
+  /**
+   * Parse a full AssemblyAI transcript response into unified format
+   *
+   * AssemblyAI times are in milliseconds — converted to seconds for unified format.
+   */
+  parseFullTranscript(transcript, raw) {
+    try {
+      const words = transcript.words ? transcript.words.map((w) => ({
+        word: w.text,
+        start: w.start / 1e3,
+        end: w.end / 1e3,
+        confidence: w.confidence,
+        speaker: w.speaker ?? void 0
+      })) : void 0;
+      const utterances = transcript.utterances ? transcript.utterances.map((u) => ({
+        text: u.text,
+        start: u.start / 1e3,
+        end: u.end / 1e3,
+        speaker: u.speaker,
+        confidence: u.confidence,
+        words: u.words.map((w) => ({
+          word: w.text,
+          start: w.start / 1e3,
+          end: w.end / 1e3,
+          confidence: w.confidence,
+          speaker: w.speaker ?? void 0
+        }))
+      })) : void 0;
+      const speakerIds = /* @__PURE__ */ new Set();
+      transcript.utterances?.forEach((u) => {
+        if (u.speaker) speakerIds.add(u.speaker);
+      });
+      const speakers = speakerIds.size > 0 ? Array.from(speakerIds).map((id) => ({ id, label: `Speaker ${id}` })) : void 0;
       return {
-        success: false,
+        success: true,
         provider: this.provider,
-        eventType: "transcription.failed",
+        eventType: "transcription.completed",
         data: {
-          id: transcriptId,
-          status: "error",
-          error: "Transcription failed"
+          id: transcript.id,
+          status: "completed",
+          text: transcript.text ?? void 0,
+          confidence: transcript.confidence ?? void 0,
+          duration: transcript.audio_duration ?? void 0,
+          language: transcript.language_code ?? void 0,
+          speakers,
+          words,
+          utterances,
+          summary: transcript.summary ?? void 0,
+          metadata: {
+            speech_model: transcript.speech_model,
+            audio_channels: transcript.audio_channels,
+            webhook_status_code: transcript.webhook_status_code
+          }
         },
         timestamp: (/* @__PURE__ */ new Date()).toISOString(),
-        raw: payload
+        raw
       };
+    } catch (error) {
+      return this.createErrorEvent(
+        raw,
+        `Failed to parse AssemblyAI transcript: ${error instanceof Error ? error.message : "Unknown error"}`
+      );
     }
-    return this.createErrorEvent(payload, `Unknown AssemblyAI status: ${status}`);
   }
   /**
    * Verify AssemblyAI webhook signature
@@ -465,24 +549,33 @@ var DeepgramWebhookHandler = class extends BaseWebhookHandler {
         end: w.end || 0,
         confidence: w.confidence
       })) : void 0;
-      const speakers = response.results.utterances && response.results.utterances.length > 0 ? response.results.utterances.map((utterance) => ({
-        id: utterance.speaker?.toString() || "unknown",
-        speaker: utterance.speaker?.toString() || "unknown",
-        text: utterance.transcript || "",
-        confidence: utterance.confidence
+      const speakerIds = /* @__PURE__ */ new Set();
+      if (response.results.utterances) {
+        for (const utterance of response.results.utterances) {
+          if (utterance.speaker !== void 0) {
+            speakerIds.add(utterance.speaker.toString());
+          }
+        }
+      }
+      const speakers = speakerIds.size > 0 ? Array.from(speakerIds).map((id) => ({
+        id,
+        label: `Speaker ${id}`
       })) : void 0;
       const utterances = response.results.utterances && response.results.utterances.length > 0 ? response.results.utterances.map((utterance) => ({
+        id: utterance.id,
         text: utterance.transcript || "",
         start: utterance.start || 0,
         end: utterance.end || 0,
         speaker: utterance.speaker?.toString(),
+        channel: utterance.channel,
         confidence: utterance.confidence,
-        words: utterance.words && utterance.words.length > 0 ? utterance.words.map((w) => ({
+        words: utterance.words?.map((w) => ({
           word: w.word || "",
           start: w.start || 0,
           end: w.end || 0,
-          confidence: w.confidence
-        })) : void 0
+          confidence: w.confidence,
+          speaker: w.speaker?.toString()
+        })) ?? []
       })) : void 0;
       const summary = alternative.summaries?.[0]?.summary;
       return {
@@ -495,7 +588,7 @@ var DeepgramWebhookHandler = class extends BaseWebhookHandler {
           text: transcript,
           confidence: alternative.confidence,
           duration,
-          language: response.metadata.models?.[0] || void 0,
+          language: channel.detected_language,
           speakers: speakers && speakers.length > 0 ? speakers : void 0,
           words: words && words.length > 0 ? words : void 0,
           utterances: utterances && utterances.length > 0 ? utterances : void 0,
@@ -682,6 +775,72 @@ function createAzureWebhookHandler() {
   return new AzureWebhookHandler();
 }
 
+// src/utils/transcription-helpers.ts
+function buildUtterancesFromWords(words) {
+  const utterances = [];
+  let currentSpeaker;
+  let currentWords = [];
+  let utteranceStart = 0;
+  for (const word of words) {
+    if (!word.speaker) continue;
+    if (word.speaker !== currentSpeaker) {
+      if (currentSpeaker && currentWords.length > 0) {
+        utterances.push({
+          text: currentWords.map((w) => w.word).join(" "),
+          start: utteranceStart,
+          end: currentWords[currentWords.length - 1].end,
+          speaker: currentSpeaker,
+          words: currentWords
+        });
+      }
+      currentSpeaker = word.speaker;
+      currentWords = [word];
+      utteranceStart = word.start;
+    } else {
+      currentWords.push(word);
+    }
+  }
+  if (currentSpeaker && currentWords.length > 0) {
+    utterances.push({
+      text: currentWords.map((w) => w.word).join(" "),
+      start: utteranceStart,
+      end: currentWords[currentWords.length - 1].end,
+      speaker: currentSpeaker,
+      words: currentWords
+    });
+  }
+  return utterances;
+}
+function buildTextFromSpeechmaticsResults(results) {
+  const parts = [];
+  let attachNext = false;
+  for (const result of results) {
+    if (result.type !== "word" && result.type !== "punctuation") continue;
+    const content = result.alternatives?.[0]?.content;
+    if (!content) continue;
+    if (result.type === "punctuation") {
+      const attaches = result.attaches_to;
+      if (attaches === "previous" || attaches === "both") {
+        parts.push(content);
+        attachNext = attaches === "both";
+      } else if (attaches === "next") {
+        if (parts.length > 0) parts.push(" ");
+        parts.push(content);
+        attachNext = true;
+      } else {
+        if (parts.length > 0 && !attachNext) parts.push(" ");
+        parts.push(content);
+        attachNext = false;
+      }
+    } else {
+      if (parts.length > 0 && !attachNext) parts.push(" ");
+      parts.push(content);
+      attachNext = false;
+    }
+  }
+  return parts.join("");
+}
+
 // src/webhooks/speechmatics-webhook.ts
 var SpeechmaticsWebhookHandler = class extends BaseWebhookHandler {
   constructor() {
@@ -763,18 +922,25 @@ var SpeechmaticsWebhookHandler = class extends BaseWebhookHandler {
     if (status === "success" && payload && typeof payload === "object") {
       const transcript = payload;
       if (transcript.results && transcript.job) {
-        const text = transcript.results.filter((r) => r.type === "word" && r.alternatives).map((r) => r.alternatives[0]?.content || "").join(" ");
+        const text = buildTextFromSpeechmaticsResults(transcript.results);
+        const wordResults = transcript.results.filter((r) => r.type === "word" && r.alternatives);
+        const words = wordResults.filter((r) => r.start_time !== void 0 && r.end_time !== void 0).map((r) => ({
+          word: r.alternatives[0]?.content || "",
+          start: r.start_time,
+          end: r.end_time,
+          confidence: r.alternatives[0]?.confidence,
+          speaker: r.alternatives[0]?.speaker
+        }));
         const speakerSet = /* @__PURE__ */ new Set();
-        transcript.results.forEach((r) => {
-          if (r.alternatives) {
-            const speaker = r.alternatives[0]?.speaker;
-            if (speaker) speakerSet.add(speaker);
-          }
+        wordResults.forEach((r) => {
+          const speaker = r.alternatives[0]?.speaker;
+          if (speaker) speakerSet.add(speaker);
         });
         const speakers = speakerSet.size > 0 ? Array.from(speakerSet).map((id) => ({
           id,
           label: `Speaker ${id}`
         })) : void 0;
+        const utterances = buildUtterancesFromWords(words);
         return {
           success: true,
           provider: this.provider,
@@ -787,6 +953,8 @@ var SpeechmaticsWebhookHandler = class extends BaseWebhookHandler {
             language: transcript.metadata.transcription_config?.language,
             duration: transcript.job.duration,
             speakers,
+            words: words.length > 0 ? words : void 0,
+            utterances: utterances.length > 0 ? utterances : void 0,
             createdAt: transcript.job.created_at
           },
           raw: payload
@@ -808,6 +976,138 @@ var SpeechmaticsWebhookHandler = class extends BaseWebhookHandler {
   }
 };
 
+// src/webhooks/elevenlabs-webhook.ts
+var ElevenLabsWebhookHandler = class extends BaseWebhookHandler {
+  constructor() {
+    super(...arguments);
+    this.provider = "elevenlabs";
+  }
+  /**
+   * Check if payload matches ElevenLabs webhook format
+   *
+   * ElevenLabs webhook payloads contain the full transcription result
+   * with `words` array and `language_code` / `language_probability` fields.
+   */
+  matches(payload, _options) {
+    if (!payload || typeof payload !== "object") {
+      return false;
+    }
+    const obj = payload;
+    if (!("words" in obj) || !("language_code" in obj) || !("language_probability" in obj)) {
+      return false;
+    }
+    if (!Array.isArray(obj.words)) {
+      return false;
+    }
+    if (!("text" in obj)) {
+      return false;
+    }
+    if (obj.words.length > 0) {
+      const firstWord = obj.words[0];
+      if (!("logprob" in firstWord) || !("type" in firstWord)) {
+        return false;
+      }
+    }
+    return true;
+  }
+  /**
+   * Parse ElevenLabs webhook payload to unified format
+   */
+  parse(payload, _options) {
+    if (!this.matches(payload)) {
+      return this.createErrorEvent(payload, "Invalid ElevenLabs webhook payload");
+    }
+    const response = payload;
+    try {
+      const transcriptionId = response.transcription_id?.toString() || "";
+      const transcript = response.text;
+      if (!transcript) {
+        return {
+          success: false,
+          provider: this.provider,
+          eventType: "transcription.failed",
+          data: {
+            id: transcriptionId,
+            status: "error",
+            error: "Empty transcript"
+          },
+          timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+          raw: payload
+        };
+      }
+      const words = response.words && response.words.length > 0 ? response.words.filter((w) => w.type === "word").map((w) => ({
+        word: w.text || "",
+        start: typeof w.start === "number" ? w.start : 0,
+        end: typeof w.end === "number" ? w.end : 0,
+        confidence: w.logprob !== void 0 ? Math.exp(w.logprob) : void 0,
+        speaker: w.speaker_id?.toString()
+      })) : void 0;
+      const speakerIds = /* @__PURE__ */ new Set();
+      if (response.words) {
+        for (const w of response.words) {
+          if (w.speaker_id !== void 0 && w.speaker_id !== null) {
+            speakerIds.add(w.speaker_id.toString());
+          }
+        }
+      }
+      const speakers = speakerIds.size > 0 ? Array.from(speakerIds).map((id) => ({
+        id,
+        label: `Speaker ${id}`
+      })) : void 0;
+      const utterances = words && words.length > 0 ? buildUtterancesFromWords(
+        words.map((w) => ({
+          word: w.word,
+          start: w.start,
+          end: w.end,
+          confidence: w.confidence,
+          speaker: w.speaker
+        }))
+      ) : void 0;
+      return {
+        success: true,
+        provider: this.provider,
+        eventType: "transcription.completed",
+        data: {
+          id: transcriptionId,
+          status: "completed",
+          text: transcript,
+          language: response.language_code,
+          speakers: speakers && speakers.length > 0 ? speakers : void 0,
+          words: words && words.length > 0 ? words : void 0,
+          utterances: utterances && utterances.length > 0 ? utterances : void 0,
+          metadata: {
+            language_probability: response.language_probability,
+            entities: response.entities,
+            channel_index: response.channel_index
+          }
+        },
+        timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+        raw: payload
+      };
+    } catch (error) {
+      return this.createErrorEvent(
+        payload,
+        `Failed to parse ElevenLabs webhook: ${error instanceof Error ? error.message : "Unknown error"}`
+      );
+    }
+  }
+  /**
+   * Verify ElevenLabs webhook signature
+   *
+   * Note: ElevenLabs does not currently provide a standard webhook signature
+   * verification mechanism for STT webhooks. For security, use HTTPS and
+   * validate the request source.
+   *
+   * @returns Always returns true (no verification available)
+   */
+  verify() {
+    return true;
+  }
+};
+function createElevenLabsWebhookHandler() {
+  return new ElevenLabsWebhookHandler();
+}
+
 // src/webhooks/webhook-router.ts
 var WebhookRouter = class {
   constructor() {
@@ -816,7 +1116,8 @@ var WebhookRouter = class {
       ["assemblyai", new AssemblyAIWebhookHandler()],
       ["deepgram", new DeepgramWebhookHandler()],
       ["azure-stt", new AzureWebhookHandler()],
-      ["speechmatics", new SpeechmaticsWebhookHandler()]
+      ["speechmatics", new SpeechmaticsWebhookHandler()],
+      ["elevenlabs", new ElevenLabsWebhookHandler()]
     ]);
   }
   /**
@@ -996,12 +1297,14 @@ function createWebhookRouter() {
   AzureWebhookHandler,
   BaseWebhookHandler,
   DeepgramWebhookHandler,
+  ElevenLabsWebhookHandler,
   GladiaWebhookHandler,
   SpeechmaticsWebhookHandler,
   WebhookRouter,
   createAssemblyAIWebhookHandler,
   createAzureWebhookHandler,
   createDeepgramWebhookHandler,
+  createElevenLabsWebhookHandler,
   createGladiaWebhookHandler,
   createWebhookRouter
 });