@teamlearners/clawops 0.16.3 → 0.16.5

package/dist/agent/index.d.cts

@@ -506,6 +506,14 @@ declare class ClawOpsAgent {
     private _handleRinging;
     private _handleFailed;
     private _onDtmfEvent;
+    /**
+     * _startCallSession 의 예외를 잡아 control WS 로 call.session_failed 전송한다.
+     *
+     * OpenAI/Gemini API 키 누락 등 session.start() 단계 실패는 media WS connect 에
+     * 도달하지 못해 call-engine 이 30 초간 무음 통화를 유지하게 만든다. 서버에 즉시
+     * 알려서 fail-fast 시키고 _activeSessions 에서 정리한다.
+     */
+    private _safeStartCallSession;
     private _startCallSession;
 }
 

package/dist/agent/index.d.ts

@@ -506,6 +506,14 @@ declare class ClawOpsAgent {
     private _handleRinging;
     private _handleFailed;
     private _onDtmfEvent;
+    /**
+     * _startCallSession 의 예외를 잡아 control WS 로 call.session_failed 전송한다.
+     *
+     * OpenAI/Gemini API 키 누락 등 session.start() 단계 실패는 media WS connect 에
+     * 도달하지 못해 call-engine 이 30 초간 무음 통화를 유지하게 만든다. 서버에 즉시
+     * 알려서 fail-fast 시키고 _activeSessions 에서 정리한다.
+     */
+    private _safeStartCallSession;
     private _startCallSession;
 }
 

package/dist/agent/index.js

@@ -1,4 +1,4 @@
-import { DEFAULT_BASE_URL, AgentError, AgentConnectionError, VERSION } from '../chunk-5IYJOP2D.js';
+import { DEFAULT_BASE_URL, AgentError, AgentConnectionError, VERSION } from '../chunk-BHGDB52N.js';
 import pino from 'pino';
 import * as fs from 'fs';
 import * as path from 'path';
@@ -1928,9 +1928,7 @@ var ClawOpsAgent = class _ClawOpsAgent {
       this._controlWs.send({ event: "call.accept", callId });
     }
     if (mediaUrl) {
-      this._startCallSession(session, mediaUrl).catch((err) => {
-        this._log.error({ err }, "Call session error: %s", callId);
-      });
+      this._safeStartCallSession(session, mediaUrl, callId);
     }
   }
   _handleEnded(event) {
@@ -1964,9 +1962,7 @@ var ClawOpsAgent = class _ClawOpsAgent {
     }
     if (mediaUrl) {
       this._log.info("Outbound call answered: %s -> %s (%s)", this._fromNumber, session.toNumber, callId);
-      this._startCallSession(session, mediaUrl).catch((err) => {
-        this._log.error({ err }, "Call session error: %s", callId);
-      });
+      this._safeStartCallSession(session, mediaUrl, callId);
     }
   }
   _handleRinging(event) {
@@ -2010,6 +2006,32 @@ var ClawOpsAgent = class _ClawOpsAgent {
       }
     }, this._passiveDtmfDebounceMs);
   }
+  /**
+   * _startCallSession 의 예외를 잡아 control WS 로 call.session_failed 전송한다.
+   *
+   * OpenAI/Gemini API 키 누락 등 session.start() 단계 실패는 media WS connect 에
+   * 도달하지 못해 call-engine 이 30 초간 무음 통화를 유지하게 만든다. 서버에 즉시
+   * 알려서 fail-fast 시키고 _activeSessions 에서 정리한다.
+   */
+  _safeStartCallSession(session, mediaWsUrl, callId) {
+    this._startCallSession(session, mediaWsUrl).catch((err) => {
+      const error = err;
+      this._log.error({ err }, "Session start failed for %s", callId);
+      if (this._controlWs) {
+        try {
+          this._controlWs.send({
+            event: "call.session_failed",
+            callId,
+            reason: error?.name ?? "Error",
+            message: error?.message ?? String(err)
+          });
+        } catch {
+        }
+      }
+      this._activeSessions.delete(callId);
+      this._callSessions.delete(callId);
+    });
+  }
   async _startCallSession(session, mediaWsUrl) {
     await withSpan(
       "clawops.call_session",
@@ -2633,6 +2655,11 @@ var OpenAIRealtime = class {
   _holdAudioChunks = null;
   constructor(options = {}) {
     this._apiKey = options.apiKey ?? process.env["OPENAI_API_KEY"] ?? "";
+    if (!this._apiKey) {
+      throw new Error(
+        "OpenAI API key is required. Set OPENAI_API_KEY env var or pass apiKey option."
+      );
+    }
     this._systemPrompt = options.systemPrompt ?? "";
     this._model = options.model ?? "gpt-realtime-2";
     this._voice = options.voice ?? "marin";
@@ -2658,15 +2685,11 @@ var OpenAIRealtime = class {
     this._closed = false;
     this._playback = null;
     this._latestMediaTs = 0;
-    if (!this._apiKey) {
-      throw new Error("OpenAI API key is required. Set OPENAI_API_KEY or pass apiKey option.");
-    }
     const { WebSocket } = await import('ws');
     const url = `${OPENAI_REALTIME_URL}${this._model}`;
     this._ws = new WebSocket(url, {
       headers: {
-        Authorization: `Bearer ${this._apiKey}`,
-        "OpenAI-Beta": "realtime=v1"
+        Authorization: `Bearer ${this._apiKey}`
       }
     });
     return new Promise((resolve, reject) => {
@@ -2736,17 +2759,24 @@ var OpenAIRealtime = class {
     this._send({
       type: "session.update",
       session: {
-        modalities: ["text", "audio"],
-        voice: this._voice,
+        type: "realtime",
+        output_modalities: ["audio"],
         instructions: this._systemPrompt,
-        input_audio_format: "g711_ulaw",
-        output_audio_format: "g711_ulaw",
-        input_audio_transcription: {
-          model: "whisper-1",
-          language: this._language
+        audio: {
+          input: {
+            format: { type: "audio/pcmu" },
+            noise_reduction: { type: "far_field" },
+            transcription: {
+              model: "gpt-realtime-whisper",
+              language: this._language
+            },
+            turn_detection: this._turnDetection
+          },
+          output: {
+            format: { type: "audio/pcmu" },
+            voice: this._voice
+          }
         },
-        input_audio_noise_reduction: { type: "far_field" },
-        turn_detection: this._turnDetection,
         tools: toolSchemas
       }
     });
@@ -2754,11 +2784,11 @@ var OpenAIRealtime = class {
   _handleMessage(msg) {
     const type = msg["type"];
     switch (type) {
-      case "response.audio.delta": {
+      case "response.output_audio.delta": {
         this._handleAudioDelta(msg);
         break;
       }
-      case "response.audio.done": {
+      case "response.output_audio.done": {
         if (this._playback) {
           this._playback.generating = false;
           if (this._playback.audioRemainder.length > 0) {
@@ -2792,7 +2822,7 @@ var OpenAIRealtime = class {
         }
         break;
       }
-      case "response.audio_transcript.done": {
+      case "response.output_audio_transcript.done": {
         if (this._call) {
           this._call._emit("transcript", "assistant", msg["transcript"] ?? "");
         }