getpatter 0.4.3 → 0.5.0

package/dist/{chunk-35EVXMGB.mjs → chunk-757NVN4L.mjs}

@@ -247,223 +247,13 @@ var ElevenLabsConvAIAdapter = class {
   }
 };
 
-// src/providers/deepgram-stt.ts
-import WebSocket3 from "ws";
-var DEEPGRAM_WS_URL = "wss://api.deepgram.com/v1/listen";
-var DeepgramSTT = class _DeepgramSTT {
-  constructor(apiKey, language = "en", model = "nova-3", encoding = "linear16", sampleRate = 16e3) {
-    this.apiKey = apiKey;
-    this.language = language;
-    this.model = model;
-    this.encoding = encoding;
-    this.sampleRate = sampleRate;
-  }
-  ws = null;
-  callbacks = [];
-  /** Request ID from Deepgram — used to query actual cost post-call. */
-  requestId = "";
-  /** Factory for Twilio calls — mulaw 8 kHz. */
-  static forTwilio(apiKey, language = "en", model = "nova-3") {
-    return new _DeepgramSTT(apiKey, language, model, "mulaw", 8e3);
-  }
-  async connect() {
-    const params = new URLSearchParams({
-      model: this.model,
-      language: this.language,
-      encoding: this.encoding,
-      sample_rate: String(this.sampleRate),
-      channels: "1",
-      interim_results: "true",
-      endpointing: "300",
-      smart_format: "true",
-      vad_events: "true",
-      no_delay: "true"
-    });
-    const url = `${DEEPGRAM_WS_URL}?${params.toString()}`;
-    this.ws = new WebSocket3(url, {
-      headers: { Authorization: `Token ${this.apiKey}` }
-    });
-    await new Promise((resolve, reject) => {
-      const timer = setTimeout(() => reject(new Error("Deepgram connect timeout")), 1e4);
-      this.ws.once("open", () => {
-        clearTimeout(timer);
-        resolve();
-      });
-      this.ws.once("error", (err) => {
-        clearTimeout(timer);
-        reject(err);
-      });
-    });
-    this.ws.on("message", (raw) => {
-      let data;
-      try {
-        data = JSON.parse(raw.toString());
-      } catch {
-        return;
-      }
-      if (data.type === "Metadata" && data.request_id) {
-        this.requestId = data.request_id;
-        return;
-      }
-      if (data.type !== "Results") return;
-      const alternatives = data.channel?.alternatives ?? [];
-      if (!alternatives.length) return;
-      const best = alternatives[0];
-      const text = (best.transcript ?? "").trim();
-      if (!text) return;
-      const transcript = {
-        text,
-        isFinal: Boolean(data.is_final) && Boolean(data.speech_final),
-        confidence: best.confidence ?? 0
-      };
-      for (const cb of this.callbacks) {
-        cb(transcript);
-      }
-    });
-  }
-  sendAudio(audio) {
-    if (!this.ws || this.ws.readyState !== WebSocket3.OPEN) return;
-    this.ws.send(audio);
-  }
-  onTranscript(callback) {
-    if (this.callbacks.length >= 10) {
-      getLogger().warn("DeepgramSTT: maximum of 10 onTranscript callbacks reached; replacing the last callback.");
-      this.callbacks[this.callbacks.length - 1] = callback;
-      return;
-    }
-    this.callbacks.push(callback);
-  }
-  close() {
-    if (this.ws) {
-      try {
-        this.ws.send(JSON.stringify({ type: "CloseStream" }));
-      } catch {
-      }
-      this.ws.close();
-      this.ws = null;
-    }
-  }
-};
-
-// src/providers/whisper-stt.ts
-var OPENAI_TRANSCRIPTION_URL = "https://api.openai.com/v1/audio/transcriptions";
-var DEFAULT_BUFFER_SIZE = 16e3 * 2;
-function wrapPcmInWav(pcm, sampleRate = 16e3, channels = 1, bitsPerSample = 16) {
-  const dataSize = pcm.length;
-  const header = Buffer.alloc(44);
-  header.write("RIFF", 0);
-  header.writeUInt32LE(36 + dataSize, 4);
-  header.write("WAVE", 8);
-  header.write("fmt ", 12);
-  header.writeUInt32LE(16, 16);
-  header.writeUInt16LE(1, 20);
-  header.writeUInt16LE(channels, 22);
-  header.writeUInt32LE(sampleRate, 24);
-  header.writeUInt32LE(sampleRate * channels * (bitsPerSample / 8), 28);
-  header.writeUInt16LE(channels * (bitsPerSample / 8), 32);
-  header.writeUInt16LE(bitsPerSample, 34);
-  header.write("data", 36);
-  header.writeUInt32LE(dataSize, 40);
-  return Buffer.concat([header, pcm]);
+// src/provider-factory.ts
+async function createSTT(agent) {
+  return agent.stt ?? null;
+}
+async function createTTS(agent) {
+  return agent.tts ?? null;
 }
-var WhisperSTT = class _WhisperSTT {
-  apiKey;
-  model;
-  language;
-  bufferSize;
-  buffer = Buffer.alloc(0);
-  callbacks = [];
-  running = false;
-  pendingTranscriptions = [];
-  constructor(apiKey, model = "whisper-1", language, bufferSize = DEFAULT_BUFFER_SIZE) {
-    this.apiKey = apiKey;
-    this.model = model;
-    this.language = language;
-    this.bufferSize = bufferSize;
-  }
-  /** Factory for Twilio calls — mulaw 8 kHz is transcoded upstream, so we still receive PCM 16-bit. */
-  static forTwilio(apiKey, language = "en", model = "whisper-1") {
-    return new _WhisperSTT(apiKey, model, language);
-  }
-  async connect() {
-    this.running = true;
-    this.buffer = Buffer.alloc(0);
-  }
-  sendAudio(audio) {
-    if (!this.running) return;
-    this.buffer = Buffer.concat([this.buffer, audio]);
-    if (this.buffer.length >= this.bufferSize) {
-      const pcm = this.buffer;
-      this.buffer = Buffer.alloc(0);
-      this.trackTranscription(this.transcribeBuffer(pcm));
-    }
-  }
-  trackTranscription(promise) {
-    const wrapped = promise.finally(() => {
-      const idx = this.pendingTranscriptions.indexOf(wrapped);
-      if (idx !== -1) this.pendingTranscriptions.splice(idx, 1);
-    });
-    this.pendingTranscriptions.push(wrapped);
-  }
-  onTranscript(callback) {
-    if (this.callbacks.length >= 10) {
-      getLogger().warn("WhisperSTT: maximum of 10 onTranscript callbacks reached; replacing the last callback.");
-      this.callbacks[this.callbacks.length - 1] = callback;
-      return;
-    }
-    this.callbacks.push(callback);
-  }
-  async close() {
-    this.running = false;
-    if (this.buffer.length >= this.bufferSize / 4) {
-      const pcm = this.buffer;
-      this.buffer = Buffer.alloc(0);
-      this.trackTranscription(this.transcribeBuffer(pcm));
-    } else {
-      this.buffer = Buffer.alloc(0);
-    }
-    await Promise.allSettled(this.pendingTranscriptions);
-    this.callbacks = [];
-  }
-  // ------------------------------------------------------------------
-  // Private
-  // ------------------------------------------------------------------
-  async transcribeBuffer(pcm) {
-    const wav = wrapPcmInWav(pcm);
-    const formData = new FormData();
-    formData.append("file", new Blob([wav.buffer.slice(wav.byteOffset, wav.byteOffset + wav.byteLength)], { type: "audio/wav" }), "audio.wav");
-    formData.append("model", this.model);
-    if (this.language) {
-      formData.append("language", this.language);
-    }
-    try {
-      const resp = await fetch(OPENAI_TRANSCRIPTION_URL, {
-        method: "POST",
-        headers: { Authorization: `Bearer ${this.apiKey}` },
-        body: formData,
-        signal: AbortSignal.timeout(15e3)
-      });
-      if (!resp.ok) {
-        const body = await resp.text();
-        getLogger().error(`WhisperSTT transcription error: ${resp.status} ${body}`);
-        return;
-      }
-      const json = await resp.json();
-      const text = (json.text ?? "").trim();
-      if (!text) return;
-      const transcript = {
-        text,
-        isFinal: true,
-        confidence: 1
-      };
-      for (const cb of this.callbacks) {
-        cb(transcript);
-      }
-    } catch (err) {
-      getLogger().error(`WhisperSTT transcription error: ${String(err)}`);
-    }
-  }
-};
 
 // src/pricing.ts
 var DEFAULT_PRICING = {
@@ -534,9 +324,15 @@ var MetricsStore = class extends EventEmitter {
   maxCalls;
   calls = [];
   activeCalls = /* @__PURE__ */ new Map();
-  constructor(maxCalls = 500) {
+  /**
+   * Accepts either a numeric ``maxCalls`` (legacy positional — matches the
+   * original TS API) or an options object ``{ maxCalls }`` to align with the
+   * Python SDK's keyword-argument style. Plain literals also work:
+   * ``new MetricsStore()`` / ``new MetricsStore(100)`` / ``new MetricsStore({ maxCalls: 100 })``.
+   */
+  constructor(maxCallsOrOpts = 500) {
     super();
-    this.maxCalls = maxCalls;
+    this.maxCalls = typeof maxCallsOrOpts === "number" ? maxCallsOrOpts : maxCallsOrOpts.maxCalls ?? 500;
   }
   publish(eventType, data) {
     this.emit("sse", { type: eventType, data });
@@ -544,22 +340,100 @@ var MetricsStore = class extends EventEmitter {
   recordCallStart(data) {
     const callId = data.call_id || "";
     if (!callId) return;
+    const existing = this.activeCalls.get(callId);
+    if (existing) {
+      existing.caller = data.caller || existing.caller;
+      existing.callee = data.callee || existing.callee;
+      existing.direction = data.direction || existing.direction;
+      existing.status = "in-progress";
+      existing.turns = existing.turns || [];
+    } else {
+      const record = {
+        call_id: callId,
+        caller: data.caller || "",
+        callee: data.callee || "",
+        direction: data.direction || "inbound",
+        started_at: Date.now() / 1e3,
+        status: "in-progress",
+        turns: []
+      };
+      this.activeCalls.set(callId, record);
+    }
+    this.publish("call_start", {
+      call_id: callId,
+      caller: data.caller || "",
+      callee: data.callee || "",
+      direction: data.direction || "inbound"
+    });
+  }
+  /**
+   * Pre-register an outbound call before any webhook fires. Lets the
+   * dashboard surface attempts that never reach media (no-answer, busy,
+   * carrier-rejected). Mirrors the Python ``record_call_initiated``.
+   */
+  recordCallInitiated(data) {
+    const callId = data.call_id || "";
+    if (!callId) return;
+    if (this.activeCalls.has(callId)) return;
     const record = {
       call_id: callId,
       caller: data.caller || "",
       callee: data.callee || "",
-      direction: data.direction || "inbound",
+      direction: data.direction || "outbound",
       started_at: Date.now() / 1e3,
+      status: "initiated",
       turns: []
     };
     this.activeCalls.set(callId, record);
-    this.publish("call_start", {
+    this.publish("call_initiated", {
       call_id: callId,
       caller: record.caller,
       callee: record.callee,
-      direction: record.direction
+      direction: record.direction,
+      status: record.status
     });
   }
+  /**
+   * Update the status of an active or completed call. Terminal states
+   * (completed, no-answer, busy, failed, canceled, webhook_error) move the
+   * row from active to completed so the UI freezes the live duration timer.
+   */
+  updateCallStatus(callId, status, extra = {}) {
+    if (!callId || !status) return;
+    const TERMINAL = /* @__PURE__ */ new Set(["completed", "no-answer", "busy", "failed", "canceled", "webhook_error"]);
+    const active = this.activeCalls.get(callId);
+    if (active) {
+      active.status = status;
+      Object.assign(active, extra);
+      if (TERMINAL.has(status)) {
+        const entry = {
+          call_id: callId,
+          caller: active.caller || "",
+          callee: active.callee || "",
+          direction: active.direction || "outbound",
+          started_at: active.started_at || 0,
+          ended_at: Date.now() / 1e3,
+          status,
+          metrics: null,
+          ...extra
+        };
+        this.activeCalls.delete(callId);
+        this.calls.push(entry);
+        if (this.calls.length > this.maxCalls) {
+          this.calls = this.calls.slice(-this.maxCalls);
+        }
+      }
+    } else {
+      for (let i = this.calls.length - 1; i >= 0; i--) {
+        if (this.calls[i].call_id === callId) {
+          this.calls[i].status = status;
+          Object.assign(this.calls[i], extra);
+          break;
+        }
+      }
+    }
+    this.publish("call_status", { call_id: callId, status, ...extra });
+  }
   recordTurn(data) {
     const callId = data.call_id || "";
     const turn = data.turn;
@@ -576,6 +450,8 @@ var MetricsStore = class extends EventEmitter {
     if (!callId) return;
     const active = this.activeCalls.get(callId);
     this.activeCalls.delete(callId);
+    const activeStatus = active?.status;
+    const resolvedStatus = activeStatus && activeStatus !== "in-progress" ? activeStatus : "completed";
     const entry = {
       call_id: callId,
       caller: data.caller || active?.caller || "",
@@ -584,6 +460,7 @@ var MetricsStore = class extends EventEmitter {
       started_at: active?.started_at || 0,
       ended_at: Date.now() / 1e3,
       transcript: data.transcript || [],
+      status: resolvedStatus,
       metrics: metrics ?? null
     };
     this.calls.push(entry);
@@ -1749,161 +1626,120 @@ function isWebSocketUrl(url) {
   return url.startsWith("ws://") || url.startsWith("wss://");
 }
 
-// src/providers/elevenlabs-tts.ts
-var ELEVENLABS_BASE_URL = "https://api.elevenlabs.io/v1";
-var ElevenLabsTTS = class {
-  constructor(apiKey, voiceId = "21m00Tcm4TlvDq8ikWAM", modelId = "eleven_turbo_v2_5", outputFormat = "pcm_16000") {
+// src/providers/deepgram-stt.ts
+import WebSocket3 from "ws";
+var DEEPGRAM_WS_URL = "wss://api.deepgram.com/v1/listen";
+var DeepgramSTT = class _DeepgramSTT {
+  ws = null;
+  callbacks = [];
+  /** Request ID from Deepgram — used to query actual cost post-call. */
+  requestId = "";
+  apiKey;
+  language;
+  model;
+  encoding;
+  sampleRate;
+  endpointingMs;
+  utteranceEndMs;
+  smartFormat;
+  interimResults;
+  vadEvents;
+  constructor(apiKey, languageOrOptions, model, encoding, sampleRate, options) {
     this.apiKey = apiKey;
-    this.voiceId = voiceId;
-    this.modelId = modelId;
-    this.outputFormat = outputFormat;
-  }
-  /**
-   * Synthesise text to speech and return the full audio as a single Buffer.
-   *
-   * For large chunks (or when latency matters) call `synthesizeStream` instead.
-   */
-  async synthesize(text) {
-    const chunks = [];
-    for await (const chunk of this.synthesizeStream(text)) {
-      chunks.push(chunk);
-    }
-    return Buffer.concat(chunks);
+    const opts = typeof languageOrOptions === "object" && languageOrOptions !== null ? languageOrOptions : options ?? {};
+    this.language = (typeof languageOrOptions === "string" ? languageOrOptions : opts.language) ?? "en";
+    this.model = model ?? opts.model ?? "nova-3";
+    this.encoding = encoding ?? opts.encoding ?? "linear16";
+    this.sampleRate = sampleRate ?? opts.sampleRate ?? 16e3;
+    this.endpointingMs = opts.endpointingMs ?? 150;
+    this.utteranceEndMs = opts.utteranceEndMs === null ? null : opts.utteranceEndMs ?? 1e3;
+    this.smartFormat = opts.smartFormat ?? true;
+    this.interimResults = opts.interimResults ?? true;
+    this.vadEvents = opts.vadEvents ?? true;
+  }
+  /** Factory for Twilio calls — mulaw 8 kHz. Forwards tuning options through. */
+  static forTwilio(apiKey, language = "en", model = "nova-3", options = {}) {
+    return new _DeepgramSTT(apiKey, language, model, "mulaw", 8e3, options);
   }
-  /**
-   * Synthesise text and yield audio chunks as they arrive (streaming).
-   *
-   * The yielded buffers are raw PCM at 16 kHz (or whatever `outputFormat` is
-   * configured to).
-   */
-  async *synthesizeStream(text) {
-    const url = `${ELEVENLABS_BASE_URL}/text-to-speech/${encodeURIComponent(this.voiceId)}/stream?output_format=${encodeURIComponent(this.outputFormat)}`;
-    const response = await fetch(url, {
-      method: "POST",
-      headers: {
-        "xi-api-key": this.apiKey,
-        "Content-Type": "application/json"
-      },
-      body: JSON.stringify({ text, model_id: this.modelId }),
-      signal: AbortSignal.timeout(3e4)
+  async connect() {
+    const params = new URLSearchParams({
+      model: this.model,
+      language: this.language,
+      encoding: this.encoding,
+      sample_rate: String(this.sampleRate),
+      channels: "1",
+      interim_results: this.interimResults ? "true" : "false",
+      endpointing: String(this.endpointingMs),
+      smart_format: this.smartFormat ? "true" : "false",
+      vad_events: this.vadEvents ? "true" : "false",
+      no_delay: "true"
     });
-    if (!response.ok) {
-      const body = await response.text();
-      throw new Error(`ElevenLabs TTS error ${response.status}: ${body}`);
-    }
-    if (!response.body) {
-      throw new Error("ElevenLabs TTS: no response body");
+    if (this.utteranceEndMs !== null) {
+      params.set("utterance_end_ms", String(Math.max(this.utteranceEndMs, 1e3)));
     }
-    const reader = response.body.getReader();
-    try {
-      while (true) {
-        const { done, value } = await reader.read();
-        if (done) break;
-        if (value && value.length > 0) {
-          yield Buffer.from(value);
-        }
-      }
-    } finally {
-      if (typeof reader.cancel === "function") await reader.cancel().catch(() => {
+    const url = `${DEEPGRAM_WS_URL}?${params.toString()}`;
+    this.ws = new WebSocket3(url, {
+      headers: { Authorization: `Token ${this.apiKey}` }
+    });
+    await new Promise((resolve, reject) => {
+      const timer = setTimeout(() => reject(new Error("Deepgram connect timeout")), 1e4);
+      this.ws.once("open", () => {
+        clearTimeout(timer);
+        resolve();
+      });
+      this.ws.once("error", (err) => {
+        clearTimeout(timer);
+        reject(err);
       });
-      reader.releaseLock();
-    }
-  }
-};
-
-// src/providers/openai-tts.ts
-var OPENAI_TTS_URL = "https://api.openai.com/v1/audio/speech";
-var OpenAITTS = class _OpenAITTS {
-  constructor(apiKey, voice = "alloy", model = "tts-1") {
-    this.apiKey = apiKey;
-    this.voice = voice;
-    this.model = model;
-  }
-  /**
-   * Synthesise text to speech and return the full audio as a single Buffer.
-   *
-   * For large chunks (or when latency matters) call `synthesizeStream` instead.
-   */
-  async synthesize(text) {
-    const chunks = [];
-    for await (const chunk of this.synthesizeStream(text)) {
-      chunks.push(chunk);
-    }
-    return Buffer.concat(chunks);
-  }
-  /**
-   * Synthesise text and yield audio chunks as they arrive (streaming).
-   *
-   * OpenAI returns 24 kHz PCM16; each chunk is resampled to 16 kHz before
-   * yielding so the output is ready for telephony pipelines.
-   */
-  async *synthesizeStream(text) {
-    const response = await fetch(OPENAI_TTS_URL, {
-      method: "POST",
-      headers: {
-        "Authorization": `Bearer ${this.apiKey}`,
-        "Content-Type": "application/json"
-      },
-      body: JSON.stringify({
-        model: this.model,
-        input: text,
-        voice: this.voice,
-        response_format: "pcm"
-      }),
-      signal: AbortSignal.timeout(3e4)
     });
-    if (!response.ok) {
-      const body = await response.text();
-      throw new Error(`OpenAI TTS error ${response.status}: ${body}`);
-    }
-    if (!response.body) {
-      throw new Error("OpenAI TTS: no response body");
-    }
-    const reader = response.body.getReader();
-    try {
-      while (true) {
-        const { done, value } = await reader.read();
-        if (done) break;
-        if (value && value.length > 0) {
-          yield _OpenAITTS.resample24kTo16k(Buffer.from(value));
-        }
+    this.ws.on("message", (raw) => {
+      let data;
+      try {
+        data = JSON.parse(raw.toString());
+      } catch {
+        return;
       }
-    } finally {
-      if (typeof reader.cancel === "function") await reader.cancel().catch(() => {
-      });
-      reader.releaseLock();
+      if (data.type === "Metadata" && data.request_id) {
+        this.requestId = data.request_id;
+        return;
+      }
+      if (data.type !== "Results") return;
+      const alternatives = data.channel?.alternatives ?? [];
+      if (!alternatives.length) return;
+      const best = alternatives[0];
+      const text = (best.transcript ?? "").trim();
+      if (!text) return;
+      const transcript = {
+        text,
+        isFinal: Boolean(data.is_final) || Boolean(data.speech_final),
+        confidence: best.confidence ?? 0
+      };
+      for (const cb of this.callbacks) {
+        cb(transcript);
+      }
+    });
+  }
+  sendAudio(audio) {
+    if (!this.ws || this.ws.readyState !== WebSocket3.OPEN) return;
+    this.ws.send(audio);
+  }
+  onTranscript(callback) {
+    if (this.callbacks.length >= 10) {
+      getLogger().warn("DeepgramSTT: maximum of 10 onTranscript callbacks reached; replacing the last callback.");
+      this.callbacks[this.callbacks.length - 1] = callback;
+      return;
     }
+    this.callbacks.push(callback);
   }
-  /**
-   * Resample 24 kHz PCM16-LE to 16 kHz by taking 2 out of every 3 samples.
-   *
-   * For each group of 3 input samples the first is kept as-is and the second
-   * output sample is the average of input samples 2 and 3.  This matches the
-   * Python SDK implementation.
-   */
-  static resample24kTo16k(audio) {
-    if (audio.length < 2) return audio;
-    const sampleCount = Math.floor(audio.length / 2);
-    const samples = new Int16Array(sampleCount);
-    for (let i = 0; i < sampleCount; i++) {
-      samples[i] = audio.readInt16LE(i * 2);
-    }
-    const resampled = [];
-    for (let i = 0; i < samples.length; i += 3) {
-      resampled.push(samples[i]);
-      if (i + 1 < samples.length) {
-        if (i + 2 < samples.length) {
-          resampled.push(Math.trunc((samples[i + 1] + samples[i + 2]) / 2));
-        } else {
-          resampled.push(samples[i + 1]);
-        }
+  close() {
+    if (this.ws) {
+      try {
+        this.ws.send(JSON.stringify({ type: "CloseStream" }));
+      } catch {
       }
+      this.ws.close();
+      this.ws = null;
     }
-    const out = Buffer.alloc(resampled.length * 2);
-    for (let i = 0; i < resampled.length; i++) {
-      out.writeInt16LE(resampled[i], i * 2);
-    }
-    return out;
   }
 };
 
@@ -2493,6 +2329,9 @@ var StreamHandler = class {
   maxDurationTimer = null;
   transcriptProcessing = false;
   transcriptQueue = [];
+  // BUG #22 throttle state — mirror Python impl.
+  lastCommitText = "";
+  lastCommitAt = 0;
   history;
   metricsAcc;
   constructor(deps, ws, caller, callee) {
@@ -2501,8 +2340,8 @@ var StreamHandler = class {
     this.caller = caller;
     this.callee = callee;
     this.history = createHistoryManager(200);
-    const sttProviderName = deps.agent.stt?.provider || (deps.agent.deepgramKey ? "deepgram" : void 0);
-    const ttsProviderName = deps.agent.tts?.provider === "elevenlabs" ? "elevenlabs" : deps.agent.tts?.provider === "openai" ? "openai_tts" : deps.agent.elevenlabsKey ? "elevenlabs" : void 0;
+    const sttProviderName = deps.agent.stt ? deps.agent.stt.constructor?.name ?? "custom" : void 0;
+    const ttsProviderName = deps.agent.tts ? deps.agent.tts.constructor?.name ?? "custom" : void 0;
     const providerMode = deps.agent.provider ?? "openai_realtime";
     this.metricsAcc = new CallMetricsAccumulator({
       callId: "",
@@ -2603,15 +2442,23 @@ var StreamHandler = class {
     this.streamSid = sid;
   }
   /** Handle an incoming audio chunk (already decoded from base64). */
-  handleAudio(audioBuffer) {
+  async handleAudio(audioBuffer) {
     const provider = this.deps.agent.provider ?? "openai_realtime";
-    if (provider === "pipeline" && this.stt && !this.isSpeaking) {
-      if (this.deps.bridge.telephonyProvider === "twilio") {
-        const pcm8k = mulawToPcm16(audioBuffer);
-        const pcm16k = resample8kTo16k(pcm8k);
-        this.stt.sendAudio(pcm16k);
+    if (provider === "pipeline" && this.stt) {
+      if (this.isSpeaking && (this.deps.agent.bargeInThresholdMs ?? 300) === 0) {
+        return;
+      }
+      const pcm8k = mulawToPcm16(audioBuffer);
+      const pcm16k = resample8kTo16k(pcm8k);
+      const hooks = this.deps.agent.hooks;
+      if (hooks) {
+        const hookExecutor = new PipelineHookExecutor(hooks);
+        const hookCtx = this.buildHookContext();
+        const processed = await hookExecutor.runBeforeSendToStt(pcm16k, hookCtx);
+        if (processed === null) return;
+        this.stt.sendAudio(processed);
       } else {
-        this.stt.sendAudio(audioBuffer);
+        this.stt.sendAudio(pcm16k);
       }
     } else if (this.adapter) {
       if (this.adapter instanceof ElevenLabsConvAIAdapter && this.deps.bridge.telephonyProvider === "twilio") {
@@ -2684,18 +2531,8 @@ var StreamHandler = class {
   // ---------------------------------------------------------------------------
   async initPipeline(resolvedPrompt) {
     const label = this.deps.bridge.label;
-    this.stt = this.deps.bridge.createStt(this.deps.agent);
-    if (this.deps.agent.tts) {
-      if (this.deps.agent.tts.provider === "elevenlabs") {
-        this.tts = new ElevenLabsTTS(this.deps.agent.tts.apiKey, this.deps.agent.tts.voice ?? "21m00Tcm4TlvDq8ikWAM");
-      }
-      if (this.deps.agent.tts.provider === "openai") {
-        this.tts = new OpenAITTS(this.deps.agent.tts.apiKey, this.deps.agent.tts.voice ?? "alloy");
-      }
-    } else if (this.deps.agent.elevenlabsKey) {
-      const voiceId = this.deps.agent.voice && this.deps.agent.voice !== "alloy" ? this.deps.agent.voice : "21m00Tcm4TlvDq8ikWAM";
-      this.tts = new ElevenLabsTTS(this.deps.agent.elevenlabsKey, voiceId);
-    }
+    this.stt = await this.deps.bridge.createStt(this.deps.agent);
+    this.tts = await createTTS(this.deps.agent);
     if (!this.stt) {
       getLogger().info(`Pipeline mode (${label}): no STT configured`);
     }
@@ -2806,7 +2643,59 @@ var StreamHandler = class {
     }
   }
   async processTranscript(transcript) {
+    if (transcript.text && this.isSpeaking) {
+      getLogger().info(
+        `Barge-in: caller spoke over agent (${sanitizeLogValue(transcript.text.slice(0, 40))})`
+      );
+      this.isSpeaking = false;
+      try {
+        this.deps.bridge.sendClear(this.ws, this.streamSid);
+      } catch (err) {
+        getLogger().debug(`sendClear during barge-in failed: ${String(err)}`);
+      }
+      this.metricsAcc.recordTurnInterrupted();
+    }
     if (!transcript.isFinal || !transcript.text) return;
+    const now = Date.now();
+    const normalised = transcript.text.trim().toLowerCase();
+    const stripped = normalised.replace(/[.,!?;: ]+$/, "").trim();
+    const sinceLastMs = now - this.lastCommitAt;
+    const HALLUCINATIONS = /* @__PURE__ */ new Set([
+      "you",
+      "thank you",
+      "thanks",
+      "yeah",
+      "yes",
+      "no",
+      "okay",
+      "ok",
+      "uh",
+      "um",
+      "mmm",
+      "hmm",
+      ".",
+      "bye",
+      "right",
+      "cool"
+    ]);
+    if (HALLUCINATIONS.has(stripped) || stripped === "") {
+      getLogger().info(`Dropped likely STT hallucination: ${sanitizeLogValue(normalised.slice(0, 40))}`);
+      return;
+    }
+    if (sinceLastMs < 2e3 && normalised === this.lastCommitText) {
+      getLogger().info(
+        `Dropped duplicate final transcript (${(sinceLastMs / 1e3).toFixed(1)}s since last): ${sanitizeLogValue(normalised.slice(0, 40))}`
+      );
+      return;
+    }
+    if (sinceLastMs < 500) {
+      getLogger().info(
+        `Dropped back-to-back final transcript (${(sinceLastMs / 1e3).toFixed(2)}s since last): ${sanitizeLogValue(normalised.slice(0, 40))}`
+      );
+      return;
+    }
+    this.lastCommitText = normalised;
+    this.lastCommitAt = now;
     const label = this.deps.bridge.label;
     getLogger().info(`User (${label} pipeline): ${sanitizeLogValue(transcript.text)}`);
     this.metricsAcc.startTurn();
@@ -3161,10 +3050,11 @@ var StreamHandler = class {
       this.maxDurationTimer = null;
     }
     await this.deps.bridge.queryTelephonyCost(this.metricsAcc, this.callId);
-    const deepgramKey = this.deps.agent.deepgramKey;
-    const deepgramRequestId = this.stt?.requestId;
-    if (deepgramKey && deepgramRequestId) {
-      await queryDeepgramCost(this.metricsAcc, deepgramKey, deepgramRequestId);
+    if (this.stt instanceof DeepgramSTT && this.stt.requestId) {
+      const dgKey = this.stt.apiKey;
+      if (dgKey) {
+        await queryDeepgramCost(this.metricsAcc, dgKey, this.stt.requestId);
+      }
     }
     const finalMetrics = this.metricsAcc.endCall();
     const callEndData = {
@@ -3321,11 +3211,16 @@ function resolveVariables(template, variables) {
   return result;
 }
 function buildAIAdapter(config, agent, resolvedPrompt) {
+  const engine = agent.engine;
   if (agent.provider === "elevenlabs_convai") {
-    const key = agent.elevenlabsKey ?? "";
+    if (!engine || engine.kind !== "elevenlabs_convai") {
+      throw new Error(
+        "ElevenLabs ConvAI mode requires `agent.engine = new ElevenLabsConvAI({...})`."
+      );
+    }
     return new ElevenLabsConvAIAdapter(
-      key,
-      agent.elevenlabsAgentId ?? "",
+      engine.apiKey,
+      engine.agentId,
       agent.voice ?? "21m00Tcm4TlvDq8ikWAM",
       "eleven_turbo_v2_5",
       agent.language ?? "en",
@@ -3338,8 +3233,9 @@ function buildAIAdapter(config, agent, resolvedPrompt) {
     parameters: t.parameters
   })) ?? [];
   const tools = [...agentTools, TRANSFER_CALL_TOOL, END_CALL_TOOL];
+  const openaiKey = engine && engine.kind === "openai_realtime" ? engine.apiKey : config.openaiKey ?? "";
   return new OpenAIRealtimeAdapter(
-    config.openaiKey ?? "",
+    openaiKey,
     agent.model,
     agent.voice,
     resolvedPrompt ?? agent.systemPrompt,
@@ -3397,16 +3293,7 @@ var TwilioBridge = class {
     }
   }
   createStt(agent) {
-    if (agent.stt) {
-      if (agent.stt.provider === "deepgram") {
-        return DeepgramSTT.forTwilio(agent.stt.apiKey, agent.stt.language ?? "en");
-      } else if (agent.stt.provider === "whisper") {
-        return WhisperSTT.forTwilio(agent.stt.apiKey, agent.stt.language ?? "en");
-      }
-    } else if (agent.deepgramKey) {
-      return DeepgramSTT.forTwilio(agent.deepgramKey, agent.language ?? "en");
-    }
-    return null;
+    return createSTT(agent);
   }
   async queryTelephonyCost(metricsAcc, callId) {
     if (this.config.twilioSid && this.config.twilioToken && callId) {
@@ -3454,12 +3341,12 @@ var TelnyxBridge = class {
   label = "Telnyx";
   telephonyProvider = "telnyx";
   sendAudio(ws, audioBase64, _streamSid) {
-    ws.send(JSON.stringify({ event_type: "media", payload: { audio: { chunk: audioBase64 } } }));
+    ws.send(JSON.stringify({ event: "media", media: { payload: audioBase64 } }));
   }
   sendMark(_ws, _markName, _streamSid) {
   }
   sendClear(ws, _streamSid) {
-    ws.send(JSON.stringify({ event_type: "media_stop" }));
+    ws.send(JSON.stringify({ event: "clear" }));
   }
   async transferCall(callId, toNumber) {
     if (!isValidTelnyxTransferTarget(toNumber)) {
@@ -3553,16 +3440,7 @@ var TelnyxBridge = class {
     ws.close();
   }
   createStt(agent) {
-    if (agent.stt) {
-      if (agent.stt.provider === "deepgram") {
-        return new DeepgramSTT(agent.stt.apiKey, agent.stt.language ?? "en", "nova-3", "linear16", 16e3);
-      } else if (agent.stt.provider === "whisper") {
-        return new WhisperSTT(agent.stt.apiKey, "whisper-1", agent.stt.language ?? "en");
-      }
-    } else if (agent.deepgramKey) {
-      return new DeepgramSTT(agent.deepgramKey, agent.language ?? "en", "nova-3", "linear16", 16e3);
-    }
-    return null;
+    return createSTT(agent);
   }
   async queryTelephonyCost(metricsAcc, callId) {
     if (this.config.telnyxKey && callId) {
@@ -3607,6 +3485,7 @@ var EmbeddedServer = class {
   server = null;
   wss = null;
   twilioTokenWarningLogged = false;
+  telnyxSigWarningLogged = false;
   metricsStore;
   pricing;
   remoteHandler = new RemoteMessageHandler();
@@ -3654,6 +3533,31 @@ var EmbeddedServer = class {
       mountApi(app, this.metricsStore, this.dashboardToken);
       getLogger().info("Dashboard: http://127.0.0.1:" + port + "/");
     }
+    app.post("/webhooks/twilio/status", (req, res) => {
+      if (this.config.twilioToken) {
+        const signature = req.headers["x-twilio-signature"] || "";
+        const url = `https://${this.config.webhookUrl}${req.originalUrl}`;
+        const params = req.body ?? {};
+        if (!validateTwilioSignature(url, params, signature, this.config.twilioToken)) {
+          res.status(403).send("Invalid signature");
+          return;
+        }
+      }
+      const body = req.body;
+      const callSid = sanitizeLogValue(body["CallSid"] ?? "");
+      const callStatus = sanitizeLogValue(body["CallStatus"] ?? "");
+      const duration = body["CallDuration"] ?? body["Duration"] ?? "";
+      getLogger().info(
+        `Twilio status ${callStatus} for call ${callSid} (duration=${duration})`
+      );
+      if (callSid && callStatus) {
+        const extra = {};
+        const parsed = parseFloat(duration);
+        if (!Number.isNaN(parsed)) extra.duration_seconds = parsed;
+        this.metricsStore.updateCallStatus(callSid, callStatus, extra);
+      }
+      res.status(204).send();
+    });
     app.post("/webhooks/twilio/recording", (req, res) => {
       if (this.config.twilioToken) {
         const signature = req.headers["x-twilio-signature"] || "";
@@ -3739,7 +3643,7 @@ var EmbeddedServer = class {
       const twiml = `<?xml version="1.0" encoding="UTF-8"?><Response><Connect><Stream url="${xmlStreamUrl}"><Parameter name="caller" value="${xmlEscape(caller)}"/><Parameter name="callee" value="${xmlEscape(callee)}"/></Stream></Connect></Response>`;
       res.type("text/xml").send(twiml);
     });
-    app.post("/webhooks/telnyx/voice", (req, res) => {
+    app.post("/webhooks/telnyx/voice", async (req, res) => {
       if (this.config.telnyxPublicKey) {
         const rawBody = req.rawBody ?? "";
         const signature = req.headers["telnyx-signature-ed25519"] ?? "";
@@ -3748,7 +3652,8 @@ var EmbeddedServer = class {
           getLogger().warn("Telnyx webhook rejected: invalid or missing Ed25519 signature");
           return res.status(403).send("Invalid signature");
         }
-      } else {
+      } else if (!this.telnyxSigWarningLogged) {
+        this.telnyxSigWarningLogged = true;
         getLogger().warn("Telnyx webhook signature verification is disabled. Set telnyxPublicKey in LocalOptions for production use.");
       }
       const body = req.body;
@@ -3758,41 +3663,77 @@ var EmbeddedServer = class {
       if (typeof body.data.event_type !== "string" || typeof body.data.payload !== "object" || body.data.payload === null) {
         return res.status(400).send("Invalid body");
       }
-      const eventType = body?.data?.event_type ?? "";
+      const eventType = body.data.event_type ?? "";
+      const payload = body.data.payload ?? {};
       if (eventType === "call.dtmf.received") {
-        const digit = String(body.data?.payload?.digit ?? "").trim();
+        const digit = String(payload.digit ?? "").trim();
         if (digit) {
           getLogger().info(`Telnyx DTMF received (webhook): ${sanitizeLogValue(digit)}`);
         }
-        return res.json({ received: true });
+        return res.status(200).send();
       }
       if (eventType === "call.recording.saved") {
-        const recordingUrl = body.data?.payload?.recording_urls?.mp3 ?? body.data?.payload?.recording_urls?.wav ?? body.data?.payload?.public_recording_urls?.mp3 ?? "";
+        const recordingUrl = payload.recording_urls?.mp3 ?? payload.recording_urls?.wav ?? payload.public_recording_urls?.mp3 ?? "";
         if (recordingUrl) {
           getLogger().info(`Telnyx recording saved (webhook): ${sanitizeLogValue(recordingUrl)}`);
         }
-        return res.json({ received: true });
+        return res.status(200).send();
       }
-      if (eventType === "call.initiated") {
-        const payload = body?.data?.payload ?? {};
-        const callControlId = payload.call_control_id ?? "";
-        const caller = payload.from ?? "";
-        const callee = payload.to ?? "";
-        const streamUrl = `wss://${this.config.webhookUrl}/ws/stream/${encodeURIComponent(callControlId)}?caller=${encodeURIComponent(caller)}&callee=${encodeURIComponent(callee)}`;
-        const commands = [
-          { command: "answer" },
-          {
-            command: "stream_start",
-            params: {
+      const callControlId = payload.call_control_id ?? "";
+      if (!callControlId) {
+        getLogger().warn("Telnyx webhook rejected: missing call_control_id");
+        return res.status(400).send("Invalid webhook payload");
+      }
+      const apiKey = this.config.telnyxKey;
+      if (!apiKey) {
+        getLogger().warn("Telnyx webhook: missing telnyxKey in LocalOptions");
+        return res.status(500).send("Missing Telnyx API key");
+      }
+      const apiBase = "https://api.telnyx.com/v2";
+      const authHeaders = {
+        "Content-Type": "application/json",
+        Authorization: `Bearer ${apiKey}`
+      };
+      try {
+        if (eventType === "call.initiated") {
+          getLogger().info(`Telnyx call.initiated ${callControlId} \u2014 answering`);
+          const resp = await fetch(`${apiBase}/calls/${encodeURIComponent(callControlId)}/actions/answer`, {
+            method: "POST",
+            headers: authHeaders,
+            body: JSON.stringify({}),
+            signal: AbortSignal.timeout(1e4)
+          });
+          if (!resp.ok) {
+            getLogger().warn(`Telnyx answer failed: ${resp.status} ${(await resp.text()).slice(0, 200)}`);
+          }
+        } else if (eventType === "call.answered") {
+          const caller = payload.from ?? "";
+          const callee = payload.to ?? "";
+          const streamUrl = `wss://${this.config.webhookUrl}/ws/stream/${encodeURIComponent(callControlId)}?caller=${encodeURIComponent(caller)}&callee=${encodeURIComponent(callee)}`;
+          getLogger().info(`Telnyx call.answered ${callControlId} \u2014 starting stream`);
+          const resp = await fetch(`${apiBase}/calls/${encodeURIComponent(callControlId)}/actions/streaming_start`, {
+            method: "POST",
+            headers: authHeaders,
+            body: JSON.stringify({
               stream_url: streamUrl,
-              stream_track: "both_tracks"
-            }
+              stream_track: "both_tracks",
+              stream_bidirectional_mode: "rtp",
+              stream_bidirectional_codec: "PCMU",
+              stream_bidirectional_sampling_rate: 8e3,
+              stream_bidirectional_target_legs: "self"
+            }),
+            signal: AbortSignal.timeout(1e4)
+          });
+          if (!resp.ok) {
+            getLogger().warn(`Telnyx streaming_start failed: ${resp.status} ${(await resp.text()).slice(0, 200)}`);
           }
-        ];
-        res.json({ commands });
-      } else {
-        res.json({ received: true });
+        } else {
+          getLogger().debug(`Telnyx event ignored: ${eventType}`);
+        }
+      } catch (e) {
+        getLogger().error(`Telnyx webhook handler error: ${String(e)}`);
       }
+      return res.status(200).send();
     });
     this.server = createServer(app);
     this.wss = new WebSocketServer({ noServer: true });
@@ -3939,11 +3880,12 @@ Connect AI agents to phone numbers in 4 lines of code
           getLogger().error("Failed to parse Telnyx WS message:", e);
           return;
         }
-        const eventType = data.event_type ?? "";
-        getLogger().info(`Telnyx event: ${eventType}`);
-        if (eventType === "stream_started" && !streamStarted) {
+        const event = data.event ?? "";
+        if (event === "connected") return;
+        getLogger().info(`Telnyx event: ${event}`);
+        if (event === "start" && !streamStarted) {
           streamStarted = true;
-          const callControlId = data.payload?.call_control_id ?? "";
+          const callControlId = data.start?.call_control_id ?? "";
           if (callControlId) this.activeCallIds.set(ws, callControlId);
           await handler.handleCallStart(callControlId);
           if (this.recording) {
@@ -3953,22 +3895,21 @@ Connect AI agents to phone numbers in 4 lines of code
               getLogger().warn(`Could not start recording: ${String(e)}`);
             }
           }
-        } else if (eventType === "media") {
-          const audioChunk = data.payload?.audio?.chunk ?? "";
+        } else if (event === "media") {
+          const track = data.media?.track ?? "inbound";
+          if (track !== "inbound") return;
+          const audioChunk = data.media?.payload ?? "";
           if (!audioChunk) return;
           handler.handleAudio(Buffer.from(audioChunk, "base64"));
-        } else if (eventType === "call.dtmf.received") {
-          const digit = String(data.payload?.digit ?? "").trim();
+        } else if (event === "dtmf") {
+          const digit = String(data.dtmf?.digit ?? "").trim();
           if (digit) {
             getLogger().info(`Telnyx DTMF received: ${digit}`);
             await handler.handleDtmf(digit);
           }
-        } else if (eventType === "call.recording.saved") {
-          const recordingUrl = data.payload?.recording_urls?.mp3 ?? data.payload?.recording_urls?.wav ?? data.payload?.public_recording_urls?.mp3 ?? "";
-          if (recordingUrl) {
-            getLogger().info(`Telnyx recording saved: ${recordingUrl}`);
-          }
-        } else if (eventType === "stream_stopped") {
+        } else if (event === "error") {
+          getLogger().warn(`Telnyx stream error: ${JSON.stringify(data)}`);
+        } else if (event === "stop") {
           await handler.handleStop();
         }
       } catch (err) {
@@ -4437,8 +4378,6 @@ var TestSession = class {
 export {
   OpenAIRealtimeAdapter,
   ElevenLabsConvAIAdapter,
-  DeepgramSTT,
-  WhisperSTT,
   DEFAULT_PRICING,
   mergePricing,
   calculateSttCost,
@@ -4454,8 +4393,7 @@ export {
   RemoteMessageHandler,
   isRemoteUrl,
   isWebSocketUrl,
-  ElevenLabsTTS,
-  OpenAITTS,
+  DeepgramSTT,
   CallMetricsAccumulator,
   mulawToPcm16,
   pcm16ToMulaw,