@livekit/agents-plugin-cartesia 1.5.3 → 1.6.0

package/dist/tts.test.cjs

@@ -1,7 +1,10 @@
 "use strict";
+var import_agents = require("@livekit/agents");
 var import_agents_plugin_openai = require("@livekit/agents-plugin-openai");
 var import_agents_plugins_test = require("@livekit/agents-plugins-test");
+var import_node_events = require("node:events");
 var import_vitest = require("vitest");
+var import_ws = require("ws");
 var import_tts = require("./tts.cjs");
 const hasCartesiaConfig = Boolean(process.env.CARTESIA_API_KEY && process.env.OPENAI_API_KEY);
 if (hasCartesiaConfig) {
@@ -14,4 +17,205 @@ if (hasCartesiaConfig) {
     });
   });
 }
+const CHUNK_BASE64 = Buffer.alloc(4800).toString("base64");
+async function startWebSocketServer() {
+  const wss = new import_ws.WebSocketServer({ host: "127.0.0.1", port: 0 });
+  await (0, import_node_events.once)(wss, "listening");
+  const address = wss.address();
+  return { wss, baseURL: `http://127.0.0.1:${address.port}` };
+}
+async function closeWebSocketServer(wss) {
+  for (const client of wss.clients) {
+    client.close();
+  }
+  await new Promise((resolve) => wss.close(() => resolve()));
+}
+async function waitFor(promise, timeoutMs = 1e3) {
+  let timeout;
+  try {
+    return await Promise.race([
+      promise,
+      new Promise((_, reject) => {
+        timeout = setTimeout(() => reject(new Error("timed out waiting for promise")), timeoutMs);
+      })
+    ]);
+  } finally {
+    if (timeout) clearTimeout(timeout);
+  }
+}
+function serveCartesia(wss, onStop) {
+  let connectionCount = 0;
+  wss.on("connection", (ws) => {
+    connectionCount++;
+    const connectionNumber = connectionCount;
+    ws.on("message", (raw) => {
+      const message = JSON.parse(raw.toString());
+      if (message.continue !== false) return;
+      const contextId = message.context_id;
+      if (onStop && !onStop(ws, contextId, connectionNumber)) return;
+      ws.send(
+        JSON.stringify({
+          type: "chunk",
+          data: CHUNK_BASE64,
+          done: false,
+          status_code: 200,
+          step_time: 0,
+          context_id: contextId
+        })
+      );
+      ws.send(
+        JSON.stringify({ type: "done", done: true, status_code: 200, context_id: contextId })
+      );
+    });
+  });
+  return { connectionCount: () => connectionCount };
+}
+async function synthesizeTurn(cartesia, text, connOptions) {
+  const stream = cartesia.stream({ connOptions });
+  stream.pushText(text);
+  stream.endInput();
+  try {
+    const events = [];
+    for await (const event of stream) {
+      if (event !== import_agents.tts.SynthesizeStream.END_OF_STREAM) events.push(event);
+    }
+    return events;
+  } finally {
+    stream.close();
+  }
+}
+(0, import_vitest.describe)("Cartesia streaming pool", () => {
+  (0, import_vitest.it)("reuses one websocket across sequential turns", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss);
+    const cartesia = new import_tts.TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      (0, import_vitest.expect)(await synthesizeTurn(cartesia, "first turn.")).not.toHaveLength(0);
+      (0, import_vitest.expect)(await synthesizeTurn(cartesia, "second turn.")).not.toHaveLength(0);
+      (0, import_vitest.expect)(server.connectionCount()).toBe(1);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  (0, import_vitest.it)("prewarms and reuses the ready websocket", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss);
+    const connected = (0, import_node_events.once)(wss, "connection");
+    const cartesia = new import_tts.TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      cartesia.prewarm();
+      await waitFor(connected);
+      (0, import_vitest.expect)(await synthesizeTurn(cartesia, "prewarmed turn.")).not.toHaveLength(0);
+      (0, import_vitest.expect)(server.connectionCount()).toBe(1);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  (0, import_vitest.it)("discards a poisoned websocket after a failure", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {
+      if (connectionNumber === 1) {
+        ws.close(1011, "provider failure");
+        return false;
+      }
+      return true;
+    });
+    const cartesia = new import_tts.TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      (0, import_vitest.expect)(
+        await synthesizeTurn(cartesia, "failing turn.", {
+          ...import_agents.DEFAULT_API_CONNECT_OPTIONS,
+          maxRetry: 0
+        })
+      ).toHaveLength(0);
+      (0, import_vitest.expect)(await synthesizeTurn(cartesia, "recovery turn.")).not.toHaveLength(0);
+      (0, import_vitest.expect)(server.connectionCount()).toBe(2);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  (0, import_vitest.it)("fails over when the socket drops mid-generation instead of ending silently", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss, (ws, contextId, connectionNumber) => {
+      if (connectionNumber === 1) {
+        ws.send(
+          JSON.stringify({
+            type: "chunk",
+            data: CHUNK_BASE64,
+            done: false,
+            status_code: 200,
+            step_time: 0,
+            context_id: contextId
+          })
+        );
+        setTimeout(() => ws.close(1011, "mid-speech drop"), 5);
+        return false;
+      }
+      return true;
+    });
+    const cartesia = new import_tts.TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      (0, import_vitest.expect)(
+        await synthesizeTurn(cartesia, "dropping turn.", {
+          ...import_agents.DEFAULT_API_CONNECT_OPTIONS,
+          maxRetry: 0
+        })
+      ).toHaveLength(0);
+      (0, import_vitest.expect)(await synthesizeTurn(cartesia, "recovery turn.")).not.toHaveLength(0);
+      (0, import_vitest.expect)(server.connectionCount()).toBe(2);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  (0, import_vitest.it)("replaces a websocket that closed while idle", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    let firstConnectionClosed;
+    const firstClosed = new Promise((resolve) => {
+      firstConnectionClosed = resolve;
+    });
+    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {
+      if (connectionNumber === 1) {
+        ws.on("close", () => firstConnectionClosed == null ? void 0 : firstConnectionClosed());
+        setTimeout(() => ws.close(), 10);
+      }
+      return true;
+    });
+    const cartesia = new import_tts.TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      (0, import_vitest.expect)(await synthesizeTurn(cartesia, "first turn.")).not.toHaveLength(0);
+      await waitFor(firstClosed);
+      await new Promise((resolve) => setTimeout(resolve, 100));
+      (0, import_vitest.expect)(
+        await waitFor(
+          synthesizeTurn(cartesia, "second turn.", { ...import_agents.DEFAULT_API_CONNECT_OPTIONS, maxRetry: 0 })
+        )
+      ).not.toHaveLength(0);
+      (0, import_vitest.expect)(server.connectionCount()).toBe(2);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  (0, import_vitest.it)("closes the pooled websocket when the TTS closes", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    serveCartesia(wss);
+    const cartesia = new import_tts.TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      await synthesizeTurn(cartesia, "closing turn.");
+      await cartesia.close();
+      await waitFor(
+        (async () => {
+          while (wss.clients.size > 0) await new Promise((r) => setTimeout(r, 5));
+        })()
+      );
+      (0, import_vitest.expect)(wss.clients.size).toBe(0);
+    } finally {
+      await closeWebSocketServer(wss);
+    }
+  });
+});
 //# sourceMappingURL=tts.test.cjs.map

package/dist/tts.test.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { STT } from '@livekit/agents-plugin-openai';\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { TTS } from './tts.js';\n\nconst hasCartesiaConfig = Boolean(process.env.CARTESIA_API_KEY && process.env.OPENAI_API_KEY);\n\nif (hasCartesiaConfig) {\n  describe('Cartesia', async () => {\n    await tts(new TTS(), new STT());\n  });\n} else {\n  describe('Cartesia', () => {\n    it.skip('requires CARTESIA_API_KEY and OPENAI_API_KEY', () => {});\n  });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAA6B;AAC7B,iBAAoB;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,oBAAoB,QAAQ,IAAI,cAAc;AAE5F,IAAI,mBAAmB;AACrB,8BAAS,YAAY,YAAY;AAC/B,cAAM,gCAAI,IAAI,eAAI,GAAG,IAAI,gCAAI,CAAC;AAAA,EAChC,CAAC;AACH,OAAO;AACL,8BAAS,YAAY,MAAM;AACzB,qBAAG,KAAK,gDAAgD,MAAM;AAAA,IAAC,CAAC;AAAA,EAClE,CAAC;AACH;","names":[]}
+{"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, tts } from '@livekit/agents';\nimport { STT } from '@livekit/agents-plugin-openai';\nimport { tts as testTts } from '@livekit/agents-plugins-test';\nimport { once } from 'node:events';\nimport type { AddressInfo } from 'node:net';\nimport { describe, expect, it } from 'vitest';\nimport { type WebSocket, WebSocketServer } from 'ws';\nimport { TTS } from './tts.js';\n\nconst hasCartesiaConfig = Boolean(process.env.CARTESIA_API_KEY && process.env.OPENAI_API_KEY);\n\nif (hasCartesiaConfig) {\n  describe('Cartesia', async () => {\n    await testTts(new TTS(), new STT());\n  });\n} else {\n  describe('Cartesia', () => {\n    it.skip('requires CARTESIA_API_KEY and OPENAI_API_KEY', () => {});\n  });\n}\n\n// A single 24 kHz mono s16le frame's worth of silence, base64-encoded the way\n// Cartesia sends audio chunks.\nconst CHUNK_BASE64 = Buffer.alloc(4800).toString('base64');\n\nasync function startWebSocketServer() {\n  const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });\n  await once(wss, 'listening');\n  const address = wss.address() as AddressInfo;\n  return { wss, baseURL: `http://127.0.0.1:${address.port}` };\n}\n\nasync function closeWebSocketServer(wss: WebSocketServer): Promise<void> {\n  for (const client of wss.clients) {\n    client.close();\n  }\n  await new Promise<void>((resolve) => wss.close(() => resolve()));\n}\n\nasync function waitFor<T>(promise: Promise<T>, timeoutMs = 1000): Promise<T> {\n  let timeout: ReturnType<typeof setTimeout> | undefined;\n  try {\n    return await Promise.race([\n      promise,\n      new Promise<never>((_, reject) => {\n        timeout = setTimeout(() => reject(new Error('timed out waiting for promise')), timeoutMs);\n      }),\n    ]);\n  } finally {\n    if (timeout) clearTimeout(timeout);\n  }\n}\n\n// A minimal Cartesia TTS WebSocket server: for every generation it replies with\n// one audio chunk and a done message, echoing the caller's context_id. `onStop`\n// lets a test override the reply (e.g. to simulate a provider failure); return\n// false to suppress the normal chunk/done reply.\nfunction serveCartesia(\n  wss: WebSocketServer,\n  onStop?: (ws: WebSocket, contextId: string, connectionNumber: number) => boolean,\n): { connectionCount: () => number } {\n  let connectionCount = 0;\n  wss.on('connection', (ws) => {\n    connectionCount++;\n    const connectionNumber = connectionCount;\n    ws.on('message', (raw) => {\n      const message = JSON.parse(raw.toString()) as { context_id: string; continue?: boolean };\n      if (message.continue !== false) return; // only reply once the turn is closed\n      const contextId = message.context_id;\n      if (onStop && !onStop(ws, contextId, connectionNumber)) return;\n      ws.send(\n        JSON.stringify({\n          type: 'chunk',\n          data: CHUNK_BASE64,\n          done: false,\n          status_code: 200,\n          step_time: 0,\n          context_id: contextId,\n        }),\n      );\n      ws.send(\n        JSON.stringify({ type: 'done', done: true, status_code: 200, context_id: contextId }),\n      );\n    });\n  });\n  return { connectionCount: () => connectionCount };\n}\n\nasync function synthesizeTurn(\n  cartesia: TTS,\n  text: string,\n  connOptions?: APIConnectOptions,\n): Promise<tts.SynthesizedAudio[]> {\n  const stream = cartesia.stream({ connOptions });\n  stream.pushText(text);\n  stream.endInput();\n\n  try {\n    const events: tts.SynthesizedAudio[] = [];\n    for await (const event of stream) {\n      if (event !== tts.SynthesizeStream.END_OF_STREAM) events.push(event);\n    }\n    return events;\n  } finally {\n    stream.close();\n  }\n}\n\ndescribe('Cartesia streaming pool', () => {\n  it('reuses one websocket across sequential turns', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    const server = serveCartesia(wss);\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      expect(await synthesizeTurn(cartesia, 'first turn.')).not.toHaveLength(0);\n      expect(await synthesizeTurn(cartesia, 'second turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(1);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('prewarms and reuses the ready websocket', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    const server = serveCartesia(wss);\n    const connected = once(wss, 'connection');\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      cartesia.prewarm();\n      await waitFor(connected);\n      expect(await synthesizeTurn(cartesia, 'prewarmed turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(1);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('discards a poisoned websocket after a failure', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    // The first connection drops the turn; the second serves it normally.\n    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {\n      if (connectionNumber === 1) {\n        ws.close(1011, 'provider failure');\n        return false;\n      }\n      return true;\n    });\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      expect(\n        await synthesizeTurn(cartesia, 'failing turn.', {\n          ...DEFAULT_API_CONNECT_OPTIONS,\n          maxRetry: 0,\n        }),\n      ).toHaveLength(0);\n      expect(await synthesizeTurn(cartesia, 'recovery turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(2);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('fails over when the socket drops mid-generation instead of ending silently', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    // Connection 1 emits one audio chunk, then drops WITHOUT a done message,\n    // i.e. mid-speech. Connection 2 serves the recovery turn normally.\n    const server = serveCartesia(wss, (ws, contextId, connectionNumber) => {\n      if (connectionNumber === 1) {\n        ws.send(\n          JSON.stringify({\n            type: 'chunk',\n            data: CHUNK_BASE64,\n            done: false,\n            status_code: 200,\n            step_time: 0,\n            context_id: contextId,\n          }),\n        );\n        setTimeout(() => ws.close(1011, 'mid-speech drop'), 5);\n        return false; // suppress the normal chunk/done reply\n      }\n      return true;\n    });\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      // The dropped turn does not complete successfully (it fails over rather\n      // than silently ending); at maxRetry: 0 that surfaces as no audio.\n      expect(\n        await synthesizeTurn(cartesia, 'dropping turn.', {\n          ...DEFAULT_API_CONNECT_OPTIONS,\n          maxRetry: 0,\n        }),\n      ).toHaveLength(0);\n      // The dead socket is discarded, so the next turn opens a fresh one.\n      expect(await synthesizeTurn(cartesia, 'recovery turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(2);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('replaces a websocket that closed while idle', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    let firstConnectionClosed: (() => void) | undefined;\n    const firstClosed = new Promise<void>((resolve) => {\n      firstConnectionClosed = resolve;\n    });\n    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {\n      if (connectionNumber === 1) {\n        ws.on('close', () => firstConnectionClosed?.());\n        // Serve the turn, then drop the idle socket so the next turn reconnects.\n        setTimeout(() => ws.close(), 10);\n      }\n      return true;\n    });\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      expect(await synthesizeTurn(cartesia, 'first turn.')).not.toHaveLength(0);\n      await waitFor(firstClosed);\n      // Let the client observe the close so the idle handler removes the socket\n      // before the next checkout, making the maxRetry: 0 assertion deterministic.\n      await new Promise((resolve) => setTimeout(resolve, 100));\n      // maxRetry: 0 proves the idle-closed socket was dropped from the pool, not\n      // handed back to burn the turn's only attempt.\n      expect(\n        await waitFor(\n          synthesizeTurn(cartesia, 'second turn.', { ...DEFAULT_API_CONNECT_OPTIONS, maxRetry: 0 }),\n        ),\n      ).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(2);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('closes the pooled websocket when the TTS closes', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    serveCartesia(wss);\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      await synthesizeTurn(cartesia, 'closing turn.');\n      await cartesia.close();\n      // close() drains the pooled socket; give the close frame a beat to land.\n      await waitFor(\n        (async () => {\n          while (wss.clients.size > 0) await new Promise((r) => setTimeout(r, 5));\n        })(),\n      );\n      expect(wss.clients.size).toBe(0);\n    } finally {\n      await closeWebSocketServer(wss);\n    }\n  });\n});\n"],"mappings":";AAGA,oBAAyE;AACzE,kCAAoB;AACpB,iCAA+B;AAC/B,yBAAqB;AAErB,oBAAqC;AACrC,gBAAgD;AAChD,iBAAoB;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,oBAAoB,QAAQ,IAAI,cAAc;AAE5F,IAAI,mBAAmB;AACrB,8BAAS,YAAY,YAAY;AAC/B,cAAM,2BAAAA,KAAQ,IAAI,eAAI,GAAG,IAAI,gCAAI,CAAC;AAAA,EACpC,CAAC;AACH,OAAO;AACL,8BAAS,YAAY,MAAM;AACzB,qBAAG,KAAK,gDAAgD,MAAM;AAAA,IAAC,CAAC;AAAA,EAClE,CAAC;AACH;AAIA,MAAM,eAAe,OAAO,MAAM,IAAI,EAAE,SAAS,QAAQ;AAEzD,eAAe,uBAAuB;AACpC,QAAM,MAAM,IAAI,0BAAgB,EAAE,MAAM,aAAa,MAAM,EAAE,CAAC;AAC9D,YAAM,yBAAK,KAAK,WAAW;AAC3B,QAAM,UAAU,IAAI,QAAQ;AAC5B,SAAO,EAAE,KAAK,SAAS,oBAAoB,QAAQ,IAAI,GAAG;AAC5D;AAEA,eAAe,qBAAqB,KAAqC;AACvE,aAAW,UAAU,IAAI,SAAS;AAChC,WAAO,MAAM;AAAA,EACf;AACA,QAAM,IAAI,QAAc,CAAC,YAAY,IAAI,MAAM,MAAM,QAAQ,CAAC,CAAC;AACjE;AAEA,eAAe,QAAW,SAAqB,YAAY,KAAkB;AAC3E,MAAI;AACJ,MAAI;AACF,WAAO,MAAM,QAAQ,KAAK;AAAA,MACxB;AAAA,MACA,IAAI,QAAe,CAAC,GAAG,WAAW;AAChC,kBAAU,WAAW,MAAM,OAAO,IAAI,MAAM,+BAA+B,CAAC,GAAG,SAAS;AAAA,MAC1F,CAAC;AAAA,IACH,CAAC;AAAA,EACH,UAAE;AACA,QAAI,QAAS,cAAa,OAAO;AAAA,EACnC;AACF;AAMA,SAAS,cACP,KACA,QACmC;AACnC,MAAI,kBAAkB;AACtB,MAAI,GAAG,cAAc,CAAC,OAAO;AAC3B;AACA,UAAM,mBAAmB;AACzB,OAAG,GAAG,WAAW,CAAC,QAAQ;AACxB,YAAM,UAAU,KAAK,MAAM,IAAI,SAAS,CAAC;AACzC,UAAI,QAAQ,aAAa,MAAO;AAChC,YAAM,YAAY,QAAQ;AAC1B,UAAI,UAAU,CAAC,OAAO,IAAI,WAAW,gBAAgB,EAAG;AACxD,SAAG;AAAA,QACD,KAAK,UAAU;AAAA,UACb,MAAM;AAAA,UACN,MAAM;AAAA,UACN,MAAM;AAAA,UACN,aAAa;AAAA,UACb,WAAW;AAAA,UACX,YAAY;AAAA,QACd,CAAC;AAAA,MACH;AACA,SAAG;AAAA,QACD,KAAK,UAAU,EAAE,MAAM,QAAQ,MAAM,MAAM,aAAa,KAAK,YAAY,UAAU,CAAC;AAAA,MACtF;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AACD,SAAO,EAAE,iBAAiB,MAAM,gBAAgB;AAClD;AAEA,eAAe,eACb,UACA,MACA,aACiC;AACjC,QAAM,SAAS,SAAS,OAAO,EAAE,YAAY,CAAC;AAC9C,SAAO,SAAS,IAAI;AACpB,SAAO,SAAS;AAEhB,MAAI;AACF,UAAM,SAAiC,CAAC;AACxC,qBAAiB,SAAS,QAAQ;AAChC,UAAI,UAAU,kBAAI,iBAAiB,cAAe,QAAO,KAAK,KAAK;AAAA,IACrE;AACA,WAAO;AAAA,EACT,UAAE;AACA,WAAO,MAAM;AAAA,EACf;AACF;AAAA,IAEA,wBAAS,2BAA2B,MAAM;AACxC,wBAAG,gDAAgD,YAAY;AAC7D,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,SAAS,cAAc,GAAG;AAEhC,UAAM,WAAW,IAAI,eAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,gCAAO,MAAM,eAAe,UAAU,aAAa,CAAC,EAAE,IAAI,aAAa,CAAC;AACxE,gCAAO,MAAM,eAAe,UAAU,cAAc,CAAC,EAAE,IAAI,aAAa,CAAC;AACzE,gCAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,wBAAG,2CAA2C,YAAY;AACxD,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,SAAS,cAAc,GAAG;AAChC,UAAM,gBAAY,yBAAK,KAAK,YAAY;AAExC,UAAM,WAAW,IAAI,eAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,eAAS,QAAQ;AACjB,YAAM,QAAQ,SAAS;AACvB,gCAAO,MAAM,eAAe,UAAU,iBAAiB,CAAC,EAAE,IAAI,aAAa,CAAC;AAC5E,gCAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,wBAAG,iDAAiD,YAAY;AAC9D,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AAEpD,UAAM,SAAS,cAAc,KAAK,CAAC,IAAI,YAAY,qBAAqB;AACtE,UAAI,qBAAqB,GAAG;AAC1B,WAAG,MAAM,MAAM,kBAAkB;AACjC,eAAO;AAAA,MACT;AACA,aAAO;AAAA,IACT,CAAC;AAED,UAAM,WAAW,IAAI,eAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF;AAAA,QACE,MAAM,eAAe,UAAU,iBAAiB;AAAA,UAC9C,GAAG;AAAA,UACH,UAAU;AAAA,QACZ,CAAC;AAAA,MACH,EAAE,aAAa,CAAC;AAChB,gCAAO,MAAM,eAAe,UAAU,gBAAgB,CAAC,EAAE,IAAI,aAAa,CAAC;AAC3E,gCAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,wBAAG,8EAA8E,YAAY;AAC3F,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AAGpD,UAAM,SAAS,cAAc,KAAK,CAAC,IAAI,WAAW,qBAAqB;AACrE,UAAI,qBAAqB,GAAG;AAC1B,WAAG;AAAA,UACD,KAAK,UAAU;AAAA,YACb,MAAM;AAAA,YACN,MAAM;AAAA,YACN,MAAM;AAAA,YACN,aAAa;AAAA,YACb,WAAW;AAAA,YACX,YAAY;AAAA,UACd,CAAC;AAAA,QACH;AACA,mBAAW,MAAM,GAAG,MAAM,MAAM,iBAAiB,GAAG,CAAC;AACrD,eAAO;AAAA,MACT;AACA,aAAO;AAAA,IACT,CAAC;AAED,UAAM,WAAW,IAAI,eAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AAGF;AAAA,QACE,MAAM,eAAe,UAAU,kBAAkB;AAAA,UAC/C,GAAG;AAAA,UACH,UAAU;AAAA,QACZ,CAAC;AAAA,MACH,EAAE,aAAa,CAAC;AAEhB,gCAAO,MAAM,eAAe,UAAU,gBAAgB,CAAC,EAAE,IAAI,aAAa,CAAC;AAC3E,gCAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,wBAAG,+CAA+C,YAAY;AAC5D,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,QAAI;AACJ,UAAM,cAAc,IAAI,QAAc,CAAC,YAAY;AACjD,8BAAwB;AAAA,IAC1B,CAAC;AACD,UAAM,SAAS,cAAc,KAAK,CAAC,IAAI,YAAY,qBAAqB;AACtE,UAAI,qBAAqB,GAAG;AAC1B,WAAG,GAAG,SAAS,MAAM,gEAAyB;AAE9C,mBAAW,MAAM,GAAG,MAAM,GAAG,EAAE;AAAA,MACjC;AACA,aAAO;AAAA,IACT,CAAC;AAED,UAAM,WAAW,IAAI,eAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,gCAAO,MAAM,eAAe,UAAU,aAAa,CAAC,EAAE,IAAI,aAAa,CAAC;AACxE,YAAM,QAAQ,WAAW;AAGzB,YAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,GAAG,CAAC;AAGvD;AAAA,QACE,MAAM;AAAA,UACJ,eAAe,UAAU,gBAAgB,EAAE,GAAG,2CAA6B,UAAU,EAAE,CAAC;AAAA,QAC1F;AAAA,MACF,EAAE,IAAI,aAAa,CAAC;AACpB,gCAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,wBAAG,mDAAmD,YAAY;AAChE,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,kBAAc,GAAG;AAEjB,UAAM,WAAW,IAAI,eAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,YAAM,eAAe,UAAU,eAAe;AAC9C,YAAM,SAAS,MAAM;AAErB,YAAM;AAAA,SACH,YAAY;AACX,iBAAO,IAAI,QAAQ,OAAO,EAAG,OAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,CAAC,CAAC;AAAA,QACxE,GAAG;AAAA,MACL;AACA,gCAAO,IAAI,QAAQ,IAAI,EAAE,KAAK,CAAC;AAAA,IACjC,UAAE;AACA,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AACH,CAAC;","names":["testTts"]}

package/dist/tts.test.js

@@ -1,11 +1,14 @@
+import { DEFAULT_API_CONNECT_OPTIONS, tts } from "@livekit/agents";
 import { STT } from "@livekit/agents-plugin-openai";
-import { tts } from "@livekit/agents-plugins-test";
-import { describe, it } from "vitest";
+import { tts as testTts } from "@livekit/agents-plugins-test";
+import { once } from "node:events";
+import { describe, expect, it } from "vitest";
+import { WebSocketServer } from "ws";
 import { TTS } from "./tts.js";
 const hasCartesiaConfig = Boolean(process.env.CARTESIA_API_KEY && process.env.OPENAI_API_KEY);
 if (hasCartesiaConfig) {
   describe("Cartesia", async () => {
-    await tts(new TTS(), new STT());
+    await testTts(new TTS(), new STT());
   });
 } else {
   describe("Cartesia", () => {
@@ -13,4 +16,205 @@ if (hasCartesiaConfig) {
     });
   });
 }
+const CHUNK_BASE64 = Buffer.alloc(4800).toString("base64");
+async function startWebSocketServer() {
+  const wss = new WebSocketServer({ host: "127.0.0.1", port: 0 });
+  await once(wss, "listening");
+  const address = wss.address();
+  return { wss, baseURL: `http://127.0.0.1:${address.port}` };
+}
+async function closeWebSocketServer(wss) {
+  for (const client of wss.clients) {
+    client.close();
+  }
+  await new Promise((resolve) => wss.close(() => resolve()));
+}
+async function waitFor(promise, timeoutMs = 1e3) {
+  let timeout;
+  try {
+    return await Promise.race([
+      promise,
+      new Promise((_, reject) => {
+        timeout = setTimeout(() => reject(new Error("timed out waiting for promise")), timeoutMs);
+      })
+    ]);
+  } finally {
+    if (timeout) clearTimeout(timeout);
+  }
+}
+function serveCartesia(wss, onStop) {
+  let connectionCount = 0;
+  wss.on("connection", (ws) => {
+    connectionCount++;
+    const connectionNumber = connectionCount;
+    ws.on("message", (raw) => {
+      const message = JSON.parse(raw.toString());
+      if (message.continue !== false) return;
+      const contextId = message.context_id;
+      if (onStop && !onStop(ws, contextId, connectionNumber)) return;
+      ws.send(
+        JSON.stringify({
+          type: "chunk",
+          data: CHUNK_BASE64,
+          done: false,
+          status_code: 200,
+          step_time: 0,
+          context_id: contextId
+        })
+      );
+      ws.send(
+        JSON.stringify({ type: "done", done: true, status_code: 200, context_id: contextId })
+      );
+    });
+  });
+  return { connectionCount: () => connectionCount };
+}
+async function synthesizeTurn(cartesia, text, connOptions) {
+  const stream = cartesia.stream({ connOptions });
+  stream.pushText(text);
+  stream.endInput();
+  try {
+    const events = [];
+    for await (const event of stream) {
+      if (event !== tts.SynthesizeStream.END_OF_STREAM) events.push(event);
+    }
+    return events;
+  } finally {
+    stream.close();
+  }
+}
+describe("Cartesia streaming pool", () => {
+  it("reuses one websocket across sequential turns", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss);
+    const cartesia = new TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      expect(await synthesizeTurn(cartesia, "first turn.")).not.toHaveLength(0);
+      expect(await synthesizeTurn(cartesia, "second turn.")).not.toHaveLength(0);
+      expect(server.connectionCount()).toBe(1);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  it("prewarms and reuses the ready websocket", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss);
+    const connected = once(wss, "connection");
+    const cartesia = new TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      cartesia.prewarm();
+      await waitFor(connected);
+      expect(await synthesizeTurn(cartesia, "prewarmed turn.")).not.toHaveLength(0);
+      expect(server.connectionCount()).toBe(1);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  it("discards a poisoned websocket after a failure", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {
+      if (connectionNumber === 1) {
+        ws.close(1011, "provider failure");
+        return false;
+      }
+      return true;
+    });
+    const cartesia = new TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      expect(
+        await synthesizeTurn(cartesia, "failing turn.", {
+          ...DEFAULT_API_CONNECT_OPTIONS,
+          maxRetry: 0
+        })
+      ).toHaveLength(0);
+      expect(await synthesizeTurn(cartesia, "recovery turn.")).not.toHaveLength(0);
+      expect(server.connectionCount()).toBe(2);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  it("fails over when the socket drops mid-generation instead of ending silently", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    const server = serveCartesia(wss, (ws, contextId, connectionNumber) => {
+      if (connectionNumber === 1) {
+        ws.send(
+          JSON.stringify({
+            type: "chunk",
+            data: CHUNK_BASE64,
+            done: false,
+            status_code: 200,
+            step_time: 0,
+            context_id: contextId
+          })
+        );
+        setTimeout(() => ws.close(1011, "mid-speech drop"), 5);
+        return false;
+      }
+      return true;
+    });
+    const cartesia = new TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      expect(
+        await synthesizeTurn(cartesia, "dropping turn.", {
+          ...DEFAULT_API_CONNECT_OPTIONS,
+          maxRetry: 0
+        })
+      ).toHaveLength(0);
+      expect(await synthesizeTurn(cartesia, "recovery turn.")).not.toHaveLength(0);
+      expect(server.connectionCount()).toBe(2);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  it("replaces a websocket that closed while idle", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    let firstConnectionClosed;
+    const firstClosed = new Promise((resolve) => {
+      firstConnectionClosed = resolve;
+    });
+    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {
+      if (connectionNumber === 1) {
+        ws.on("close", () => firstConnectionClosed == null ? void 0 : firstConnectionClosed());
+        setTimeout(() => ws.close(), 10);
+      }
+      return true;
+    });
+    const cartesia = new TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      expect(await synthesizeTurn(cartesia, "first turn.")).not.toHaveLength(0);
+      await waitFor(firstClosed);
+      await new Promise((resolve) => setTimeout(resolve, 100));
+      expect(
+        await waitFor(
+          synthesizeTurn(cartesia, "second turn.", { ...DEFAULT_API_CONNECT_OPTIONS, maxRetry: 0 })
+        )
+      ).not.toHaveLength(0);
+      expect(server.connectionCount()).toBe(2);
+    } finally {
+      await cartesia.close();
+      await closeWebSocketServer(wss);
+    }
+  });
+  it("closes the pooled websocket when the TTS closes", async () => {
+    const { wss, baseURL } = await startWebSocketServer();
+    serveCartesia(wss);
+    const cartesia = new TTS({ apiKey: "test-key", baseUrl: baseURL });
+    try {
+      await synthesizeTurn(cartesia, "closing turn.");
+      await cartesia.close();
+      await waitFor(
+        (async () => {
+          while (wss.clients.size > 0) await new Promise((r) => setTimeout(r, 5));
+        })()
+      );
+      expect(wss.clients.size).toBe(0);
+    } finally {
+      await closeWebSocketServer(wss);
+    }
+  });
+});
 //# sourceMappingURL=tts.test.js.map

package/dist/tts.test.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { STT } from '@livekit/agents-plugin-openai';\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { TTS } from './tts.js';\n\nconst hasCartesiaConfig = Boolean(process.env.CARTESIA_API_KEY && process.env.OPENAI_API_KEY);\n\nif (hasCartesiaConfig) {\n  describe('Cartesia', async () => {\n    await tts(new TTS(), new STT());\n  });\n} else {\n  describe('Cartesia', () => {\n    it.skip('requires CARTESIA_API_KEY and OPENAI_API_KEY', () => {});\n  });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,UAAU,UAAU;AAC7B,SAAS,WAAW;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,oBAAoB,QAAQ,IAAI,cAAc;AAE5F,IAAI,mBAAmB;AACrB,WAAS,YAAY,YAAY;AAC/B,UAAM,IAAI,IAAI,IAAI,GAAG,IAAI,IAAI,CAAC;AAAA,EAChC,CAAC;AACH,OAAO;AACL,WAAS,YAAY,MAAM;AACzB,OAAG,KAAK,gDAAgD,MAAM;AAAA,IAAC,CAAC;AAAA,EAClE,CAAC;AACH;","names":[]}
+{"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS, tts } from '@livekit/agents';\nimport { STT } from '@livekit/agents-plugin-openai';\nimport { tts as testTts } from '@livekit/agents-plugins-test';\nimport { once } from 'node:events';\nimport type { AddressInfo } from 'node:net';\nimport { describe, expect, it } from 'vitest';\nimport { type WebSocket, WebSocketServer } from 'ws';\nimport { TTS } from './tts.js';\n\nconst hasCartesiaConfig = Boolean(process.env.CARTESIA_API_KEY && process.env.OPENAI_API_KEY);\n\nif (hasCartesiaConfig) {\n  describe('Cartesia', async () => {\n    await testTts(new TTS(), new STT());\n  });\n} else {\n  describe('Cartesia', () => {\n    it.skip('requires CARTESIA_API_KEY and OPENAI_API_KEY', () => {});\n  });\n}\n\n// A single 24 kHz mono s16le frame's worth of silence, base64-encoded the way\n// Cartesia sends audio chunks.\nconst CHUNK_BASE64 = Buffer.alloc(4800).toString('base64');\n\nasync function startWebSocketServer() {\n  const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });\n  await once(wss, 'listening');\n  const address = wss.address() as AddressInfo;\n  return { wss, baseURL: `http://127.0.0.1:${address.port}` };\n}\n\nasync function closeWebSocketServer(wss: WebSocketServer): Promise<void> {\n  for (const client of wss.clients) {\n    client.close();\n  }\n  await new Promise<void>((resolve) => wss.close(() => resolve()));\n}\n\nasync function waitFor<T>(promise: Promise<T>, timeoutMs = 1000): Promise<T> {\n  let timeout: ReturnType<typeof setTimeout> | undefined;\n  try {\n    return await Promise.race([\n      promise,\n      new Promise<never>((_, reject) => {\n        timeout = setTimeout(() => reject(new Error('timed out waiting for promise')), timeoutMs);\n      }),\n    ]);\n  } finally {\n    if (timeout) clearTimeout(timeout);\n  }\n}\n\n// A minimal Cartesia TTS WebSocket server: for every generation it replies with\n// one audio chunk and a done message, echoing the caller's context_id. `onStop`\n// lets a test override the reply (e.g. to simulate a provider failure); return\n// false to suppress the normal chunk/done reply.\nfunction serveCartesia(\n  wss: WebSocketServer,\n  onStop?: (ws: WebSocket, contextId: string, connectionNumber: number) => boolean,\n): { connectionCount: () => number } {\n  let connectionCount = 0;\n  wss.on('connection', (ws) => {\n    connectionCount++;\n    const connectionNumber = connectionCount;\n    ws.on('message', (raw) => {\n      const message = JSON.parse(raw.toString()) as { context_id: string; continue?: boolean };\n      if (message.continue !== false) return; // only reply once the turn is closed\n      const contextId = message.context_id;\n      if (onStop && !onStop(ws, contextId, connectionNumber)) return;\n      ws.send(\n        JSON.stringify({\n          type: 'chunk',\n          data: CHUNK_BASE64,\n          done: false,\n          status_code: 200,\n          step_time: 0,\n          context_id: contextId,\n        }),\n      );\n      ws.send(\n        JSON.stringify({ type: 'done', done: true, status_code: 200, context_id: contextId }),\n      );\n    });\n  });\n  return { connectionCount: () => connectionCount };\n}\n\nasync function synthesizeTurn(\n  cartesia: TTS,\n  text: string,\n  connOptions?: APIConnectOptions,\n): Promise<tts.SynthesizedAudio[]> {\n  const stream = cartesia.stream({ connOptions });\n  stream.pushText(text);\n  stream.endInput();\n\n  try {\n    const events: tts.SynthesizedAudio[] = [];\n    for await (const event of stream) {\n      if (event !== tts.SynthesizeStream.END_OF_STREAM) events.push(event);\n    }\n    return events;\n  } finally {\n    stream.close();\n  }\n}\n\ndescribe('Cartesia streaming pool', () => {\n  it('reuses one websocket across sequential turns', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    const server = serveCartesia(wss);\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      expect(await synthesizeTurn(cartesia, 'first turn.')).not.toHaveLength(0);\n      expect(await synthesizeTurn(cartesia, 'second turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(1);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('prewarms and reuses the ready websocket', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    const server = serveCartesia(wss);\n    const connected = once(wss, 'connection');\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      cartesia.prewarm();\n      await waitFor(connected);\n      expect(await synthesizeTurn(cartesia, 'prewarmed turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(1);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('discards a poisoned websocket after a failure', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    // The first connection drops the turn; the second serves it normally.\n    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {\n      if (connectionNumber === 1) {\n        ws.close(1011, 'provider failure');\n        return false;\n      }\n      return true;\n    });\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      expect(\n        await synthesizeTurn(cartesia, 'failing turn.', {\n          ...DEFAULT_API_CONNECT_OPTIONS,\n          maxRetry: 0,\n        }),\n      ).toHaveLength(0);\n      expect(await synthesizeTurn(cartesia, 'recovery turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(2);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('fails over when the socket drops mid-generation instead of ending silently', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    // Connection 1 emits one audio chunk, then drops WITHOUT a done message,\n    // i.e. mid-speech. Connection 2 serves the recovery turn normally.\n    const server = serveCartesia(wss, (ws, contextId, connectionNumber) => {\n      if (connectionNumber === 1) {\n        ws.send(\n          JSON.stringify({\n            type: 'chunk',\n            data: CHUNK_BASE64,\n            done: false,\n            status_code: 200,\n            step_time: 0,\n            context_id: contextId,\n          }),\n        );\n        setTimeout(() => ws.close(1011, 'mid-speech drop'), 5);\n        return false; // suppress the normal chunk/done reply\n      }\n      return true;\n    });\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      // The dropped turn does not complete successfully (it fails over rather\n      // than silently ending); at maxRetry: 0 that surfaces as no audio.\n      expect(\n        await synthesizeTurn(cartesia, 'dropping turn.', {\n          ...DEFAULT_API_CONNECT_OPTIONS,\n          maxRetry: 0,\n        }),\n      ).toHaveLength(0);\n      // The dead socket is discarded, so the next turn opens a fresh one.\n      expect(await synthesizeTurn(cartesia, 'recovery turn.')).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(2);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('replaces a websocket that closed while idle', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    let firstConnectionClosed: (() => void) | undefined;\n    const firstClosed = new Promise<void>((resolve) => {\n      firstConnectionClosed = resolve;\n    });\n    const server = serveCartesia(wss, (ws, _contextId, connectionNumber) => {\n      if (connectionNumber === 1) {\n        ws.on('close', () => firstConnectionClosed?.());\n        // Serve the turn, then drop the idle socket so the next turn reconnects.\n        setTimeout(() => ws.close(), 10);\n      }\n      return true;\n    });\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      expect(await synthesizeTurn(cartesia, 'first turn.')).not.toHaveLength(0);\n      await waitFor(firstClosed);\n      // Let the client observe the close so the idle handler removes the socket\n      // before the next checkout, making the maxRetry: 0 assertion deterministic.\n      await new Promise((resolve) => setTimeout(resolve, 100));\n      // maxRetry: 0 proves the idle-closed socket was dropped from the pool, not\n      // handed back to burn the turn's only attempt.\n      expect(\n        await waitFor(\n          synthesizeTurn(cartesia, 'second turn.', { ...DEFAULT_API_CONNECT_OPTIONS, maxRetry: 0 }),\n        ),\n      ).not.toHaveLength(0);\n      expect(server.connectionCount()).toBe(2);\n    } finally {\n      await cartesia.close();\n      await closeWebSocketServer(wss);\n    }\n  });\n\n  it('closes the pooled websocket when the TTS closes', async () => {\n    const { wss, baseURL } = await startWebSocketServer();\n    serveCartesia(wss);\n\n    const cartesia = new TTS({ apiKey: 'test-key', baseUrl: baseURL });\n    try {\n      await synthesizeTurn(cartesia, 'closing turn.');\n      await cartesia.close();\n      // close() drains the pooled socket; give the close frame a beat to land.\n      await waitFor(\n        (async () => {\n          while (wss.clients.size > 0) await new Promise((r) => setTimeout(r, 5));\n        })(),\n      );\n      expect(wss.clients.size).toBe(0);\n    } finally {\n      await closeWebSocketServer(wss);\n    }\n  });\n});\n"],"mappings":"AAGA,SAAiC,6BAA6B,WAAW;AACzE,SAAS,WAAW;AACpB,SAAS,OAAO,eAAe;AAC/B,SAAS,YAAY;AAErB,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAyB,uBAAuB;AAChD,SAAS,WAAW;AAEpB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,oBAAoB,QAAQ,IAAI,cAAc;AAE5F,IAAI,mBAAmB;AACrB,WAAS,YAAY,YAAY;AAC/B,UAAM,QAAQ,IAAI,IAAI,GAAG,IAAI,IAAI,CAAC;AAAA,EACpC,CAAC;AACH,OAAO;AACL,WAAS,YAAY,MAAM;AACzB,OAAG,KAAK,gDAAgD,MAAM;AAAA,IAAC,CAAC;AAAA,EAClE,CAAC;AACH;AAIA,MAAM,eAAe,OAAO,MAAM,IAAI,EAAE,SAAS,QAAQ;AAEzD,eAAe,uBAAuB;AACpC,QAAM,MAAM,IAAI,gBAAgB,EAAE,MAAM,aAAa,MAAM,EAAE,CAAC;AAC9D,QAAM,KAAK,KAAK,WAAW;AAC3B,QAAM,UAAU,IAAI,QAAQ;AAC5B,SAAO,EAAE,KAAK,SAAS,oBAAoB,QAAQ,IAAI,GAAG;AAC5D;AAEA,eAAe,qBAAqB,KAAqC;AACvE,aAAW,UAAU,IAAI,SAAS;AAChC,WAAO,MAAM;AAAA,EACf;AACA,QAAM,IAAI,QAAc,CAAC,YAAY,IAAI,MAAM,MAAM,QAAQ,CAAC,CAAC;AACjE;AAEA,eAAe,QAAW,SAAqB,YAAY,KAAkB;AAC3E,MAAI;AACJ,MAAI;AACF,WAAO,MAAM,QAAQ,KAAK;AAAA,MACxB;AAAA,MACA,IAAI,QAAe,CAAC,GAAG,WAAW;AAChC,kBAAU,WAAW,MAAM,OAAO,IAAI,MAAM,+BAA+B,CAAC,GAAG,SAAS;AAAA,MAC1F,CAAC;AAAA,IACH,CAAC;AAAA,EACH,UAAE;AACA,QAAI,QAAS,cAAa,OAAO;AAAA,EACnC;AACF;AAMA,SAAS,cACP,KACA,QACmC;AACnC,MAAI,kBAAkB;AACtB,MAAI,GAAG,cAAc,CAAC,OAAO;AAC3B;AACA,UAAM,mBAAmB;AACzB,OAAG,GAAG,WAAW,CAAC,QAAQ;AACxB,YAAM,UAAU,KAAK,MAAM,IAAI,SAAS,CAAC;AACzC,UAAI,QAAQ,aAAa,MAAO;AAChC,YAAM,YAAY,QAAQ;AAC1B,UAAI,UAAU,CAAC,OAAO,IAAI,WAAW,gBAAgB,EAAG;AACxD,SAAG;AAAA,QACD,KAAK,UAAU;AAAA,UACb,MAAM;AAAA,UACN,MAAM;AAAA,UACN,MAAM;AAAA,UACN,aAAa;AAAA,UACb,WAAW;AAAA,UACX,YAAY;AAAA,QACd,CAAC;AAAA,MACH;AACA,SAAG;AAAA,QACD,KAAK,UAAU,EAAE,MAAM,QAAQ,MAAM,MAAM,aAAa,KAAK,YAAY,UAAU,CAAC;AAAA,MACtF;AAAA,IACF,CAAC;AAAA,EACH,CAAC;AACD,SAAO,EAAE,iBAAiB,MAAM,gBAAgB;AAClD;AAEA,eAAe,eACb,UACA,MACA,aACiC;AACjC,QAAM,SAAS,SAAS,OAAO,EAAE,YAAY,CAAC;AAC9C,SAAO,SAAS,IAAI;AACpB,SAAO,SAAS;AAEhB,MAAI;AACF,UAAM,SAAiC,CAAC;AACxC,qBAAiB,SAAS,QAAQ;AAChC,UAAI,UAAU,IAAI,iBAAiB,cAAe,QAAO,KAAK,KAAK;AAAA,IACrE;AACA,WAAO;AAAA,EACT,UAAE;AACA,WAAO,MAAM;AAAA,EACf;AACF;AAEA,SAAS,2BAA2B,MAAM;AACxC,KAAG,gDAAgD,YAAY;AAC7D,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,SAAS,cAAc,GAAG;AAEhC,UAAM,WAAW,IAAI,IAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,aAAO,MAAM,eAAe,UAAU,aAAa,CAAC,EAAE,IAAI,aAAa,CAAC;AACxE,aAAO,MAAM,eAAe,UAAU,cAAc,CAAC,EAAE,IAAI,aAAa,CAAC;AACzE,aAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,KAAG,2CAA2C,YAAY;AACxD,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,UAAM,SAAS,cAAc,GAAG;AAChC,UAAM,YAAY,KAAK,KAAK,YAAY;AAExC,UAAM,WAAW,IAAI,IAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,eAAS,QAAQ;AACjB,YAAM,QAAQ,SAAS;AACvB,aAAO,MAAM,eAAe,UAAU,iBAAiB,CAAC,EAAE,IAAI,aAAa,CAAC;AAC5E,aAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,KAAG,iDAAiD,YAAY;AAC9D,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AAEpD,UAAM,SAAS,cAAc,KAAK,CAAC,IAAI,YAAY,qBAAqB;AACtE,UAAI,qBAAqB,GAAG;AAC1B,WAAG,MAAM,MAAM,kBAAkB;AACjC,eAAO;AAAA,MACT;AACA,aAAO;AAAA,IACT,CAAC;AAED,UAAM,WAAW,IAAI,IAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF;AAAA,QACE,MAAM,eAAe,UAAU,iBAAiB;AAAA,UAC9C,GAAG;AAAA,UACH,UAAU;AAAA,QACZ,CAAC;AAAA,MACH,EAAE,aAAa,CAAC;AAChB,aAAO,MAAM,eAAe,UAAU,gBAAgB,CAAC,EAAE,IAAI,aAAa,CAAC;AAC3E,aAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,KAAG,8EAA8E,YAAY;AAC3F,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AAGpD,UAAM,SAAS,cAAc,KAAK,CAAC,IAAI,WAAW,qBAAqB;AACrE,UAAI,qBAAqB,GAAG;AAC1B,WAAG;AAAA,UACD,KAAK,UAAU;AAAA,YACb,MAAM;AAAA,YACN,MAAM;AAAA,YACN,MAAM;AAAA,YACN,aAAa;AAAA,YACb,WAAW;AAAA,YACX,YAAY;AAAA,UACd,CAAC;AAAA,QACH;AACA,mBAAW,MAAM,GAAG,MAAM,MAAM,iBAAiB,GAAG,CAAC;AACrD,eAAO;AAAA,MACT;AACA,aAAO;AAAA,IACT,CAAC;AAED,UAAM,WAAW,IAAI,IAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AAGF;AAAA,QACE,MAAM,eAAe,UAAU,kBAAkB;AAAA,UAC/C,GAAG;AAAA,UACH,UAAU;AAAA,QACZ,CAAC;AAAA,MACH,EAAE,aAAa,CAAC;AAEhB,aAAO,MAAM,eAAe,UAAU,gBAAgB,CAAC,EAAE,IAAI,aAAa,CAAC;AAC3E,aAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,KAAG,+CAA+C,YAAY;AAC5D,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,QAAI;AACJ,UAAM,cAAc,IAAI,QAAc,CAAC,YAAY;AACjD,8BAAwB;AAAA,IAC1B,CAAC;AACD,UAAM,SAAS,cAAc,KAAK,CAAC,IAAI,YAAY,qBAAqB;AACtE,UAAI,qBAAqB,GAAG;AAC1B,WAAG,GAAG,SAAS,MAAM,gEAAyB;AAE9C,mBAAW,MAAM,GAAG,MAAM,GAAG,EAAE;AAAA,MACjC;AACA,aAAO;AAAA,IACT,CAAC;AAED,UAAM,WAAW,IAAI,IAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,aAAO,MAAM,eAAe,UAAU,aAAa,CAAC,EAAE,IAAI,aAAa,CAAC;AACxE,YAAM,QAAQ,WAAW;AAGzB,YAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,GAAG,CAAC;AAGvD;AAAA,QACE,MAAM;AAAA,UACJ,eAAe,UAAU,gBAAgB,EAAE,GAAG,6BAA6B,UAAU,EAAE,CAAC;AAAA,QAC1F;AAAA,MACF,EAAE,IAAI,aAAa,CAAC;AACpB,aAAO,OAAO,gBAAgB,CAAC,EAAE,KAAK,CAAC;AAAA,IACzC,UAAE;AACA,YAAM,SAAS,MAAM;AACrB,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AAED,KAAG,mDAAmD,YAAY;AAChE,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,kBAAc,GAAG;AAEjB,UAAM,WAAW,IAAI,IAAI,EAAE,QAAQ,YAAY,SAAS,QAAQ,CAAC;AACjE,QAAI;AACF,YAAM,eAAe,UAAU,eAAe;AAC9C,YAAM,SAAS,MAAM;AAErB,YAAM;AAAA,SACH,YAAY;AACX,iBAAO,IAAI,QAAQ,OAAO,EAAG,OAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,CAAC,CAAC;AAAA,QACxE,GAAG;AAAA,MACL;AACA,aAAO,IAAI,QAAQ,IAAI,EAAE,KAAK,CAAC;AAAA,IACjC,UAAE;AACA,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AACH,CAAC;","names":[]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livekit/agents-plugin-cartesia",
-  "version": "1.5.3",
+  "version": "1.6.0",
   "description": "Cartesia plugin for LiveKit Node Agents",
   "main": "dist/index.js",
   "require": "dist/index.cjs",
@@ -17,7 +17,10 @@
   },
   "author": "LiveKit",
   "type": "module",
-  "repository": "git@github.com:livekit/agents-js.git",
+  "repository": {
+    "type": "git",
+    "url": "git@github.com:livekit/agents-js.git"
+  },
   "license": "Apache-2.0",
   "files": [
     "dist",
@@ -30,18 +33,18 @@
     "@types/ws": "^8.5.10",
     "tsup": "^8.3.5",
     "typescript": "^5.0.0",
-    "@livekit/agents": "1.5.3",
-    "@livekit/agents-plugin-openai": "1.5.3",
-    "@livekit/agents-plugin-silero": "1.5.3",
-    "@livekit/agents-plugins-test": "1.5.3"
+    "@livekit/agents": "1.6.0",
+    "@livekit/agents-plugin-openai": "1.6.0",
+    "@livekit/agents-plugin-silero": "1.6.0",
+    "@livekit/agents-plugins-test": "1.6.0"
   },
   "dependencies": {
-    "ws": "^8.18.0"
+    "ws": "^8.21.0"
   },
   "peerDependencies": {
     "@livekit/rtc-node": "^0.13.31",
     "zod": "^3.25.76 || ^4.1.8",
-    "@livekit/agents": "1.5.3"
+    "@livekit/agents": "1.6.0"
   },
   "scripts": {
     "build": "tsup --onSuccess \"pnpm build:types\"",