@dtelecom/agents-js 0.1.14 → 0.1.16

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -296,8 +296,63 @@ var OpenRouterLLM = class {
296
296
  let buffer = "";
297
297
  const structured = !!this.responseFormat;
298
298
  let jsonBuffer = "";
299
- let lastSegmentIndex = 0;
300
- const segmentRe = /\{"lang"\s*:\s*"(\w+)"\s*,\s*"text"\s*:\s*"((?:[^"\\]|\\.)*)"\s*\}/g;
299
+ let segmentsYielded = false;
300
+ let lastUsage;
301
+ let inSegmentsArray = false;
302
+ let objectStart = -1;
303
+ let braceDepth = 0;
304
+ let scanIndex = 0;
305
+ let inString = false;
306
+ let escaped = false;
307
+ function extractSegments(buf) {
308
+ const results = [];
309
+ for (let i = scanIndex; i < buf.length; i++) {
310
+ const ch = buf[i];
311
+ if (escaped) {
312
+ escaped = false;
313
+ continue;
314
+ }
315
+ if (ch === "\\" && inString) {
316
+ escaped = true;
317
+ continue;
318
+ }
319
+ if (ch === '"') {
320
+ inString = !inString;
321
+ continue;
322
+ }
323
+ if (inString) continue;
324
+ if (!inSegmentsArray) {
325
+ if (ch === "[") {
326
+ const before = buf.slice(0, i).trimEnd();
327
+ if (before.endsWith(":") && buf.slice(0, i).includes('"segments"')) {
328
+ inSegmentsArray = true;
329
+ }
330
+ }
331
+ continue;
332
+ }
333
+ if (ch === "{") {
334
+ if (braceDepth === 0) objectStart = i;
335
+ braceDepth++;
336
+ } else if (ch === "}") {
337
+ braceDepth--;
338
+ if (braceDepth === 0 && objectStart >= 0) {
339
+ const objStr = buf.slice(objectStart, i + 1);
340
+ try {
341
+ const seg = JSON.parse(objStr);
342
+ if (seg.lang && seg.text) {
343
+ results.push({ lang: seg.lang, text: seg.text });
344
+ }
345
+ } catch {
346
+ }
347
+ objectStart = -1;
348
+ }
349
+ } else if (ch === "]" && braceDepth === 0) {
350
+ inSegmentsArray = false;
351
+ }
352
+ }
353
+ scanIndex = buf.length;
354
+ return results;
355
+ }
301
356
  try {
302
357
  while (true) {
303
358
  if (signal?.aborted) break;
@@ -310,10 +365,7 @@ var OpenRouterLLM = class {
310
365
  const trimmed = line.trim();
311
366
  if (!trimmed || !trimmed.startsWith("data: ")) continue;
312
367
  const data = trimmed.slice(6);
313
- if (data === "[DONE]") {
314
- yield { type: "done" };
315
- return;
316
- }
368
+ if (data === "[DONE]") break;
317
369
  try {
318
370
  const parsed = JSON.parse(data);
319
371
  const choice = parsed.choices?.[0];
@@ -322,31 +374,20 @@ var OpenRouterLLM = class {
322
374
  if (delta?.content) {
323
375
  if (structured) {
324
376
  jsonBuffer += delta.content;
325
- segmentRe.lastIndex = lastSegmentIndex;
326
- let match;
327
- while ((match = segmentRe.exec(jsonBuffer)) !== null) {
328
- const lang = match[1];
329
- const text = match[2].replace(/\\(.)/g, (_, c) => {
330
- if (c === "n") return "\n";
331
- if (c === "t") return " ";
332
- return c;
333
- });
334
- lastSegmentIndex = segmentRe.lastIndex;
335
- yield { type: "segment", segment: { lang, text } };
377
+ const segments = extractSegments(jsonBuffer);
378
+ for (const seg of segments) {
379
+ yield { type: "segment", segment: seg };
380
+ segmentsYielded = true;
336
381
  }
337
382
  } else {
338
383
  yield { type: "token", token: delta.content };
339
384
  }
340
385
  }
341
386
  if (parsed.usage) {
342
- yield {
343
- type: "done",
344
- usage: {
345
- promptTokens: parsed.usage.prompt_tokens,
346
- completionTokens: parsed.usage.completion_tokens
347
- }
387
+ lastUsage = {
388
+ promptTokens: parsed.usage.prompt_tokens,
389
+ completionTokens: parsed.usage.completion_tokens
348
390
  };
349
- return;
350
391
  }
351
392
  } catch {
352
393
  }
@@ -355,10 +396,10 @@ var OpenRouterLLM = class {
355
396
  } finally {
356
397
  reader.releaseLock();
357
398
  }
358
- if (structured && lastSegmentIndex === 0 && jsonBuffer.length > 0) {
359
- log2.warn(`Structured response yielded no segments. Raw buffer (first 200 chars): "${jsonBuffer.slice(0, 200)}"`);
399
+ if (structured && !segmentsYielded && jsonBuffer.length > 0) {
400
+ log2.warn(`LLM returned no segments. Raw JSON: "${jsonBuffer.slice(0, 300)}"`);
360
401
  }
361
- yield { type: "done" };
402
+ yield { type: "done", ...lastUsage ? { usage: lastUsage } : {} };
362
403
  }
363
404
  };
364
405
 
@@ -681,13 +722,21 @@ var DeepgramTTS = class {
681
722
  }
682
723
  async *synthesize(text, signal) {
683
724
  if (signal?.aborted) return;
684
- const segments = this.multiLanguage ? parseLangSegments(text, this.defaultLang) : [{ lang: this.defaultLang, text }];
725
+ const rawSegments = this.multiLanguage ? parseLangSegments(text, this.defaultLang) : [{ lang: this.defaultLang, text }];
726
+ const segments = [];
727
+ for (const seg of rawSegments) {
728
+ if (!seg.text.trim()) continue;
729
+ if (!/\p{L}/u.test(seg.text) && segments.length > 0) {
730
+ segments[segments.length - 1].text += seg.text;
731
+ } else {
732
+ segments.push(seg);
733
+ }
734
+ }
685
735
  const silenceBytes = Math.round(this.sampleRate * 0.2) * 2;
686
736
  const silence = Buffer.alloc(silenceBytes);
687
737
  let prevLang = null;
688
738
  for (const segment of segments) {
689
739
  if (signal?.aborted) break;
690
- if (!segment.text.trim()) continue;
691
740
  const lang = this.models[segment.lang] ? segment.lang : this.defaultLang;
692
741
  if (prevLang !== null && lang !== prevLang) {
693
742
  yield silence;
@@ -815,10 +864,399 @@ var DeepgramTTS = class {
815
864
  return promise;
816
865
  }
817
866
  };
867
+
868
+ // src/providers/dtelecom-stt.ts
869
+ import WebSocket4 from "ws";
870
+ var log5 = createLogger("DtelecomSTT");
871
+ var KEEPALIVE_INTERVAL_MS2 = 5e3;
872
+ var DtelecomSTT = class {
873
+ options;
874
+ constructor(options) {
875
+ if (!options.serverUrl) {
876
+ throw new Error("DtelecomSTT requires a serverUrl");
877
+ }
878
+ this.options = options;
879
+ }
880
+ createStream(options) {
881
+ const language = options?.language ?? this.options.language ?? "auto";
882
+ return new DtelecomSTTStream(this.options, language);
883
+ }
884
+ };
885
+ var DtelecomSTTStream = class extends BaseSTTStream {
886
+ ws = null;
887
+ serverUrl;
888
+ forceWhisper;
889
+ _ready = false;
890
+ _closed = false;
891
+ pendingAudio = [];
892
+ keepAliveTimer = null;
893
+ language;
894
+ constructor(options, language) {
895
+ super();
896
+ this.serverUrl = options.serverUrl;
897
+ this.language = language;
898
+ this.forceWhisper = options.forceWhisper ?? false;
899
+ this.connect();
900
+ }
901
+ sendAudio(pcm16) {
902
+ if (this._closed) return;
903
+ if (!this._ready) {
904
+ this.pendingAudio.push(pcm16);
905
+ return;
906
+ }
907
+ if (this.ws?.readyState === WebSocket4.OPEN) {
908
+ this.ws.send(pcm16);
909
+ }
910
+ }
911
+ /**
912
+ * Switch language mid-session.
913
+ * Sends a reconfigure message to the server; clears buffers and updates model routing.
914
+ */
915
+ setLanguage(language, options) {
916
+ if (this._closed) return;
917
+ this.language = language;
918
+ const config = { type: "config", language };
919
+ if (options?.forceWhisper) {
920
+ config.model = "whisper";
921
+ }
922
+ if (this.ws?.readyState === WebSocket4.OPEN) {
923
+ this.ws.send(JSON.stringify(config));
924
+ log5.info(`Reconfiguring STT: language=${language}${options?.forceWhisper ? ", model=whisper" : ""}`);
925
+ }
926
+ }
927
+ async close() {
928
+ if (this._closed) return;
929
+ this._closed = true;
930
+ this._ready = false;
931
+ this.pendingAudio = [];
932
+ this.stopKeepAlive();
933
+ if (this.ws) {
934
+ this.ws.close();
935
+ this.ws = null;
936
+ }
937
+ log5.debug("DtelecomSTT stream closed");
938
+ }
939
+ connect() {
940
+ log5.debug(`Connecting to dTelecom STT: ${this.serverUrl}`);
941
+ this.ws = new WebSocket4(this.serverUrl);
942
+ this.ws.on("open", () => {
943
+ log5.info("dTelecom STT WebSocket connected");
944
+ const config = { type: "config", language: this.language };
945
+ if (this.forceWhisper) {
946
+ config.model = "whisper";
947
+ }
948
+ this.ws.send(JSON.stringify(config));
949
+ });
950
+ this.ws.on("message", (data, isBinary) => {
951
+ if (isBinary) return;
952
+ try {
953
+ const msg = JSON.parse(data.toString());
954
+ this.handleMessage(msg);
955
+ } catch (err) {
956
+ log5.error("Failed to parse dTelecom STT message:", err);
957
+ }
958
+ });
959
+ this.ws.on("error", (err) => {
960
+ log5.error("dTelecom STT WebSocket error:", err);
961
+ this.emit("error", err instanceof Error ? err : new Error(String(err)));
962
+ });
963
+ this.ws.on("close", (code, reason) => {
964
+ log5.debug(`dTelecom STT WebSocket closed: ${code} ${reason.toString()}`);
965
+ this._ready = false;
966
+ this.stopKeepAlive();
967
+ if (!this._closed) {
968
+ log5.info("dTelecom STT connection lost, reconnecting in 1s...");
969
+ setTimeout(() => {
970
+ if (!this._closed) this.connect();
971
+ }, 1e3);
972
+ }
973
+ });
974
+ }
975
+ handleMessage(msg) {
976
+ const type = msg.type;
977
+ if (type === "ready") {
978
+ this.handleReady(msg);
979
+ } else if (type === "transcription") {
980
+ this.handleTranscription(msg);
981
+ } else if (type === "vad_event") {
982
+ this.handleVadEvent(msg);
983
+ } else if (type === "pong") {
984
+ } else if (type === "error") {
985
+ const errData = msg.error;
986
+ const errorMsg = msg.message || (typeof errData === "string" ? errData : JSON.stringify(errData)) || "Unknown STT error";
987
+ log5.error(`dTelecom STT error: ${errorMsg}`);
988
+ this.emit("error", new Error(errorMsg));
989
+ }
990
+ }
991
+ handleReady(msg) {
992
+ const clientId = msg.client_id;
993
+ const lang = msg.language;
994
+ log5.info(`dTelecom STT ready: client_id=${clientId}, language=${lang}`);
995
+ this._ready = true;
996
+ for (const buf of this.pendingAudio) {
997
+ if (this.ws?.readyState === WebSocket4.OPEN) {
998
+ this.ws.send(buf);
999
+ }
1000
+ }
1001
+ this.pendingAudio = [];
1002
+ this.startKeepAlive();
1003
+ }
1004
+ handleTranscription(msg) {
1005
+ const text = msg.text ?? "";
1006
+ const isFinal = msg.is_final ?? false;
1007
+ const language = msg.language;
1008
+ const latencyMs = msg.latency_ms;
1009
+ if (!text) return;
1010
+ if (isFinal && latencyMs !== void 0) {
1011
+ log5.info(`stt_final: ${latencyMs.toFixed(0)}ms "${text.slice(0, 50)}"`);
1012
+ }
1013
+ this.emit("transcription", {
1014
+ text,
1015
+ isFinal,
1016
+ language,
1017
+ sttDuration: isFinal ? latencyMs : void 0
1018
+ });
1019
+ }
1020
+ handleVadEvent(msg) {
1021
+ const event = msg.event;
1022
+ log5.debug(`VAD event: ${event}`);
1023
+ if (event === "speech_start") {
1024
+ this.emit("transcription", {
1025
+ text: "",
1026
+ isFinal: false
1027
+ });
1028
+ }
1029
+ }
1030
+ startKeepAlive() {
1031
+ this.stopKeepAlive();
1032
+ this.keepAliveTimer = setInterval(() => {
1033
+ if (this.ws?.readyState === WebSocket4.OPEN) {
1034
+ this.ws.send(JSON.stringify({ type: "ping" }));
1035
+ }
1036
+ }, KEEPALIVE_INTERVAL_MS2);
1037
+ }
1038
+ stopKeepAlive() {
1039
+ if (this.keepAliveTimer) {
1040
+ clearInterval(this.keepAliveTimer);
1041
+ this.keepAliveTimer = null;
1042
+ }
1043
+ }
1044
+ };
1045
+
1046
+ // src/providers/dtelecom-tts.ts
1047
+ import WebSocket5 from "ws";
1048
+ var log6 = createLogger("DtelecomTTS");
1049
+ var DtelecomTTS = class {
1050
+ serverUrl;
1051
+ voices;
1052
+ defaultLang;
1053
+ speed;
1054
+ ws = null;
1055
+ connectPromise = null;
1056
+ flushState = null;
1057
+ /** Default language code for untagged text (e.g. 'en'). */
1058
+ get defaultLanguage() {
1059
+ return this.defaultLang;
1060
+ }
1061
+ constructor(options) {
1062
+ if (!options.serverUrl) {
1063
+ throw new Error("DtelecomTTS requires a serverUrl");
1064
+ }
1065
+ if (!options.voices || Object.keys(options.voices).length === 0) {
1066
+ throw new Error("DtelecomTTS requires at least one voice config");
1067
+ }
1068
+ this.serverUrl = options.serverUrl;
1069
+ this.voices = { ...options.voices };
1070
+ this.defaultLang = options.defaultLanguage ?? Object.keys(this.voices)[0];
1071
+ this.speed = options.speed ?? 1;
1072
+ }
1073
+ /** Pre-connect WebSocket to TTS server. */
1074
+ async warmup() {
1075
+ log6.info("Warming up TTS connection...");
1076
+ const start = performance.now();
1077
+ try {
1078
+ await this.ensureConnection();
1079
+ log6.info(`TTS warmup complete in ${(performance.now() - start).toFixed(0)}ms`);
1080
+ } catch (err) {
1081
+ log6.warn("TTS warmup failed (non-fatal):", err);
1082
+ }
1083
+ }
1084
+ /** Close WebSocket connection. */
1085
+ close() {
1086
+ if (this.ws) {
1087
+ log6.debug("Closing TTS WebSocket");
1088
+ this.ws.close();
1089
+ this.ws = null;
1090
+ }
1091
+ this.connectPromise = null;
1092
+ this.flushState = null;
1093
+ }
1094
+ /** Strip SSML lang tags from text for display/events. */
1095
+ cleanText(text) {
1096
+ return parseLangSegments(text, this.defaultLang).map((s) => s.text).join(" ").replace(/\s+/g, " ").trim();
1097
+ }
1098
+ async *synthesize(text, signal) {
1099
+ if (signal?.aborted) return;
1100
+ const rawSegments = parseLangSegments(text, this.defaultLang);
1101
+ const segments = [];
1102
+ for (const seg of rawSegments) {
1103
+ if (!seg.text.trim()) continue;
1104
+ if (!/\p{L}/u.test(seg.text) && segments.length > 0) {
1105
+ segments[segments.length - 1].text += seg.text;
1106
+ } else {
1107
+ segments.push(seg);
1108
+ }
1109
+ }
1110
+ const silenceBytes = Math.round(48e3 * 0.2) * 2;
1111
+ const silence = Buffer.alloc(silenceBytes);
1112
+ let prevLang = null;
1113
+ for (const segment of segments) {
1114
+ if (signal?.aborted) break;
1115
+ const lang = this.voices[segment.lang] ? segment.lang : this.defaultLang;
1116
+ if (prevLang !== null && lang !== prevLang) {
1117
+ yield silence;
1118
+ }
1119
+ prevLang = lang;
1120
+ yield* this.synthesizeSegment(lang, segment.text, signal);
1121
+ }
1122
+ }
1123
+ async *synthesizeSegment(lang, text, signal) {
1124
+ await this.ensureConnection();
1125
+ const ws = this.ws;
1126
+ if (!ws || ws.readyState !== WebSocket5.OPEN) {
1127
+ throw new Error("dTelecom TTS WebSocket not connected");
1128
+ }
1129
+ const state = { chunks: [], done: false, cleared: false, error: null, wake: null };
1130
+ this.flushState = state;
1131
+ const onAbort = () => {
1132
+ state.done = true;
1133
+ state.wake?.();
1134
+ if (ws.readyState === WebSocket5.OPEN) {
1135
+ try {
1136
+ ws.send(JSON.stringify({ type: "clear" }));
1137
+ } catch {
1138
+ }
1139
+ }
1140
+ };
1141
+ signal?.addEventListener("abort", onAbort, { once: true });
1142
+ const voiceConfig = this.voices[lang];
1143
+ const msg = { text };
1144
+ if (voiceConfig) {
1145
+ msg.voice = voiceConfig.voice;
1146
+ msg.lang_code = voiceConfig.langCode;
1147
+ msg.speed = this.speed;
1148
+ }
1149
+ log6.info(`TTS send [${lang}]: voice=${voiceConfig?.voice ?? "default"} lang_code=${voiceConfig?.langCode ?? "default"} "${text.slice(0, 60)}"`);
1150
+ ws.send(JSON.stringify(msg));
1151
+ try {
1152
+ while (true) {
1153
+ if (signal?.aborted) break;
1154
+ if (state.error) throw state.error;
1155
+ if (state.chunks.length > 0) {
1156
+ yield state.chunks.shift();
1157
+ continue;
1158
+ }
1159
+ if (state.done) break;
1160
+ await new Promise((resolve) => {
1161
+ state.wake = resolve;
1162
+ });
1163
+ state.wake = null;
1164
+ }
1165
+ while (state.chunks.length > 0) {
1166
+ yield state.chunks.shift();
1167
+ }
1168
+ } finally {
1169
+ signal?.removeEventListener("abort", onAbort);
1170
+ this.flushState = null;
1171
+ }
1172
+ }
1173
+ /** Ensure a WebSocket connection exists and is open. */
1174
+ ensureConnection() {
1175
+ if (this.ws && this.ws.readyState === WebSocket5.OPEN) {
1176
+ return Promise.resolve();
1177
+ }
1178
+ if (this.connectPromise) return this.connectPromise;
1179
+ this.connectPromise = new Promise((resolve, reject) => {
1180
+ log6.debug(`Connecting to dTelecom TTS: ${this.serverUrl}`);
1181
+ const ws = new WebSocket5(this.serverUrl);
1182
+ ws.on("open", () => {
1183
+ this.ws = ws;
1184
+ this.connectPromise = null;
1185
+ const defaultVoice = this.voices[this.defaultLang];
1186
+ if (defaultVoice) {
1187
+ ws.send(JSON.stringify({
1188
+ config: {
1189
+ voice: defaultVoice.voice,
1190
+ lang_code: defaultVoice.langCode,
1191
+ speed: this.speed
1192
+ }
1193
+ }));
1194
+ }
1195
+ log6.info("dTelecom TTS WebSocket connected");
1196
+ resolve();
1197
+ });
1198
+ ws.on("message", (data, isBinary) => {
1199
+ const state = this.flushState;
1200
+ if (!state) return;
1201
+ if (isBinary) {
1202
+ const buf = Buffer.isBuffer(data) ? data : Buffer.from(data);
1203
+ state.chunks.push(buf);
1204
+ state.wake?.();
1205
+ } else {
1206
+ try {
1207
+ const msg = JSON.parse(data.toString());
1208
+ if (msg.type === "done") {
1209
+ state.done = true;
1210
+ state.wake?.();
1211
+ } else if (msg.type === "cleared") {
1212
+ state.cleared = true;
1213
+ state.done = true;
1214
+ state.wake?.();
1215
+ } else if (msg.type === "generating") {
1216
+ log6.debug(`TTS generating: "${msg.text?.slice(0, 40)}"`);
1217
+ } else if (msg.type === "error") {
1218
+ const errorMsg = msg.message || "Unknown TTS error";
1219
+ log6.error(`dTelecom TTS error: ${errorMsg}`);
1220
+ state.error = new Error(errorMsg);
1221
+ state.wake?.();
1222
+ }
1223
+ } catch {
1224
+ log6.warn("Failed to parse dTelecom TTS message");
1225
+ }
1226
+ }
1227
+ });
1228
+ ws.on("error", (err) => {
1229
+ const error = err instanceof Error ? err : new Error(String(err));
1230
+ log6.error("dTelecom TTS WebSocket error:", error);
1231
+ const state = this.flushState;
1232
+ if (state) {
1233
+ state.error = error;
1234
+ state.wake?.();
1235
+ }
1236
+ this.ws = null;
1237
+ this.connectPromise = null;
1238
+ reject(error);
1239
+ });
1240
+ ws.on("close", (code, reason) => {
1241
+ log6.debug(`dTelecom TTS WebSocket closed: ${code} ${reason.toString()}`);
1242
+ this.ws = null;
1243
+ this.connectPromise = null;
1244
+ const state = this.flushState;
1245
+ if (state) {
1246
+ state.done = true;
1247
+ state.wake?.();
1248
+ }
1249
+ });
1250
+ });
1251
+ return this.connectPromise;
1252
+ }
1253
+ };
818
1254
  export {
819
1255
  CartesiaTTS,
820
1256
  DeepgramSTT,
821
1257
  DeepgramTTS,
1258
+ DtelecomSTT,
1259
+ DtelecomTTS,
822
1260
  OpenRouterLLM
823
1261
  };
824
1262
  //# sourceMappingURL=index.mjs.map