@openclaw/voice-call 2026.7.2-beta.5 → 2026.7.33

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.
Files changed (49) hide show
  1. package/README.md +3 -2
  2. package/dist/.setup/config-BBpso-fA.mjs +674 -0
  3. package/dist/.setup/config-compat-CYRE-SSo.mjs +156 -0
  4. package/dist/{guarded-json-api-Dxhe25Zi.js → .setup/guarded-json-api-BukKSVNW.mjs} +50 -24
  5. package/dist/{mock-YsvBTX-7.js → .setup/mock-DY4yTqR2.mjs} +1 -1
  6. package/dist/{plivo-B-KvPVji.js → .setup/plivo-DFHh72MH.mjs} +3 -2
  7. package/dist/{realtime-handler-hIVPEZxN.js → .setup/realtime-handler-DWLxvZuO.mjs} +197 -128
  8. package/dist/{response-generator-z6M9229t.js → .setup/response-generator-DoC6lfcm.mjs} +6 -21
  9. package/dist/{runtime-entry-CQtitvRd.js → .setup/runtime-entry-CH-CV0Sf.mjs} +384 -868
  10. package/dist/{store-path-nYL-yM0S.js → .setup/store-1UfyY3HS.mjs} +6 -13
  11. package/dist/{telnyx-BG56lL7z.js → .setup/telnyx-Bb8X-nin.mjs} +1 -1
  12. package/dist/{twilio-B3FQ_P9x.js → .setup/twilio-Bsbf3O9A.mjs} +11 -15
  13. package/dist/{webhook-replay-aaP79dI8.js → .setup/webhook-replay-g3cDA5wr.mjs} +1 -1
  14. package/dist/cli-metadata.js +1 -5
  15. package/dist/doctor-contract-api.js +14 -36
  16. package/dist/index.js +38 -28
  17. package/dist/runtime-entry.js +1 -1
  18. package/dist/setup-api.js +2 -86
  19. package/node_modules/commander/LICENSE +22 -0
  20. package/node_modules/commander/Readme.md +1172 -0
  21. package/node_modules/commander/index.js +21 -0
  22. package/node_modules/commander/lib/argument.js +147 -0
  23. package/node_modules/commander/lib/command.js +2790 -0
  24. package/node_modules/commander/lib/error.js +36 -0
  25. package/node_modules/commander/lib/help.js +731 -0
  26. package/node_modules/commander/lib/option.js +377 -0
  27. package/node_modules/commander/lib/suggestSimilar.js +99 -0
  28. package/node_modules/commander/package-support.json +19 -0
  29. package/node_modules/commander/package.json +64 -0
  30. package/node_modules/commander/typings/index.d.ts +1113 -0
  31. package/node_modules/typebox/build/compile/validator.d.mts +4 -1
  32. package/node_modules/typebox/build/guard/guard.d.mts +1 -1
  33. package/node_modules/typebox/build/guard/guard.mjs +2 -5
  34. package/node_modules/typebox/build/guard/string.mjs +4 -16
  35. package/node_modules/typebox/build/schema/engine/_functions.mjs +15 -34
  36. package/node_modules/typebox/build/type/action/readonly_object.d.mts +0 -5
  37. package/node_modules/typebox/build/type/action/readonly_object.mjs +0 -8
  38. package/node_modules/typebox/build/typebox.d.mts +1 -1
  39. package/node_modules/typebox/build/typebox.mjs +1 -1
  40. package/node_modules/typebox/package.json +1 -1
  41. package/node_modules/typebox/readme.md +32 -29
  42. package/node_modules/ws/lib/permessage-deflate.js +3 -1
  43. package/node_modules/ws/package.json +1 -1
  44. package/openclaw.plugin.json +15 -2
  45. package/package.json +8 -6
  46. package/dist/cli-output-mode.js +0 -27
  47. package/dist/twilio-region-4fkgz3UG.js +0 -25
  48. /package/dist/{realtime-transcription.runtime-CbJAs5t_.js → .setup/realtime-transcription.runtime-CbJAs5t_.mjs} +0 -0
  49. /package/dist/{realtime-voice.runtime-vtdCOWg-.js → .setup/realtime-voice.runtime-vtdCOWg-.mjs} +0 -0
@@ -1,659 +1,29 @@
1
- import { TtsConfigSchema, isBlockedHostnameOrIp, isRequestBodyLimitError, readRequestBodyWithLimit, requestBodyErrorToText } from "./runtime-api.js";
2
- import "./api.js";
3
- import { t as TWILIO_REGIONS } from "./twilio-region-4fkgz3UG.js";
4
- import { _ as TerminalStates, d as loadActiveCallsFromStore, g as setVoiceCallStateRuntime, l as findCallMatchesInStore, p as persistCallRecord, t as resolveDefaultVoiceCallStoreDir, u as getCallHistoryFromStore } from "./store-path-nYL-yM0S.js";
1
+ import { isBlockedHostnameOrIp, isRequestBodyLimitError, readRequestBodyWithLimit, requestBodyErrorToText } from "../runtime-api.js";
2
+ import "../api.js";
3
+ import { a as resolveVoiceCallEffectiveConfig, c as validateProviderConfig, d as deepMergeDefined, i as resolveVoiceCallConfig, n as normalizeVoiceCallConfig, o as resolveVoiceCallNumberRouteKeyForCall, r as resolveTwilioAuthToken, s as resolveVoiceCallSessionKey } from "./config-BBpso-fA.mjs";
4
+ import { c as findCallMatchesInStore, f as persistCallRecord, g as TerminalStates, h as setVoiceCallStateRuntime, l as getCallHistoryFromStore, u as loadActiveCallsFromStore } from "./store-1UfyY3HS.mjs";
5
5
  import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
6
6
  import { isLoopbackHost } from "openclaw/plugin-sdk/gateway-runtime";
7
7
  import { MAX_TIMER_TIMEOUT_MS, asDateTimestampMs, resolveExpiresAtMsFromDurationMs, resolveTimerTimeoutMs } from "openclaw/plugin-sdk/number-runtime";
8
- import { normalizeAgentId, parseAgentSessionKey } from "openclaw/plugin-sdk/routing";
8
+ import { normalizeAgentId } from "openclaw/plugin-sdk/routing";
9
9
  import { normalizeLowercaseStringOrEmpty, normalizeOptionalLowercaseString, normalizeOptionalString, normalizeStringEntries } from "openclaw/plugin-sdk/string-coerce-runtime";
10
10
  import { createSubsystemLogger } from "openclaw/plugin-sdk/runtime-env";
11
- import { resolveDefaultAgentId } from "openclaw/plugin-sdk/agent-runtime";
12
11
  import { createLazyRuntimeModule } from "openclaw/plugin-sdk/lazy-runtime";
13
- import { REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME, REALTIME_VOICE_AGENT_CONSULT_TOOL_POLICIES, assertRealtimeVoiceAgentConsultModelSelectionUnlocked, buildRealtimeVoiceAgentConsultPolicyInstructions, consultRealtimeVoiceAgent, convertPcmToMulaw8k, createTalkSessionController, recordTalkObservabilityEvent, resolveRealtimeVoiceAgentConsultTools, resolveRealtimeVoiceAgentConsultToolsAllow, resolveRealtimeVoiceFastContextConsult } from "openclaw/plugin-sdk/realtime-voice";
14
- import { mergeDeep } from "openclaw/plugin-sdk/plugin-config-runtime";
15
- import { buildSecretInputSchema, hasConfiguredSecretInput, normalizeResolvedSecretInputString } from "openclaw/plugin-sdk/secret-input";
16
- import { canonicalizeMainSessionAlias } from "openclaw/plugin-sdk/session-store-runtime";
17
- import { resolveSpeechProviderApiKey } from "openclaw/plugin-sdk/speech-core";
12
+ import { REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME, buildRealtimeVoiceAgentConsultPolicyInstructions, consultRealtimeVoiceAgent, convertPcmToMulaw8k, createTalkSessionController, recordTalkObservabilityEvent, resolveRealtimeVoiceAgentConsultTools, resolveRealtimeVoiceAgentConsultToolsAllow, resolveRealtimeVoiceFastContextConsult } from "openclaw/plugin-sdk/realtime-voice";
18
13
  import { WEBHOOK_BODY_READ_DEFAULTS, createWebhookInFlightLimiter, normalizeWebhookPath } from "openclaw/plugin-sdk/webhook-ingress";
19
- import { z } from "zod";
20
14
  import fs from "node:fs";
15
+ import os from "node:os";
16
+ import path from "node:path";
21
17
  import crypto from "node:crypto";
22
18
  import { redactIdentifier } from "openclaw/plugin-sdk/logging-core";
23
- import path from "node:path";
24
- import os from "node:os";
25
19
  import { root } from "openclaw/plugin-sdk/security-runtime";
26
20
  import { sliceUtf16Safe, truncateUtf16Safe } from "openclaw/plugin-sdk/text-utility-runtime";
21
+ import { parseTtsDirectives } from "openclaw/plugin-sdk/speech";
27
22
  import { spawn } from "node:child_process";
28
- import { runCommandWithTimeout } from "openclaw/plugin-sdk/process-runtime";
29
23
  import http from "node:http";
30
24
  import { URL as URL$1 } from "node:url";
31
25
  import { resolveConfiguredCapabilityProvider } from "openclaw/plugin-sdk/provider-selection-runtime";
32
26
  import { WebSocket, WebSocketServer } from "ws";
33
- import { canonicalizeBase64 } from "openclaw/plugin-sdk/media-runtime";
34
- //#region extensions/voice-call/src/realtime-defaults.ts
35
- /** Baseline instructions that keep realtime calls brief and route deep work to agent consult. */
36
- const DEFAULT_VOICE_CALL_REALTIME_INSTRUCTIONS = `You are OpenClaw's phone-call realtime voice interface. Keep spoken replies brief and natural. When a question needs deeper reasoning, current information, or tools, call ${REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME} before answering.`;
37
- //#endregion
38
- //#region extensions/voice-call/src/config.ts
39
- /**
40
- * E.164 phone number format: +[country code][number]
41
- * Examples use 555 prefix (reserved for fictional numbers)
42
- */
43
- const E164Schema = z.string().regex(/^\+[1-9]\d{1,14}$/, "Expected E.164 format, e.g. +15550001234");
44
- /**
45
- * Controls how inbound calls are handled:
46
- * - "disabled": Block all inbound calls (outbound only)
47
- * - "allowlist": Only accept calls from numbers in allowFrom
48
- * - "pairing": Unknown callers can request pairing (future)
49
- * - "open": Accept all inbound calls (dangerous!)
50
- */
51
- const InboundPolicySchema = z.enum([
52
- "disabled",
53
- "allowlist",
54
- "pairing",
55
- "open"
56
- ]);
57
- const SecretInputSchema = buildSecretInputSchema();
58
- const TelnyxConfigSchema = z.object({
59
- /** Telnyx API v2 key */
60
- apiKey: z.string().min(1).optional(),
61
- /** Telnyx connection ID (from Call Control app) */
62
- connectionId: z.string().min(1).optional(),
63
- /** Public key for webhook signature verification */
64
- publicKey: z.string().min(1).optional()
65
- }).strict();
66
- const TwilioConfigSchema = z.object({
67
- /** Twilio Account SID */
68
- accountSid: z.string().min(1).optional(),
69
- /** Twilio Auth Token */
70
- authToken: SecretInputSchema.optional(),
71
- /** Twilio processing Region (for example, ie1) */
72
- region: z.enum(TWILIO_REGIONS).optional()
73
- }).strict();
74
- const PlivoConfigSchema = z.object({
75
- /** Plivo Auth ID (starts with MA/SA) */
76
- authId: z.string().min(1).optional(),
77
- /** Plivo Auth Token */
78
- authToken: z.string().min(1).optional()
79
- }).strict();
80
- const VoiceCallNumberRouteConfigSchema = z.object({
81
- /** Greeting message for inbound calls to this number. */
82
- inboundGreeting: z.string().optional(),
83
- /** TTS override for inbound calls to this number. Deep-merges with global voice-call TTS. */
84
- tts: TtsConfigSchema,
85
- /** Agent ID to use for voice response generation for this number. */
86
- agentId: z.string().min(1).optional(),
87
- /** Optional model override for voice responses for this number. */
88
- responseModel: z.string().optional(),
89
- /** System prompt for voice responses for this number. */
90
- responseSystemPrompt: z.string().optional(),
91
- /** Timeout for response generation in ms for this number. */
92
- responseTimeoutMs: z.number().int().positive().optional()
93
- }).strict();
94
- const VoiceCallServeConfigSchema = z.object({
95
- /** Port to listen on */
96
- port: z.number().int().positive().default(3334),
97
- /** Bind address */
98
- bind: z.string().default("127.0.0.1"),
99
- /** Webhook path */
100
- path: z.string().min(1).default("/voice/webhook")
101
- }).strict().default({
102
- port: 3334,
103
- bind: "127.0.0.1",
104
- path: "/voice/webhook"
105
- });
106
- const VoiceCallTailscaleConfigSchema = z.object({
107
- /**
108
- * Tailscale exposure mode:
109
- * - "off": No Tailscale exposure
110
- * - "serve": Tailscale serve (private to tailnet)
111
- * - "funnel": Tailscale funnel (public HTTPS)
112
- */
113
- mode: z.enum([
114
- "off",
115
- "serve",
116
- "funnel"
117
- ]).default("off"),
118
- /** Path for Tailscale serve/funnel (should usually match serve.path) */
119
- path: z.string().min(1).default("/voice/webhook")
120
- }).strict().default({
121
- mode: "off",
122
- path: "/voice/webhook"
123
- });
124
- const VoiceCallTunnelConfigSchema = z.object({
125
- /**
126
- * Tunnel provider:
127
- * - "none": No tunnel (use publicUrl if set, or manual setup)
128
- * - "ngrok": Use ngrok for public HTTPS tunnel
129
- * - "tailscale-serve": Tailscale serve (private to tailnet)
130
- * - "tailscale-funnel": Tailscale funnel (public HTTPS)
131
- */
132
- provider: z.enum([
133
- "none",
134
- "ngrok",
135
- "tailscale-serve",
136
- "tailscale-funnel"
137
- ]).default("none"),
138
- /** ngrok auth token (optional, enables longer sessions and more features) */
139
- ngrokAuthToken: z.string().min(1).optional(),
140
- /** ngrok custom domain (paid feature, e.g., "myapp.ngrok.io") */
141
- ngrokDomain: z.string().min(1).optional(),
142
- /**
143
- * Allow ngrok free tier compatibility mode.
144
- * When true, forwarded headers may be trusted for loopback requests
145
- * to reconstruct the public ngrok URL used for signing.
146
- *
147
- * IMPORTANT: This does NOT bypass signature verification.
148
- */
149
- allowNgrokFreeTierLoopbackBypass: z.boolean().default(false)
150
- }).strict().default({
151
- provider: "none",
152
- allowNgrokFreeTierLoopbackBypass: false
153
- });
154
- const VoiceCallWebhookSecurityConfigSchema = z.object({
155
- /**
156
- * Allowed hostnames for webhook URL reconstruction.
157
- * Only these hosts are accepted from forwarding headers.
158
- */
159
- allowedHosts: z.array(z.string().min(1)).default([]),
160
- /**
161
- * Trust X-Forwarded-* headers without a hostname allowlist.
162
- * WARNING: Only enable if you trust your proxy configuration.
163
- */
164
- trustForwardingHeaders: z.boolean().default(false),
165
- /**
166
- * Trusted proxy IP addresses. Forwarded headers are only trusted when
167
- * the remote IP matches one of these addresses.
168
- */
169
- trustedProxyIPs: z.array(z.string().min(1)).default([])
170
- }).strict().default({
171
- allowedHosts: [],
172
- trustForwardingHeaders: false,
173
- trustedProxyIPs: []
174
- });
175
- /**
176
- * Call mode determines how outbound calls behave:
177
- * - "notify": Deliver message and auto-hangup after delay (one-way notification)
178
- * - "conversation": Stay open for back-and-forth until explicit end or timeout
179
- */
180
- const CallModeSchema = z.enum(["notify", "conversation"]);
181
- const VoiceCallSessionScopeSchema = z.enum(["per-phone", "per-call"]);
182
- const OutboundConfigSchema = z.object({
183
- /** Default call mode for outbound calls */
184
- defaultMode: CallModeSchema.default("notify"),
185
- /** Seconds to wait after TTS before auto-hangup in notify mode */
186
- notifyHangupDelaySec: z.number().int().nonnegative().default(3)
187
- }).strict().default({
188
- defaultMode: "notify",
189
- notifyHangupDelaySec: 3
190
- });
191
- const RealtimeToolSchema = z.object({
192
- type: z.literal("function"),
193
- name: z.string().min(1),
194
- description: z.string(),
195
- parameters: z.object({
196
- type: z.literal("object"),
197
- properties: z.record(z.string(), z.unknown()),
198
- required: z.array(z.string()).optional()
199
- })
200
- }).strict();
201
- const VoiceCallRealtimeProvidersConfigSchema = z.record(z.string(), z.record(z.string(), z.unknown())).default({});
202
- const VoiceCallRealtimeToolPolicySchema = z.enum(REALTIME_VOICE_AGENT_CONSULT_TOOL_POLICIES);
203
- const VoiceCallRealtimeConsultPolicySchema = z.enum([
204
- "auto",
205
- "substantive",
206
- "always"
207
- ]);
208
- const VoiceCallRealtimeFastContextSourceSchema = z.enum(["memory", "sessions"]);
209
- const VoiceCallRealtimeFastContextConfigSchema = z.object({
210
- /** Enable bounded memory/session lookup before the full consult agent. */
211
- enabled: z.boolean().default(false),
212
- /** Hard deadline for the fast context lookup. */
213
- timeoutMs: z.number().int().positive().default(800),
214
- /** Maximum memory/session hits to inject into the realtime tool result. */
215
- maxResults: z.number().int().positive().default(3),
216
- /** Indexed sources used by the fast context lookup. */
217
- sources: z.array(VoiceCallRealtimeFastContextSourceSchema).min(1).default(["memory", "sessions"]),
218
- /** Fall back to the full agent consult when fast context has no answer. */
219
- fallbackToConsult: z.boolean().default(false)
220
- }).strict().default({
221
- enabled: false,
222
- timeoutMs: 800,
223
- maxResults: 3,
224
- sources: ["memory", "sessions"],
225
- fallbackToConsult: false
226
- });
227
- const VoiceCallRealtimeAgentContextConfigSchema = z.object({
228
- /** Inject a compact agent persona/context capsule into realtime voice instructions. */
229
- enabled: z.boolean().default(false),
230
- /** Maximum number of characters from the generated capsule to append. */
231
- maxChars: z.number().int().positive().default(6e3),
232
- /** Include configured agent identity fields. */
233
- includeIdentity: z.boolean().default(true),
234
- /** Include selected workspace files such as SOUL.md and IDENTITY.md. */
235
- includeWorkspaceFiles: z.boolean().default(true),
236
- /** Workspace-relative files to include, bounded by maxChars. */
237
- files: z.array(z.string().min(1)).default([
238
- "SOUL.md",
239
- "IDENTITY.md",
240
- "USER.md"
241
- ])
242
- }).strict().default({
243
- enabled: false,
244
- maxChars: 6e3,
245
- includeIdentity: true,
246
- includeWorkspaceFiles: true,
247
- files: [
248
- "SOUL.md",
249
- "IDENTITY.md",
250
- "USER.md"
251
- ]
252
- });
253
- const VoiceCallRealtimeConsultThinkingLevelSchema = z.enum([
254
- "off",
255
- "minimal",
256
- "low",
257
- "medium",
258
- "high",
259
- "xhigh",
260
- "adaptive",
261
- "max",
262
- "ultra"
263
- ]);
264
- const VoiceCallStreamingProvidersConfigSchema = z.record(z.string(), z.record(z.string(), z.unknown())).default({});
265
- const VoiceCallRealtimeConfigSchema = z.object({
266
- /** Enable realtime voice-to-voice mode. */
267
- enabled: z.boolean().default(false),
268
- /** Provider id from registered realtime voice providers. */
269
- provider: z.string().min(1).optional(),
270
- /** Optional override for the local WebSocket route path. */
271
- streamPath: z.string().min(1).optional(),
272
- /** System instructions passed to the realtime provider. */
273
- instructions: z.string().default(DEFAULT_VOICE_CALL_REALTIME_INSTRUCTIONS),
274
- /** Tool policy for the shared OpenClaw agent consult tool. */
275
- toolPolicy: VoiceCallRealtimeToolPolicySchema.default("safe-read-only"),
276
- /** Guidance for when the realtime model should call the OpenClaw agent consult tool. */
277
- consultPolicy: VoiceCallRealtimeConsultPolicySchema.default("auto"),
278
- /** Optional thinking level override for the regular agent behind realtime consults. */
279
- consultThinkingLevel: VoiceCallRealtimeConsultThinkingLevelSchema.optional(),
280
- /** Optional fast mode override for the regular agent behind realtime consults. */
281
- consultFastMode: z.boolean().optional(),
282
- /** Tool definitions exposed to the realtime provider. */
283
- tools: z.array(RealtimeToolSchema).default([]),
284
- /** Low-latency memory/session context for the consult tool. */
285
- fastContext: VoiceCallRealtimeFastContextConfigSchema,
286
- /** Bounded agent persona/context injection for the fast realtime voice path. */
287
- agentContext: VoiceCallRealtimeAgentContextConfigSchema,
288
- /** Provider-owned raw config blobs keyed by provider id. */
289
- providers: VoiceCallRealtimeProvidersConfigSchema
290
- }).strict().default({
291
- enabled: false,
292
- instructions: DEFAULT_VOICE_CALL_REALTIME_INSTRUCTIONS,
293
- toolPolicy: "safe-read-only",
294
- consultPolicy: "auto",
295
- tools: [],
296
- fastContext: {
297
- enabled: false,
298
- timeoutMs: 800,
299
- maxResults: 3,
300
- sources: ["memory", "sessions"],
301
- fallbackToConsult: false
302
- },
303
- agentContext: {
304
- enabled: false,
305
- maxChars: 6e3,
306
- includeIdentity: true,
307
- includeWorkspaceFiles: true,
308
- files: [
309
- "SOUL.md",
310
- "IDENTITY.md",
311
- "USER.md"
312
- ]
313
- },
314
- providers: {}
315
- });
316
- const VoiceCallStreamingConfigSchema = z.object({
317
- /** Enable Twilio Media Streams for real-time transcription. */
318
- enabled: z.boolean().default(false),
319
- /** Provider id from registered realtime transcription providers. */
320
- provider: z.string().min(1).optional(),
321
- /** WebSocket path for media stream connections */
322
- streamPath: z.string().min(1).default("/voice/stream"),
323
- /** Provider-owned raw config blobs keyed by provider id. */
324
- providers: VoiceCallStreamingProvidersConfigSchema,
325
- /**
326
- * Close unauthenticated media stream sockets if no valid `start` frame arrives in time.
327
- * Protects against pre-auth idle connection hold attacks.
328
- */
329
- preStartTimeoutMs: z.number().int().positive().default(5e3),
330
- /** Maximum number of concurrently pending (pre-start) media stream sockets. */
331
- maxPendingConnections: z.number().int().positive().default(32),
332
- /** Maximum pending media stream sockets per source IP. */
333
- maxPendingConnectionsPerIp: z.number().int().positive().default(4),
334
- /** Hard cap for all open media stream sockets (pending + active). */
335
- maxConnections: z.number().int().positive().default(128)
336
- }).strict().default({
337
- enabled: false,
338
- streamPath: "/voice/stream",
339
- providers: {},
340
- preStartTimeoutMs: 5e3,
341
- maxPendingConnections: 32,
342
- maxPendingConnectionsPerIp: 4,
343
- maxConnections: 128
344
- });
345
- const VoiceCallConfigSchema = z.object({
346
- /** Enable voice call functionality */
347
- enabled: z.boolean().default(false),
348
- /** Active provider (telnyx, twilio, plivo, or mock) */
349
- provider: z.enum([
350
- "telnyx",
351
- "twilio",
352
- "plivo",
353
- "mock"
354
- ]).optional(),
355
- /** Telnyx-specific configuration */
356
- telnyx: TelnyxConfigSchema.optional(),
357
- /** Twilio-specific configuration */
358
- twilio: TwilioConfigSchema.optional(),
359
- /** Plivo-specific configuration */
360
- plivo: PlivoConfigSchema.optional(),
361
- /** Phone number to call from (E.164) */
362
- fromNumber: E164Schema.optional(),
363
- /** Default phone number to call (E.164) */
364
- toNumber: E164Schema.optional(),
365
- /** Inbound call policy */
366
- inboundPolicy: InboundPolicySchema.default("disabled"),
367
- /** Allowlist of phone numbers for inbound calls (E.164) */
368
- allowFrom: z.array(E164Schema).default([]),
369
- /** Greeting message for inbound calls */
370
- inboundGreeting: z.string().optional(),
371
- /** Per-dialed-number overrides for inbound calls. Keys are E.164 numbers. */
372
- numbers: z.record(E164Schema, VoiceCallNumberRouteConfigSchema).default({}),
373
- /** Outbound call configuration */
374
- outbound: OutboundConfigSchema,
375
- /** Maximum call duration in seconds */
376
- maxDurationSeconds: z.number().int().positive().default(300),
377
- /**
378
- * Maximum age of a call in seconds before it is automatically reaped.
379
- * Catches calls stuck before answer (for example, local mock calls that
380
- * never receive provider webhooks). Set to 0 to disable.
381
- */
382
- staleCallReaperSeconds: z.number().int().nonnegative().default(120),
383
- /** Silence timeout for end-of-speech detection (ms) */
384
- silenceTimeoutMs: z.number().int().positive().default(800),
385
- /** Timeout for user transcript (ms) */
386
- transcriptTimeoutMs: z.number().int().positive().default(18e4),
387
- /** Ring timeout for outbound calls (ms) */
388
- ringTimeoutMs: z.number().int().positive().default(3e4),
389
- /** Maximum concurrent calls */
390
- maxConcurrentCalls: z.number().int().positive().default(1),
391
- /** Webhook server configuration */
392
- serve: VoiceCallServeConfigSchema,
393
- /** @deprecated Prefer tunnel config. */
394
- tailscale: VoiceCallTailscaleConfigSchema,
395
- /** Tunnel configuration (unified ngrok/tailscale) */
396
- tunnel: VoiceCallTunnelConfigSchema,
397
- /** Webhook signature reconstruction and proxy trust configuration */
398
- webhookSecurity: VoiceCallWebhookSecurityConfigSchema,
399
- /** Real-time audio streaming configuration */
400
- streaming: VoiceCallStreamingConfigSchema,
401
- /** Realtime voice-to-voice configuration */
402
- realtime: VoiceCallRealtimeConfigSchema,
403
- /** Session memory scope for voice conversations. */
404
- sessionScope: VoiceCallSessionScopeSchema.default("per-phone"),
405
- /** Public webhook URL override (if set, bypasses tunnel auto-detection) */
406
- publicUrl: z.string().url().optional(),
407
- /** Skip webhook signature verification (development only, NOT for production) */
408
- skipSignatureVerification: z.boolean().default(false),
409
- /** TTS override (deep-merges with core tts) */
410
- tts: TtsConfigSchema,
411
- /** Store path for call logs */
412
- store: z.string().optional(),
413
- /** Agent ID to use for voice response generation. Defaults to "main". */
414
- agentId: z.string().min(1).optional(),
415
- /** Optional model override for generating voice responses. */
416
- responseModel: z.string().optional(),
417
- /** System prompt for voice responses */
418
- responseSystemPrompt: z.string().optional(),
419
- /** Timeout for response generation in ms (default 30s) */
420
- responseTimeoutMs: z.number().int().positive().default(3e4)
421
- }).strict();
422
- const TWILIO_AUTH_TOKEN_PATH = "plugins.entries.voice-call.config.twilio.authToken";
423
- const DEFAULT_VOICE_CALL_CONFIG = VoiceCallConfigSchema.parse({});
424
- function cloneDefaultVoiceCallConfig() {
425
- return structuredClone(DEFAULT_VOICE_CALL_CONFIG);
426
- }
427
- function defaultRealtimeStreamPathForServePath(servePath) {
428
- const normalized = normalizeWebhookPath(servePath);
429
- if (normalized.endsWith("/webhook")) return `${normalized.slice(0, -8)}/stream/realtime`;
430
- if (normalized === "/") return "/voice/stream/realtime";
431
- return `${normalized}/stream/realtime`;
432
- }
433
- function normalizeVoiceCallTtsConfig(defaults, overrides) {
434
- if (!defaults && !overrides) return;
435
- return TtsConfigSchema.parse(mergeDeep(defaults ?? {}, overrides ?? {}));
436
- }
437
- function normalizePhoneRouteKey(phone) {
438
- return phone?.replace(/\D/g, "") ?? "";
439
- }
440
- function resolveVoiceCallNumberRouteKey(config, phone) {
441
- const routes = config.numbers;
442
- if (!routes) return;
443
- if (phone && Object.hasOwn(routes, phone)) return phone;
444
- const normalizedPhone = normalizePhoneRouteKey(phone);
445
- if (!normalizedPhone) return;
446
- return Object.keys(routes).find((routeKey) => normalizePhoneRouteKey(routeKey) === normalizedPhone);
447
- }
448
- /** Resolve inbound-only number routing from a persisted call record. */
449
- function resolveVoiceCallNumberRouteKeyForCall(call) {
450
- if (call.direction !== "inbound") return;
451
- const storedRouteKey = call.metadata?.numberRouteKey;
452
- if (typeof storedRouteKey === "string") return storedRouteKey;
453
- return call.to;
454
- }
455
- function resolveVoiceCallEffectiveConfig(config, phoneOrRouteKey) {
456
- const numberRouteKey = resolveVoiceCallNumberRouteKey(config, phoneOrRouteKey);
457
- if (!numberRouteKey) return { config };
458
- const route = config.numbers[numberRouteKey];
459
- if (!route) return { config };
460
- return {
461
- numberRouteKey,
462
- config: {
463
- ...config,
464
- ...route,
465
- tts: normalizeVoiceCallTtsConfig(config.tts, route.tts),
466
- numbers: config.numbers
467
- }
468
- };
469
- }
470
- function sanitizeVoiceCallProviderConfigs(value) {
471
- if (!value) return {};
472
- return Object.fromEntries(Object.entries(value).filter((entry) => entry[1] !== void 0));
473
- }
474
- function sanitizeVoiceCallNumberRoutes(value) {
475
- if (!value) return {};
476
- return Object.fromEntries(Object.entries(value).filter((entry) => entry[1] !== void 0).map(([key, route]) => [key, VoiceCallNumberRouteConfigSchema.parse(route)]));
477
- }
478
- function resolveTwilioAuthToken(config) {
479
- return normalizeResolvedSecretInputString({
480
- value: config.twilio?.authToken,
481
- path: TWILIO_AUTH_TOKEN_PATH
482
- });
483
- }
484
- function normalizeVoiceCallConfig(config) {
485
- const defaults = cloneDefaultVoiceCallConfig();
486
- const serve = {
487
- ...defaults.serve,
488
- ...config.serve
489
- };
490
- const streamingProvider = config.streaming?.provider;
491
- const streamingProviders = sanitizeVoiceCallProviderConfigs(config.streaming?.providers ?? defaults.streaming.providers);
492
- const realtimeProvider = config.realtime?.provider ?? defaults.realtime.provider;
493
- const realtimeProviders = sanitizeVoiceCallProviderConfigs(config.realtime?.providers ?? defaults.realtime.providers);
494
- const realtimeFastContext = {
495
- ...defaults.realtime.fastContext,
496
- ...config.realtime?.fastContext,
497
- sources: config.realtime?.fastContext?.sources ?? defaults.realtime.fastContext.sources
498
- };
499
- const realtimeAgentContext = {
500
- ...defaults.realtime.agentContext,
501
- ...config.realtime?.agentContext,
502
- files: config.realtime?.agentContext?.files ?? defaults.realtime.agentContext.files
503
- };
504
- return {
505
- ...defaults,
506
- ...config,
507
- allowFrom: config.allowFrom ?? defaults.allowFrom,
508
- numbers: sanitizeVoiceCallNumberRoutes(config.numbers ?? defaults.numbers),
509
- outbound: {
510
- ...defaults.outbound,
511
- ...config.outbound
512
- },
513
- serve,
514
- tailscale: {
515
- ...defaults.tailscale,
516
- ...config.tailscale
517
- },
518
- tunnel: {
519
- ...defaults.tunnel,
520
- ...config.tunnel
521
- },
522
- webhookSecurity: {
523
- ...defaults.webhookSecurity,
524
- ...config.webhookSecurity,
525
- allowedHosts: config.webhookSecurity?.allowedHosts ?? defaults.webhookSecurity.allowedHosts,
526
- trustedProxyIPs: config.webhookSecurity?.trustedProxyIPs ?? defaults.webhookSecurity.trustedProxyIPs
527
- },
528
- streaming: {
529
- ...defaults.streaming,
530
- ...config.streaming,
531
- provider: streamingProvider,
532
- providers: streamingProviders
533
- },
534
- realtime: {
535
- ...defaults.realtime,
536
- ...config.realtime,
537
- provider: realtimeProvider,
538
- streamPath: config.realtime?.streamPath ?? defaultRealtimeStreamPathForServePath(serve.path ?? defaults.serve.path),
539
- tools: config.realtime?.tools ?? defaults.realtime.tools,
540
- consultThinkingLevel: VoiceCallRealtimeConsultThinkingLevelSchema.optional().parse(config.realtime?.consultThinkingLevel ?? defaults.realtime.consultThinkingLevel),
541
- consultFastMode: config.realtime?.consultFastMode ?? defaults.realtime.consultFastMode,
542
- fastContext: realtimeFastContext,
543
- agentContext: realtimeAgentContext,
544
- providers: realtimeProviders
545
- },
546
- tts: normalizeVoiceCallTtsConfig(defaults.tts, config.tts)
547
- };
548
- }
549
- function resolveVoiceCallSessionKey(params) {
550
- const explicit = params.explicitSessionKey?.trim();
551
- if (explicit) return resolveVoiceCallAgentSessionKey({
552
- config: params.config,
553
- sessionKey: explicit,
554
- coreSession: params.coreSession
555
- });
556
- const prefix = `agent:${normalizeAgentId(params.config.agentId)}:voice`;
557
- if (params.config.sessionScope === "per-call") return `${prefix}:call:${params.callId}`.toLowerCase();
558
- const normalizedPhone = params.phone?.replace(/\D/g, "");
559
- return (normalizedPhone ? `${prefix}:${normalizedPhone}` : `${prefix}:${params.callId}`).toLowerCase();
560
- }
561
- /** Resolve persisted or integration-provided keys into the configured agent namespace. */
562
- function resolveVoiceCallAgentSessionKey(params) {
563
- const sessionKey = params.sessionKey.trim();
564
- if (!sessionKey) throw new Error("Voice Call session key cannot be empty");
565
- const lower = sessionKey.toLowerCase();
566
- const agentId = normalizeAgentId(params.config.agentId);
567
- if (lower === "global" || lower === "unknown") return lower;
568
- const parsedInput = parseAgentSessionKey(sessionKey);
569
- let normalizedScopedKey;
570
- if (parsedInput && normalizeAgentId(parsedInput.agentId) === parsedInput.agentId && parsedInput.agentId === agentId) normalizedScopedKey = `agent:${parsedInput.agentId}:${parsedInput.rest}`;
571
- else {
572
- const wrappedInput = parseAgentSessionKey(`agent:${agentId}:${sessionKey}`);
573
- if (!wrappedInput) throw new Error("Voice Call session key could not be normalized");
574
- normalizedScopedKey = `agent:${agentId}:${wrappedInput.rest}`;
575
- }
576
- const canonicalMain = canonicalizeMainSessionAlias({
577
- cfg: { session: params.coreSession },
578
- agentId,
579
- sessionKey: normalizedScopedKey
580
- });
581
- return canonicalMain === normalizedScopedKey ? normalizedScopedKey : canonicalMain;
582
- }
583
- /**
584
- * Resolves the configuration by merging environment variables into missing fields.
585
- * Returns a new configuration object with environment variables applied.
586
- */
587
- function resolveVoiceCallConfig(config) {
588
- const resolved = normalizeVoiceCallConfig(config);
589
- if (resolved.provider === "telnyx") {
590
- resolved.telnyx = resolved.telnyx ?? {};
591
- resolved.telnyx.apiKey = resolved.telnyx.apiKey ?? resolveSpeechProviderApiKey(process.env.TELNYX_API_KEY);
592
- resolved.telnyx.connectionId = resolved.telnyx.connectionId ?? resolveSpeechProviderApiKey(process.env.TELNYX_CONNECTION_ID);
593
- resolved.telnyx.publicKey = resolved.telnyx.publicKey ?? resolveSpeechProviderApiKey(process.env.TELNYX_PUBLIC_KEY);
594
- }
595
- if (resolved.provider === "twilio") {
596
- resolved.fromNumber = resolved.fromNumber ?? resolveSpeechProviderApiKey(process.env.TWILIO_FROM_NUMBER);
597
- resolved.twilio = resolved.twilio ?? {};
598
- resolved.twilio.accountSid = resolved.twilio.accountSid ?? resolveSpeechProviderApiKey(process.env.TWILIO_ACCOUNT_SID);
599
- resolved.twilio.authToken = resolved.twilio.authToken ?? resolveSpeechProviderApiKey(process.env.TWILIO_AUTH_TOKEN);
600
- }
601
- if (resolved.provider === "plivo") {
602
- resolved.plivo = resolved.plivo ?? {};
603
- resolved.plivo.authId = resolved.plivo.authId ?? resolveSpeechProviderApiKey(process.env.PLIVO_AUTH_ID);
604
- resolved.plivo.authToken = resolved.plivo.authToken ?? resolveSpeechProviderApiKey(process.env.PLIVO_AUTH_TOKEN);
605
- }
606
- resolved.tunnel = resolved.tunnel ?? {
607
- provider: "none",
608
- allowNgrokFreeTierLoopbackBypass: false
609
- };
610
- resolved.tunnel.allowNgrokFreeTierLoopbackBypass = resolved.tunnel.allowNgrokFreeTierLoopbackBypass ?? false;
611
- resolved.tunnel.ngrokAuthToken = resolved.tunnel.ngrokAuthToken ?? resolveSpeechProviderApiKey(process.env.NGROK_AUTHTOKEN);
612
- resolved.tunnel.ngrokDomain = resolved.tunnel.ngrokDomain ?? resolveSpeechProviderApiKey(process.env.NGROK_DOMAIN);
613
- resolved.webhookSecurity = resolved.webhookSecurity ?? {
614
- allowedHosts: [],
615
- trustForwardingHeaders: false,
616
- trustedProxyIPs: []
617
- };
618
- resolved.webhookSecurity.allowedHosts = resolved.webhookSecurity.allowedHosts ?? [];
619
- resolved.webhookSecurity.trustForwardingHeaders = resolved.webhookSecurity.trustForwardingHeaders ?? false;
620
- resolved.webhookSecurity.trustedProxyIPs = resolved.webhookSecurity.trustedProxyIPs ?? [];
621
- return normalizeVoiceCallConfig(resolved);
622
- }
623
- /**
624
- * Validate that the configuration has all required fields for the selected provider.
625
- */
626
- function validateProviderConfig(config) {
627
- const errors = [];
628
- if (!config.enabled) return {
629
- valid: true,
630
- errors: []
631
- };
632
- if (!config.provider) errors.push("plugins.entries.voice-call.config.provider is required");
633
- if (!config.fromNumber && config.provider !== "mock") errors.push(config.provider === "twilio" ? "plugins.entries.voice-call.config.fromNumber is required (or set TWILIO_FROM_NUMBER env)" : "plugins.entries.voice-call.config.fromNumber is required");
634
- if (config.provider === "telnyx") {
635
- if (!config.telnyx?.apiKey) errors.push("plugins.entries.voice-call.config.telnyx.apiKey is required (or set TELNYX_API_KEY env)");
636
- if (!config.telnyx?.connectionId) errors.push("plugins.entries.voice-call.config.telnyx.connectionId is required (or set TELNYX_CONNECTION_ID env)");
637
- if (!config.skipSignatureVerification && !config.telnyx?.publicKey) errors.push("plugins.entries.voice-call.config.telnyx.publicKey is required (or set TELNYX_PUBLIC_KEY env)");
638
- }
639
- if (config.provider === "twilio") {
640
- if (!config.twilio?.accountSid) errors.push("plugins.entries.voice-call.config.twilio.accountSid is required (or set TWILIO_ACCOUNT_SID env)");
641
- if (!hasConfiguredSecretInput(config.twilio?.authToken)) errors.push("plugins.entries.voice-call.config.twilio.authToken is required (or set TWILIO_AUTH_TOKEN env)");
642
- }
643
- if (config.provider === "plivo") {
644
- if (!config.plivo?.authId) errors.push("plugins.entries.voice-call.config.plivo.authId is required (or set PLIVO_AUTH_ID env)");
645
- if (!config.plivo?.authToken) errors.push("plugins.entries.voice-call.config.plivo.authToken is required (or set PLIVO_AUTH_TOKEN env)");
646
- }
647
- if (config.realtime.enabled && config.inboundPolicy === "disabled") errors.push("plugins.entries.voice-call.config.inboundPolicy must not be \"disabled\" when realtime.enabled is true");
648
- if (config.realtime.enabled && config.streaming.enabled) errors.push("plugins.entries.voice-call.config.realtime.enabled and plugins.entries.voice-call.config.streaming.enabled cannot both be true");
649
- if (config.streaming.enabled && config.provider && config.provider !== "twilio") errors.push("plugins.entries.voice-call.config.provider must be \"twilio\" when streaming.enabled is true");
650
- if (config.realtime.enabled && config.provider && config.provider !== "twilio" && config.provider !== "telnyx" && config.provider !== "mock") errors.push("plugins.entries.voice-call.config.provider must be \"twilio\", \"telnyx\", or \"mock\" when realtime.enabled is true");
651
- return {
652
- valid: errors.length === 0,
653
- errors
654
- };
655
- }
656
- //#endregion
657
27
  //#region extensions/voice-call/src/allowlist.ts
658
28
  /** Normalize a phone number to digits only. */
659
29
  function normalizePhoneNumber(input) {
@@ -876,7 +246,7 @@ const DEFAULT_POLLY_VOICE = "Polly.Joanna";
876
246
  function mapVoiceToPolly(voice) {
877
247
  if (!voice) return DEFAULT_POLLY_VOICE;
878
248
  if (voice.startsWith("Polly.") || voice.startsWith("Google.")) return voice;
879
- return OPENAI_TO_POLLY_MAP[normalizeLowercaseStringOrEmpty(voice)] || DEFAULT_POLLY_VOICE;
249
+ return OPENAI_TO_POLLY_MAP[normalizeLowercaseStringOrEmpty(voice)] || "Polly.Joanna";
880
250
  }
881
251
  //#endregion
882
252
  //#region extensions/voice-call/src/manager/twiml.ts
@@ -1509,18 +879,16 @@ function processEvent(ctx, event) {
1509
879
  endedAt: event.timestamp
1510
880
  });
1511
881
  return { kind: "processed" };
1512
- case "call.error":
1513
- if (!event.retryable) {
1514
- finalizeCall({
1515
- ctx,
1516
- call,
1517
- endReason: "error",
1518
- endedAt: event.timestamp,
1519
- transcriptRejectReason: `Call error: ${event.error}`
1520
- });
1521
- return { kind: "processed" };
1522
- }
1523
- break;
882
+ case "call.error": if (!event.retryable) {
883
+ finalizeCall({
884
+ ctx,
885
+ call,
886
+ endReason: "error",
887
+ endedAt: event.timestamp,
888
+ transcriptRejectReason: `Call error: ${event.error}`
889
+ });
890
+ return { kind: "processed" };
891
+ }
1524
892
  }
1525
893
  persistCallRecord(ctx.storePath, call);
1526
894
  return result;
@@ -1554,7 +922,14 @@ function resolveRestoredMaxDurationAnchor(call) {
1554
922
  function resolveDefaultStoreBase(config, storePath) {
1555
923
  const rawOverride = storePath?.trim() || config.store?.trim();
1556
924
  if (rawOverride) return resolveUserPath(rawOverride);
1557
- return resolveDefaultVoiceCallStoreDir();
925
+ const preferred = path.join(os.homedir(), ".openclaw", "voice-calls");
926
+ return [preferred].map((dir) => resolveUserPath(dir)).find((dir) => {
927
+ try {
928
+ return fs.existsSync(path.join(dir, "calls.jsonl")) || fs.existsSync(dir);
929
+ } catch {
930
+ return false;
931
+ }
932
+ }) ?? resolveUserPath(preferred);
1558
933
  }
1559
934
  /**
1560
935
  * Manages voice calls: state ownership and delegation to manager helper modules.
@@ -1840,7 +1215,7 @@ async function buildRealtimeVoiceInstructions(params) {
1840
1215
  if (consultGuidance) sections.push(consultGuidance);
1841
1216
  const contextConfig = config.realtime.agentContext;
1842
1217
  if (!contextConfig.enabled) return sections.filter(Boolean).join("\n\n");
1843
- const { agentId } = params;
1218
+ const agentId = config.agentId ?? "main";
1844
1219
  const capsule = [
1845
1220
  "OpenClaw agent voice context:",
1846
1221
  `- Agent id: ${agentId}`,
@@ -1914,26 +1289,26 @@ function chunkAudio(audio, chunkSize = 160) {
1914
1289
  /** Default timeout for one telephony synthesis request. */
1915
1290
  const TELEPHONY_DEFAULT_TTS_TIMEOUT_MS = 8e3;
1916
1291
  /** Create a TTS provider that honors voice-call overrides and converts PCM to mulaw. */
1917
- async function createTelephonyTtsProvider(params) {
1292
+ function createTelephonyTtsProvider(params) {
1918
1293
  const { coreConfig, ttsOverride, runtime, logger } = params;
1919
- const preparedConfig = await runtime.prepareTtsRequest({
1920
- cfg: coreConfig,
1921
- override: ttsOverride,
1922
- text: ""
1923
- });
1294
+ const mergedConfig = applyTtsOverride(coreConfig, ttsOverride);
1295
+ const ttsConfig = mergedConfig.messages?.tts;
1296
+ const modelOverrides = resolveTelephonyModelOverridePolicy(readTelephonyModelOverrides(ttsConfig));
1297
+ const providerConfigs = collectTelephonyProviderConfigs(ttsConfig);
1298
+ const activeProvider = normalizeProviderId(ttsConfig?.provider);
1924
1299
  return {
1925
- synthesisTimeoutMs: resolveTimerTimeoutMs(preparedConfig.cfg.tts?.timeoutMs, TELEPHONY_DEFAULT_TTS_TIMEOUT_MS),
1300
+ synthesisTimeoutMs: resolveTimerTimeoutMs(mergedConfig.messages?.tts?.timeoutMs, TELEPHONY_DEFAULT_TTS_TIMEOUT_MS),
1926
1301
  synthesizeForTelephony: async (text) => {
1927
- const prepared = await runtime.prepareTtsRequest({
1928
- cfg: preparedConfig.cfg,
1929
- text
1302
+ const directives = parseTtsDirectives(text, modelOverrides, {
1303
+ cfg: mergedConfig,
1304
+ providerConfigs,
1305
+ preferredProviderId: activeProvider
1930
1306
  });
1931
- const directives = prepared.directives;
1932
1307
  if (directives.warnings.length > 0) logger?.warn?.(`[voice-call] Ignored telephony TTS directive overrides (${directives.warnings.join("; ")})`);
1933
1308
  const cleanText = directives.hasDirective ? directives.ttsText?.trim() || directives.cleanedText.trim() : text;
1934
1309
  const result = await runtime.textToSpeechTelephony({
1935
1310
  text: cleanText,
1936
- cfg: prepared.cfg,
1311
+ cfg: mergedConfig,
1937
1312
  overrides: directives.overrides
1938
1313
  });
1939
1314
  if (!result.success || !result.audioBuffer || !result.sampleRate) throw new Error(result.error ?? "TTS conversion failed");
@@ -1945,6 +1320,94 @@ async function createTelephonyTtsProvider(params) {
1945
1320
  }
1946
1321
  };
1947
1322
  }
1323
+ /** Apply voice-call TTS overrides to core config without mutating the original object. */
1324
+ function applyTtsOverride(coreConfig, override) {
1325
+ if (!override) return coreConfig;
1326
+ const base = coreConfig.messages?.tts;
1327
+ const merged = mergeTtsConfig(base, override);
1328
+ if (!merged) return coreConfig;
1329
+ return {
1330
+ ...coreConfig,
1331
+ messages: {
1332
+ ...coreConfig.messages,
1333
+ tts: merged
1334
+ }
1335
+ };
1336
+ }
1337
+ /** Merge core and voice-call TTS config, keeping undefined override fields out. */
1338
+ function mergeTtsConfig(base, override) {
1339
+ if (!base && !override) return;
1340
+ if (!override) return base;
1341
+ if (!base) return override;
1342
+ return deepMergeDefined(base, override);
1343
+ }
1344
+ /** Resolve directive override policy for telephony synthesis. */
1345
+ function resolveTelephonyModelOverridePolicy(overrides) {
1346
+ if (!(overrides?.enabled ?? true)) return {
1347
+ enabled: false,
1348
+ allowText: false,
1349
+ allowProvider: false,
1350
+ allowVoice: false,
1351
+ allowModelId: false,
1352
+ allowVoiceSettings: false,
1353
+ allowNormalization: false,
1354
+ allowSeed: false
1355
+ };
1356
+ const allow = (value, defaultValue = true) => value ?? defaultValue;
1357
+ return {
1358
+ enabled: true,
1359
+ allowText: allow(overrides?.allowText),
1360
+ allowProvider: allow(overrides?.allowProvider, false),
1361
+ allowVoice: allow(overrides?.allowVoice),
1362
+ allowModelId: allow(overrides?.allowModelId),
1363
+ allowVoiceSettings: allow(overrides?.allowVoiceSettings),
1364
+ allowNormalization: allow(overrides?.allowNormalization),
1365
+ allowSeed: allow(overrides?.allowSeed)
1366
+ };
1367
+ }
1368
+ /** Read model override policy from TTS config when present. */
1369
+ function readTelephonyModelOverrides(ttsConfig) {
1370
+ const value = ttsConfig?.modelOverrides;
1371
+ return value && typeof value === "object" && !Array.isArray(value) ? value : void 0;
1372
+ }
1373
+ /** Normalize provider ids for config lookup. */
1374
+ function normalizeProviderId(value) {
1375
+ return typeof value === "string" ? value.trim().toLowerCase() || void 0 : void 0;
1376
+ }
1377
+ /** Coerce provider config objects while rejecting arrays and primitives. */
1378
+ function asProviderConfig(value) {
1379
+ return value && typeof value === "object" && !Array.isArray(value) ? value : {};
1380
+ }
1381
+ /** Collect named provider configs from canonical and legacy TTS config shapes. */
1382
+ function collectTelephonyProviderConfigs(ttsConfig) {
1383
+ if (!ttsConfig) return {};
1384
+ const entries = {};
1385
+ const rawProviders = ttsConfig.providers && typeof ttsConfig.providers === "object" && !Array.isArray(ttsConfig.providers) ? ttsConfig.providers : {};
1386
+ for (const [providerId, value] of Object.entries(rawProviders)) {
1387
+ const normalized = normalizeProviderId(providerId) ?? providerId;
1388
+ entries[normalized] = asProviderConfig(value);
1389
+ }
1390
+ const reservedKeys = /* @__PURE__ */ new Set([
1391
+ "auto",
1392
+ "enabled",
1393
+ "maxTextLength",
1394
+ "mode",
1395
+ "modelOverrides",
1396
+ "persona",
1397
+ "personas",
1398
+ "prefsPath",
1399
+ "provider",
1400
+ "providers",
1401
+ "summaryModel",
1402
+ "timeoutMs"
1403
+ ]);
1404
+ for (const [key, value] of Object.entries(ttsConfig)) {
1405
+ if (reservedKeys.has(key) || typeof value !== "object" || value === null || Array.isArray(value)) continue;
1406
+ const normalized = normalizeProviderId(key) ?? key;
1407
+ entries[normalized] ??= asProviderConfig(value);
1408
+ }
1409
+ return entries;
1410
+ }
1948
1411
  //#endregion
1949
1412
  //#region extensions/voice-call/src/bounded-child-output.ts
1950
1413
  const DEFAULT_MAX_OUTPUT_CHARS = 16384;
@@ -1973,33 +1436,78 @@ function formatBoundedChildOutput(output) {
1973
1436
  }
1974
1437
  //#endregion
1975
1438
  //#region extensions/voice-call/src/webhook/tailscale.ts
1976
- const TAILSCALE_COMMAND_STDOUT_MAX_BYTES = 4 * 1024 * 1024;
1977
- async function runTailscaleCommand(args, timeoutMs = 2500) {
1978
- try {
1979
- const result = await runCommandWithTimeout(["tailscale", ...args], {
1980
- killProcessTree: true,
1981
- maxOutputBytes: {
1982
- stdout: TAILSCALE_COMMAND_STDOUT_MAX_BYTES,
1983
- stderr: 1
1984
- },
1985
- outputCapture: "head",
1986
- terminateOnOutputLimit: { stdout: true },
1987
- timeoutMs
1988
- });
1989
- if (result.termination !== "exit" || result.outputLimitExceeded) return {
1990
- code: -1,
1991
- stdout: ""
1992
- };
1993
- return {
1994
- code: result.code ?? -1,
1995
- stdout: result.stdout
1439
+ const TAILSCALE_COMMAND_STDOUT_MAX_BYTES = 4194304;
1440
+ function appendTailscaleCommandStdout(current, data, maxBytes = TAILSCALE_COMMAND_STDOUT_MAX_BYTES) {
1441
+ if (current.exceeded) return current;
1442
+ const buffer = Buffer.isBuffer(data) ? data : Buffer.from(data);
1443
+ const bytes = current.bytes + buffer.byteLength;
1444
+ if (bytes > maxBytes) return {
1445
+ bytes,
1446
+ exceeded: true,
1447
+ text: ""
1448
+ };
1449
+ return {
1450
+ bytes,
1451
+ exceeded: false,
1452
+ text: `${current.text}${buffer.toString("utf8")}`
1453
+ };
1454
+ }
1455
+ function runTailscaleCommand(args, timeoutMs = 2500) {
1456
+ return new Promise((resolve) => {
1457
+ const proc = spawn("tailscale", args, { stdio: [
1458
+ "ignore",
1459
+ "pipe",
1460
+ "ignore"
1461
+ ] });
1462
+ let stdout = {
1463
+ bytes: 0,
1464
+ exceeded: false,
1465
+ text: ""
1996
1466
  };
1997
- } catch {
1998
- return {
1999
- code: -1,
2000
- stdout: ""
1467
+ let settled = false;
1468
+ const finish = (result) => {
1469
+ if (settled) return;
1470
+ settled = true;
1471
+ clearTimeout(timer);
1472
+ resolve(result);
2001
1473
  };
2002
- }
1474
+ const timer = setTimeout(() => {
1475
+ proc.kill("SIGKILL");
1476
+ finish({
1477
+ code: -1,
1478
+ stdout: ""
1479
+ });
1480
+ }, timeoutMs);
1481
+ proc.stdout.on("data", (data) => {
1482
+ stdout = appendTailscaleCommandStdout(stdout, data);
1483
+ if (stdout.exceeded) {
1484
+ proc.kill("SIGKILL");
1485
+ finish({
1486
+ code: -1,
1487
+ stdout: ""
1488
+ });
1489
+ }
1490
+ });
1491
+ proc.stdout.on("error", () => {
1492
+ proc.kill("SIGKILL");
1493
+ finish({
1494
+ code: -1,
1495
+ stdout: ""
1496
+ });
1497
+ });
1498
+ proc.on("error", () => {
1499
+ finish({
1500
+ code: -1,
1501
+ stdout: ""
1502
+ });
1503
+ });
1504
+ proc.on("close", (code) => {
1505
+ finish({
1506
+ code: code ?? -1,
1507
+ stdout: stdout.text
1508
+ });
1509
+ });
1510
+ });
2003
1511
  }
2004
1512
  async function getTailscaleSelfInfo() {
2005
1513
  const { code, stdout } = await runTailscaleCommand([
@@ -2070,32 +1578,6 @@ async function cleanupTailscaleExposure(config) {
2070
1578
  //#endregion
2071
1579
  //#region extensions/voice-call/src/tunnel.ts
2072
1580
  const NGROK_LOG_BUFFER_MAX_CHARS = 16384;
2073
- const NGROK_ERROR_MARKER = "ERR_NGROK";
2074
- const TUNNEL_COMMAND_OUTPUT_MAX_BYTES = 16384;
2075
- const NGROK_STOP_GRACE_MS = 2e3;
2076
- const NGROK_FORCE_KILL_WAIT_MS = 1e3;
2077
- async function terminateNgrokProcess(proc, isClosed) {
2078
- if (isClosed()) return;
2079
- await new Promise((resolve) => {
2080
- let finished = false;
2081
- let forceKillWaitTimer;
2082
- const finish = () => {
2083
- if (finished) return;
2084
- finished = true;
2085
- if (forceKillTimer) clearTimeout(forceKillTimer);
2086
- if (forceKillWaitTimer) clearTimeout(forceKillWaitTimer);
2087
- proc.off("close", finish);
2088
- resolve();
2089
- };
2090
- proc.once("close", finish);
2091
- const forceKillTimer = setTimeout(() => {
2092
- forceKillWaitTimer = setTimeout(finish, NGROK_FORCE_KILL_WAIT_MS);
2093
- if (!isClosed()) proc.kill("SIGKILL");
2094
- }, NGROK_STOP_GRACE_MS);
2095
- proc.kill("SIGTERM");
2096
- if (isClosed()) finish();
2097
- });
2098
- }
2099
1581
  function listenForChildStreamErrors(proc, onError) {
2100
1582
  proc.stdout.on("error", (error) => onError("stdout", error));
2101
1583
  proc.stderr.on("error", (error) => onError("stderr", error));
@@ -2131,25 +1613,22 @@ async function startNgrokTunnel(config) {
2131
1613
  "pipe",
2132
1614
  "pipe"
2133
1615
  ] });
2134
- let startupSettled = false;
2135
- let childClosed = false;
1616
+ let resolved = false;
1617
+ let closed = false;
2136
1618
  let publicUrl = null;
2137
1619
  let outputBuffer = "";
2138
- let stderrTail = "";
2139
1620
  const timeout = setTimeout(() => {
2140
- if (!startupSettled) {
2141
- startupSettled = true;
2142
- terminateNgrokProcess(proc, () => childClosed).then(() => {
2143
- reject(/* @__PURE__ */ new Error("ngrok startup timed out (30s)"));
2144
- });
1621
+ if (!resolved) {
1622
+ resolved = true;
1623
+ proc.kill("SIGTERM");
1624
+ reject(/* @__PURE__ */ new Error("ngrok startup timed out (30s)"));
2145
1625
  }
2146
1626
  }, 3e4);
2147
- timeout.unref();
2148
1627
  const rejectIfPending = (message, kill = false) => {
2149
- if (!startupSettled) {
2150
- startupSettled = true;
1628
+ if (!resolved) {
1629
+ resolved = true;
2151
1630
  clearTimeout(timeout);
2152
- if (kill && !childClosed) proc.kill("SIGKILL");
1631
+ if (kill && !closed) proc.kill("SIGKILL");
2153
1632
  reject(new Error(message));
2154
1633
  }
2155
1634
  };
@@ -2158,8 +1637,8 @@ async function startNgrokTunnel(config) {
2158
1637
  const log = JSON.parse(line);
2159
1638
  if (log.msg === "started tunnel" && log.url) publicUrl = log.url;
2160
1639
  if (log.addr && log.url && !publicUrl) publicUrl = log.url;
2161
- if (publicUrl && !startupSettled) {
2162
- startupSettled = true;
1640
+ if (publicUrl && !resolved) {
1641
+ resolved = true;
2163
1642
  clearTimeout(timeout);
2164
1643
  const fullUrl = publicUrl + config.path;
2165
1644
  console.log(`[voice-call] ngrok tunnel active: ${fullUrl}`);
@@ -2167,24 +1646,39 @@ async function startNgrokTunnel(config) {
2167
1646
  publicUrl: fullUrl,
2168
1647
  provider: "ngrok",
2169
1648
  stop: async () => {
2170
- await terminateNgrokProcess(proc, () => childClosed);
1649
+ if (closed) return;
1650
+ await new Promise((res) => {
1651
+ let finished = false;
1652
+ const finish = () => {
1653
+ if (finished) return;
1654
+ finished = true;
1655
+ clearTimeout(fallback);
1656
+ proc.off("close", finish);
1657
+ res();
1658
+ };
1659
+ if (closed) {
1660
+ res();
1661
+ return;
1662
+ }
1663
+ proc.once("close", finish);
1664
+ const fallback = setTimeout(finish, 2e3);
1665
+ proc.kill("SIGTERM");
1666
+ if (closed) finish();
1667
+ });
2171
1668
  }
2172
1669
  });
2173
1670
  }
2174
1671
  } catch {}
2175
1672
  };
2176
- proc.stdout.setEncoding("utf8");
2177
- proc.stderr.setEncoding("utf8");
2178
- proc.stdout.on("data", (chunk) => {
2179
- const lines = (outputBuffer + chunk).split("\n");
1673
+ proc.stdout.on("data", (data) => {
1674
+ const lines = (outputBuffer + data.toString()).split("\n");
2180
1675
  outputBuffer = lines.pop() || "";
2181
1676
  if (outputBuffer.length > NGROK_LOG_BUFFER_MAX_CHARS) outputBuffer = sliceUtf16Safe(outputBuffer, -16384);
2182
1677
  for (const line of lines) if (line.trim()) processLine(line);
2183
1678
  });
2184
- proc.stderr.on("data", (chunk) => {
2185
- const combined = stderrTail + chunk;
2186
- if (combined.includes(NGROK_ERROR_MARKER)) rejectIfPending(`ngrok error: ${formatBoundedChildOutput(appendBoundedChildOutput(emptyBoundedChildOutput(), combined))}`, true);
2187
- stderrTail = sliceUtf16Safe(combined, -8);
1679
+ proc.stderr.on("data", (data) => {
1680
+ const msg = data.toString();
1681
+ if (msg.includes("ERR_NGROK")) rejectIfPending(`ngrok error: ${formatBoundedChildOutput(appendBoundedChildOutput(emptyBoundedChildOutput(), msg))}`, true);
2188
1682
  });
2189
1683
  listenForChildStreamErrors(proc, (stream, error) => {
2190
1684
  rejectIfPending(`ngrok ${stream} error: ${error.message}`, true);
@@ -2193,9 +1687,9 @@ async function startNgrokTunnel(config) {
2193
1687
  rejectIfPending(`Failed to start ngrok: ${err.message}`);
2194
1688
  });
2195
1689
  proc.on("close", (code) => {
2196
- childClosed = true;
2197
- if (!startupSettled) {
2198
- startupSettled = true;
1690
+ closed = true;
1691
+ if (!resolved) {
1692
+ resolved = true;
2199
1693
  clearTimeout(timeout);
2200
1694
  reject(/* @__PURE__ */ new Error(`ngrok exited unexpectedly with code ${code}`));
2201
1695
  }
@@ -2206,22 +1700,41 @@ async function startNgrokTunnel(config) {
2206
1700
  * Run an ngrok command and wait for completion.
2207
1701
  */
2208
1702
  async function runNgrokCommand(args) {
2209
- const result = await runCommandWithTimeout(["ngrok", ...args], {
2210
- killProcessTree: true,
2211
- maxOutputBytes: TUNNEL_COMMAND_OUTPUT_MAX_BYTES,
2212
- outputCapture: "tail",
2213
- timeoutMs: 3e4
1703
+ return new Promise((resolve, reject) => {
1704
+ const proc = spawn("ngrok", args, { stdio: [
1705
+ "ignore",
1706
+ "pipe",
1707
+ "pipe"
1708
+ ] });
1709
+ let stdout = emptyBoundedChildOutput();
1710
+ let stderr = emptyBoundedChildOutput();
1711
+ let settled = false;
1712
+ const rejectIfPending = (error, kill = false) => {
1713
+ if (settled) return;
1714
+ settled = true;
1715
+ if (kill) proc.kill("SIGKILL");
1716
+ reject(error);
1717
+ };
1718
+ proc.stdout.on("data", (data) => {
1719
+ stdout = appendBoundedChildOutput(stdout, data.toString());
1720
+ });
1721
+ proc.stderr.on("data", (data) => {
1722
+ stderr = appendBoundedChildOutput(stderr, data.toString());
1723
+ });
1724
+ listenForChildStreamErrors(proc, (stream, error) => {
1725
+ rejectIfPending(/* @__PURE__ */ new Error(`ngrok command ${stream} error: ${error.message}`), true);
1726
+ });
1727
+ proc.on("close", (code) => {
1728
+ if (settled) return;
1729
+ settled = true;
1730
+ if (code === 0) resolve(stdout.text);
1731
+ else {
1732
+ const output = stderr.text ? stderr : stdout;
1733
+ reject(/* @__PURE__ */ new Error(`ngrok command failed: ${formatBoundedChildOutput(output)}`));
1734
+ }
1735
+ });
1736
+ proc.on("error", (error) => rejectIfPending(error));
2214
1737
  });
2215
- if (result.termination === "timeout") throw new Error("ngrok command timed out");
2216
- if (result.code === 0) return result.stdout;
2217
- const output = result.stderr ? {
2218
- text: result.stderr,
2219
- truncated: Boolean(result.stderrTruncatedBytes)
2220
- } : {
2221
- text: result.stdout,
2222
- truncated: Boolean(result.stdoutTruncatedBytes)
2223
- };
2224
- throw new Error(`ngrok command failed: ${formatBoundedChildOutput(output)}`);
2225
1738
  }
2226
1739
  /**
2227
1740
  * Start a Tailscale serve/funnel tunnel.
@@ -2231,56 +1744,89 @@ async function startTailscaleTunnel(config) {
2231
1744
  if (!dnsName) throw new Error("Could not get Tailscale DNS name. Is Tailscale running?");
2232
1745
  const path = config.path.startsWith("/") ? config.path : `/${config.path}`;
2233
1746
  const localUrl = `http://127.0.0.1:${config.port}${path}`;
2234
- const result = await runCommandWithTimeout([
2235
- "tailscale",
2236
- config.mode,
2237
- "--bg",
2238
- "--yes",
2239
- "--set-path",
2240
- path,
2241
- localUrl
2242
- ], {
2243
- killProcessTree: true,
2244
- maxOutputBytes: TUNNEL_COMMAND_OUTPUT_MAX_BYTES,
2245
- outputCapture: "tail",
2246
- timeoutMs: 1e4
2247
- });
2248
- if (result.termination === "timeout") throw new Error(`Tailscale ${config.mode} timed out`);
2249
- if (result.code !== 0) {
2250
- const output = result.stderr ? {
2251
- text: result.stderr,
2252
- truncated: Boolean(result.stderrTruncatedBytes)
2253
- } : {
2254
- text: result.stdout,
2255
- truncated: Boolean(result.stdoutTruncatedBytes)
1747
+ return new Promise((resolve, reject) => {
1748
+ const proc = spawn("tailscale", [
1749
+ config.mode,
1750
+ "--bg",
1751
+ "--yes",
1752
+ "--set-path",
1753
+ path,
1754
+ localUrl
1755
+ ], { stdio: [
1756
+ "ignore",
1757
+ "pipe",
1758
+ "pipe"
1759
+ ] });
1760
+ let resolved = false;
1761
+ let stdout = emptyBoundedChildOutput();
1762
+ let stderr = emptyBoundedChildOutput();
1763
+ const rejectIfPending = (error, kill = false) => {
1764
+ if (resolved) return;
1765
+ resolved = true;
1766
+ clearTimeout(timeout);
1767
+ if (kill) proc.kill("SIGKILL");
1768
+ reject(error);
2256
1769
  };
2257
- const detail = output.text ? `: ${formatBoundedChildOutput(output)}` : "";
2258
- throw new Error(`Tailscale ${config.mode} failed with code ${result.code}${detail}`);
2259
- }
2260
- const publicUrl = `https://${dnsName}${path}`;
2261
- console.log(`[voice-call] Tailscale ${config.mode} active: ${publicUrl}`);
2262
- return {
2263
- publicUrl,
2264
- provider: `tailscale-${config.mode}`,
2265
- stop: async () => {
2266
- await stopTailscaleTunnel(config.mode, path);
2267
- }
2268
- };
1770
+ const timeout = setTimeout(() => {
1771
+ rejectIfPending(/* @__PURE__ */ new Error(`Tailscale ${config.mode} timed out`), true);
1772
+ }, 1e4);
1773
+ proc.stdout.on("data", (data) => {
1774
+ stdout = appendBoundedChildOutput(stdout, data.toString());
1775
+ });
1776
+ proc.stderr.on("data", (data) => {
1777
+ stderr = appendBoundedChildOutput(stderr, data.toString());
1778
+ });
1779
+ listenForChildStreamErrors(proc, (stream, error) => {
1780
+ rejectIfPending(/* @__PURE__ */ new Error(`Tailscale ${config.mode} ${stream} error: ${error.message}`), true);
1781
+ });
1782
+ proc.on("close", (code) => {
1783
+ clearTimeout(timeout);
1784
+ if (resolved) return;
1785
+ resolved = true;
1786
+ if (code === 0) {
1787
+ const publicUrl = `https://${dnsName}${path}`;
1788
+ console.log(`[voice-call] Tailscale ${config.mode} active: ${publicUrl}`);
1789
+ resolve({
1790
+ publicUrl,
1791
+ provider: `tailscale-${config.mode}`,
1792
+ stop: async () => {
1793
+ await stopTailscaleTunnel(config.mode, path);
1794
+ }
1795
+ });
1796
+ } else {
1797
+ const output = stderr.text ? stderr : stdout;
1798
+ const detail = output.text ? `: ${formatBoundedChildOutput(output)}` : "";
1799
+ reject(/* @__PURE__ */ new Error(`Tailscale ${config.mode} failed with code ${code}${detail}`));
1800
+ }
1801
+ });
1802
+ proc.on("error", (err) => {
1803
+ rejectIfPending(err);
1804
+ });
1805
+ });
2269
1806
  }
2270
1807
  /**
2271
1808
  * Stop a Tailscale serve/funnel tunnel.
2272
1809
  */
2273
1810
  async function stopTailscaleTunnel(mode, path) {
2274
- await runCommandWithTimeout([
2275
- "tailscale",
2276
- mode,
2277
- "off",
2278
- path
2279
- ], {
2280
- killProcessTree: true,
2281
- maxOutputBytes: 1,
2282
- timeoutMs: 5e3
2283
- }).catch(() => {});
1811
+ return new Promise((resolve) => {
1812
+ const proc = spawn("tailscale", [
1813
+ mode,
1814
+ "off",
1815
+ path
1816
+ ], { stdio: "ignore" });
1817
+ const timeout = setTimeout(() => {
1818
+ proc.kill("SIGKILL");
1819
+ resolve();
1820
+ }, 5e3);
1821
+ proc.on("close", () => {
1822
+ clearTimeout(timeout);
1823
+ resolve();
1824
+ });
1825
+ proc.on("error", () => {
1826
+ clearTimeout(timeout);
1827
+ resolve();
1828
+ });
1829
+ });
2284
1830
  }
2285
1831
  /**
2286
1832
  * Start a tunnel based on configuration.
@@ -2369,18 +1915,13 @@ function getHeader(headers, name) {
2369
1915
  return value;
2370
1916
  }
2371
1917
  //#endregion
2372
- //#region extensions/voice-call/src/media-base64.ts
2373
- function canonicalizeVoiceCallMediaBase64(payloadBase64) {
2374
- return canonicalizeBase64(payloadBase64.replaceAll("-", "+").replaceAll("_", "/"));
2375
- }
2376
- //#endregion
2377
1918
  //#region extensions/voice-call/src/media-stream.ts
2378
1919
  const DEFAULT_PRE_START_TIMEOUT_MS = 5e3;
2379
1920
  const DEFAULT_MAX_PENDING_CONNECTIONS = 32;
2380
1921
  const DEFAULT_MAX_PENDING_CONNECTIONS_PER_IP = 4;
2381
1922
  const DEFAULT_MAX_CONNECTIONS = 128;
2382
- const MAX_INBOUND_MESSAGE_BYTES = 64 * 1024;
2383
- const MAX_WS_BUFFERED_BYTES = 1024 * 1024;
1923
+ const MAX_INBOUND_MESSAGE_BYTES = 65536;
1924
+ const MAX_WS_BUFFERED_BYTES = 1048576;
2384
1925
  const CLOSE_REASON_LOG_MAX_CHARS = 120;
2385
1926
  function sanitizeLogText(value, maxChars) {
2386
1927
  const sanitized = value.replace(/\p{Cc}/gu, " ").replace(/\s+/g, " ").trim();
@@ -2488,9 +2029,7 @@ var MediaStreamHandler = class {
2488
2029
  break;
2489
2030
  case "media":
2490
2031
  if (session && message.media?.payload) {
2491
- const canonicalPayload = canonicalizeVoiceCallMediaBase64(message.media.payload);
2492
- if (!canonicalPayload) break;
2493
- const audioBuffer = Buffer.from(canonicalPayload, "base64");
2032
+ const audioBuffer = Buffer.from(message.media.payload, "base64");
2494
2033
  const turnId = this.ensureActiveTurn(session);
2495
2034
  this.emitTalkEvent(session, {
2496
2035
  type: "input.audio.delta",
@@ -2504,14 +2043,10 @@ var MediaStreamHandler = class {
2504
2043
  session.sttSession.sendAudio(audioBuffer);
2505
2044
  }
2506
2045
  break;
2507
- case "stop":
2508
- if (session) {
2509
- this.handleStop(session);
2510
- session = null;
2511
- }
2512
- break;
2513
- case "clear":
2514
- case "mark": break;
2046
+ case "stop": if (session) {
2047
+ this.handleStop(session);
2048
+ session = null;
2049
+ }
2515
2050
  }
2516
2051
  } catch (error) {
2517
2052
  console.error("[MediaStream] Error processing message:", error);
@@ -2771,7 +2306,9 @@ Content-Length: ${Buffer.byteLength(body)}\r\n\r
2771
2306
  bufferedAfterBytes: session.ws.bufferedAmount
2772
2307
  };
2773
2308
  if (bufferedBeforeBytes > MAX_WS_BUFFERED_BYTES) {
2774
- session.ws.close(1013, "Backpressure: send buffer exceeded");
2309
+ try {
2310
+ session.ws.close(1013, "Backpressure: send buffer exceeded");
2311
+ } catch {}
2775
2312
  return {
2776
2313
  sent: false,
2777
2314
  readyState,
@@ -2783,7 +2320,9 @@ Content-Length: ${Buffer.byteLength(body)}\r\n\r
2783
2320
  session.ws.send(JSON.stringify(message));
2784
2321
  const bufferedAfterBytes = session.ws.bufferedAmount;
2785
2322
  if (bufferedAfterBytes > MAX_WS_BUFFERED_BYTES) {
2786
- session.ws.close(1013, "Backpressure: send buffer exceeded");
2323
+ try {
2324
+ session.ws.close(1013, "Backpressure: send buffer exceeded");
2325
+ } catch {}
2787
2326
  return {
2788
2327
  sent: false,
2789
2328
  readyState,
@@ -3043,19 +2582,15 @@ function startStaleCallReaper(params) {
3043
2582
  const maxAgeSeconds = params.staleCallReaperSeconds;
3044
2583
  if (!maxAgeSeconds || maxAgeSeconds <= 0) return null;
3045
2584
  const maxAgeMs = maxAgeSeconds * 1e3;
3046
- const callsBeingReaped = /* @__PURE__ */ new Set();
3047
2585
  const interval = setInterval(() => {
3048
2586
  const now = Date.now();
3049
2587
  for (const call of params.manager.getActiveCalls()) {
3050
2588
  if (call.answeredAt || TerminalStates.has(call.state) || LiveConversationStates.has(call.state)) continue;
3051
2589
  const age = now - call.startedAt;
3052
- if (age > maxAgeMs && !callsBeingReaped.has(call.callId)) {
3053
- callsBeingReaped.add(call.callId);
2590
+ if (age > maxAgeMs) {
3054
2591
  console.log(`[voice-call] Reaping stale call ${call.callId} (age: ${Math.round(age / 1e3)}s, state: ${call.state})`);
3055
2592
  params.manager.endCall(call.callId).catch((err) => {
3056
2593
  console.warn(`[voice-call] Reaper failed to end call ${call.callId}:`, err);
3057
- }).finally(() => {
3058
- callsBeingReaped.delete(call.callId);
3059
2594
  });
3060
2595
  }
3061
2596
  }
@@ -3070,8 +2605,8 @@ const MAX_WEBHOOK_BODY_BYTES = WEBHOOK_BODY_READ_DEFAULTS.preAuth.maxBytes;
3070
2605
  const WEBHOOK_BODY_TIMEOUT_MS = WEBHOOK_BODY_READ_DEFAULTS.preAuth.timeoutMs;
3071
2606
  const MISSING_REMOTE_ADDRESS_IN_FLIGHT_KEY = "__voice_call_no_remote__";
3072
2607
  const STREAM_DISCONNECT_HANGUP_GRACE_MS = 2e3;
3073
- const loadRealtimeTranscriptionRuntime = createLazyRuntimeModule(() => import("./realtime-transcription.runtime-CbJAs5t_.js"));
3074
- const loadResponseGeneratorModule = createLazyRuntimeModule(() => import("./response-generator-z6M9229t.js"));
2608
+ const loadRealtimeTranscriptionRuntime = createLazyRuntimeModule(() => import("./realtime-transcription.runtime-CbJAs5t_.mjs"));
2609
+ const loadResponseGeneratorModule = createLazyRuntimeModule(() => import("./response-generator-DoC6lfcm.mjs"));
3075
2610
  function appendRecentTalkEventMetadata(call, event) {
3076
2611
  const metadata = call.metadata ?? {};
3077
2612
  const recent = Array.isArray(metadata.recentTalkEvents) ? metadata.recentTalkEvents.filter((entry) => Boolean(entry) && typeof entry === "object" && !Array.isArray(entry)) : [];
@@ -3128,7 +2663,7 @@ function buildTwilioReplayTwiML() {
3128
2663
  body: "<?xml version=\"1.0\" encoding=\"UTF-8\"?><Response></Response>"
3129
2664
  };
3130
2665
  }
3131
- const WEBHOOK_REPLAY_RESPONSE_TTL_MS = 600 * 1e3;
2666
+ const WEBHOOK_REPLAY_RESPONSE_TTL_MS = 6e5;
3132
2667
  const WEBHOOK_REPLAY_RESPONSE_MAX_ENTRIES = 1e4;
3133
2668
  const WEBHOOK_REPLAY_RESPONSE_PRUNE_INTERVAL = 64;
3134
2669
  function cloneWebhookResponsePayload(payload) {
@@ -3263,14 +2798,12 @@ var VoiceCallWebhookServer = class {
3263
2798
  maxConnections: streaming.maxConnections,
3264
2799
  resolveClientIp: (request) => this.resolveMediaStreamClientIp(request),
3265
2800
  shouldAcceptStream: ({ callId, token }) => {
3266
- if (this.provider.name !== "twilio") {
3267
- this.logger.warn(`Rejecting media stream: provider ${this.provider.name} does not support authenticated classic streaming`);
3268
- return false;
3269
- }
3270
2801
  if (!this.manager.getCallByProviderCallId(callId)) return false;
3271
- if (!this.provider.isValidStreamToken(callId, token)) {
3272
- this.logger.warn(`Rejecting media stream: invalid token for ${callId}`);
3273
- return false;
2802
+ if (this.provider.name === "twilio") {
2803
+ if (!this.provider.isValidStreamToken(callId, token)) {
2804
+ this.logger.warn(`Rejecting media stream: invalid token for ${callId}`);
2805
+ return false;
2806
+ }
3274
2807
  }
3275
2808
  return true;
3276
2809
  },
@@ -3715,6 +3248,7 @@ var VoiceCallWebhookServer = class {
3715
3248
  callId,
3716
3249
  sessionKey: call.sessionKey,
3717
3250
  from: call.from,
3251
+ senderIsOwner: call.direction === "inbound" ? false : void 0,
3718
3252
  agentId: resolveCallAgentId(call, effectiveConfig),
3719
3253
  transcript: call.transcript,
3720
3254
  userMessage,
@@ -3746,12 +3280,12 @@ const REALTIME_VOICE_CONSULT_SYSTEM_PROMPT = [
3746
3280
  "Do not print secret values or dump environment variables; only check whether required configuration is present.",
3747
3281
  "Be accurate, brief, and speakable."
3748
3282
  ].join(" ");
3749
- const loadTelnyxProvider = createLazyRuntimeModule(() => import("./telnyx-BG56lL7z.js"));
3750
- const loadTwilioProvider = createLazyRuntimeModule(() => import("./twilio-B3FQ_P9x.js"));
3751
- const loadPlivoProvider = createLazyRuntimeModule(() => import("./plivo-B-KvPVji.js"));
3752
- const loadMockProvider = createLazyRuntimeModule(() => import("./mock-YsvBTX-7.js"));
3753
- const loadRealtimeVoiceRuntime = createLazyRuntimeModule(() => import("./realtime-voice.runtime-vtdCOWg-.js"));
3754
- const loadRealtimeHandler = createLazyRuntimeModule(() => import("./realtime-handler-hIVPEZxN.js"));
3283
+ const loadTelnyxProvider = createLazyRuntimeModule(() => import("./telnyx-Bb8X-nin.mjs"));
3284
+ const loadTwilioProvider = createLazyRuntimeModule(() => import("./twilio-Bsbf3O9A.mjs"));
3285
+ const loadPlivoProvider = createLazyRuntimeModule(() => import("./plivo-DFHh72MH.mjs"));
3286
+ const loadMockProvider = createLazyRuntimeModule(() => import("./mock-DY4yTqR2.mjs"));
3287
+ const loadRealtimeVoiceRuntime = createLazyRuntimeModule(() => import("./realtime-voice.runtime-vtdCOWg-.mjs"));
3288
+ const loadRealtimeHandler = createLazyRuntimeModule(() => import("./realtime-handler-DWLxvZuO.mjs"));
3755
3289
  function resolveVoiceCallConsultSessionKey(call) {
3756
3290
  return resolveVoiceCallSessionKey({
3757
3291
  config: call.config,
@@ -3876,8 +3410,7 @@ async function createRealtimeInstructionsResolver(params) {
3876
3410
  baseInstructions: params.config.realtime.instructions,
3877
3411
  config: genericConfig,
3878
3412
  coreConfig: params.coreConfig,
3879
- agentRuntime: params.agentRuntime,
3880
- agentId: params.config.agentId
3413
+ agentRuntime: params.agentRuntime
3881
3414
  });
3882
3415
  const entries = await Promise.all(listRealtimeAgentIds(params.config, params.coreConfig).map(async (agentId) => {
3883
3416
  return [agentId, await buildRealtimeVoiceInstructions({
@@ -3887,8 +3420,7 @@ async function createRealtimeInstructionsResolver(params) {
3887
3420
  agentId
3888
3421
  },
3889
3422
  coreConfig: params.coreConfig,
3890
- agentRuntime: params.agentRuntime,
3891
- agentId
3423
+ agentRuntime: params.agentRuntime
3892
3424
  })];
3893
3425
  }));
3894
3426
  const instructionsByAgentId = new Map(entries);
@@ -3906,13 +3438,8 @@ async function createVoiceCallRuntime(params) {
3906
3438
  error: console.error,
3907
3439
  debug: console.debug
3908
3440
  };
3441
+ const config = resolveVoiceCallConfig(rawConfig);
3909
3442
  const cfg = fullConfig ?? coreConfig;
3910
- const unresolvedConfig = resolveVoiceCallConfig(rawConfig);
3911
- const configuredAgentId = unresolvedConfig.agentId ? normalizeAgentId(unresolvedConfig.agentId) : resolveDefaultAgentId(cfg);
3912
- const config = {
3913
- ...unresolvedConfig,
3914
- agentId: configuredAgentId
3915
- };
3916
3443
  if (!config.enabled) throw new Error("Voice call disabled. Enable the plugin entry in config.");
3917
3444
  if (config.skipSignatureVerification) log.warn("[voice-call] SECURITY WARNING: skipSignatureVerification=true disables webhook signature verification (development only). Do not use in production.");
3918
3445
  const validation = validateProviderConfig(config);
@@ -3951,14 +3478,6 @@ async function createVoiceCallRuntime(params) {
3951
3478
  coreSession: cfg.session
3952
3479
  });
3953
3480
  const requesterSessionKey = typeof call.metadata?.requesterSessionKey === "string" ? call.metadata.requesterSessionKey : void 0;
3954
- const modelLockParams = {
3955
- cfg,
3956
- agentRuntime,
3957
- agentId,
3958
- sessionKey,
3959
- spawnedBy: requesterSessionKey
3960
- };
3961
- assertRealtimeVoiceAgentConsultModelSelectionUnlocked(modelLockParams);
3962
3481
  const fastContext = await resolveRealtimeFastContextConsult({
3963
3482
  cfg,
3964
3483
  agentId,
@@ -3967,10 +3486,7 @@ async function createVoiceCallRuntime(params) {
3967
3486
  args,
3968
3487
  logger: log
3969
3488
  });
3970
- if (fastContext.handled) {
3971
- assertRealtimeVoiceAgentConsultModelSelectionUnlocked(modelLockParams);
3972
- return fastContext.result;
3973
- }
3489
+ if (fastContext.handled) return fastContext.result;
3974
3490
  const { provider: agentProvider, model } = resolveVoiceResponseModel({
3975
3491
  voiceConfig: effectiveConfig,
3976
3492
  agentRuntime
@@ -4038,8 +3554,8 @@ async function createVoiceCallRuntime(params) {
4038
3554
  if (provider.name === "twilio" && config.streaming?.enabled) {
4039
3555
  const twilioProvider = provider;
4040
3556
  if (ttsRuntime?.textToSpeechTelephony) try {
4041
- const ttsProvider = await createTelephonyTtsProvider({
4042
- coreConfig: cfg,
3557
+ const ttsProvider = createTelephonyTtsProvider({
3558
+ coreConfig,
4043
3559
  ttsOverride: config.tts,
4044
3560
  runtime: ttsRuntime,
4045
3561
  logger: log
@@ -4077,4 +3593,4 @@ async function createVoiceCallRuntime(params) {
4077
3593
  }
4078
3594
  }
4079
3595
  //#endregion
4080
- export { validateProviderConfig as S, escapeXml as _, normalizeProviderStatus as a, resolveVoiceCallConfig as b, resolveWebhookExposureStatus as c, setupTailscaleExposureRoute as d, TELEPHONY_DEFAULT_TTS_TIMEOUT_MS as f, resolveUserPath as g, resolveCallAgentId as h, mapProviderStatusToEndReason as i, cleanupTailscaleExposureRoute as l, resolveVoiceResponseModel as m, normalizeProxyIp as n, canonicalizeVoiceCallMediaBase64 as o, chunkAudio as p, isProviderStatusTerminal as r, getHeader as s, createVoiceCallRuntime as t, getTailscaleSelfInfo as u, mapVoiceToPolly as v, resolveVoiceCallSessionKey as x, VoiceCallConfigSchema as y };
3596
+ export { mapVoiceToPolly as _, normalizeProviderStatus as a, cleanupTailscaleExposureRoute as c, TELEPHONY_DEFAULT_TTS_TIMEOUT_MS as d, chunkAudio as f, escapeXml as g, resolveUserPath as h, mapProviderStatusToEndReason as i, getTailscaleSelfInfo as l, resolveCallAgentId as m, normalizeProxyIp as n, getHeader as o, resolveVoiceResponseModel as p, isProviderStatusTerminal as r, resolveWebhookExposureStatus as s, createVoiceCallRuntime as t, setupTailscaleExposureRoute as u };