@lawpath-tech/openclaw 2026.2.21-19 → 2026.2.21-21

package/dist/{web-DPHo2cWX.js → web-CVSol55V.js}

@@ -8,13 +8,13 @@ import { c as resolveIdentityNamePrefix, l as resolveMessagePrefix, t as createR
 import "./github-copilot-token-ttqQRqMA.js";
 import { r as formatCliCommand } from "./env-DulyEJtM.js";
 import "./tokens-DnMmat9I.js";
-import { C as shouldComputeCommandAuthorized, S as hasControlCommand, T as normalizeMentionText, _ as enqueueSystemEvent, a as resolveMentionGating, b as formatInboundEnvelope, c as recordPendingHistoryEntryIfEnabled, f as createDedupeCache, g as formatDurationPrecise, h as parseActivationCommand, i as buildPairingReply, l as shouldAckReactionForWhatsApp, m as normalizeGroupActivation, n as computeBackoff, o as DEFAULT_GROUP_HISTORY_LIMIT, p as getReplyFromConfig, r as sleepWithAbort, s as buildHistoryContextFromEntries, u as dispatchReplyWithBufferedBlockDispatcher, v as createInboundDebouncer, w as buildMentionRegexes, x as resolveEnvelopeFormatOptions, y as resolveInboundDebounceMs } from "./pi-embedded-CwQsXrVT.js";
+import { C as shouldComputeCommandAuthorized, S as hasControlCommand, T as normalizeMentionText, _ as enqueueSystemEvent, a as resolveMentionGating, b as formatInboundEnvelope, c as recordPendingHistoryEntryIfEnabled, f as createDedupeCache, g as formatDurationPrecise, h as parseActivationCommand, i as buildPairingReply, l as shouldAckReactionForWhatsApp, m as normalizeGroupActivation, n as computeBackoff, o as DEFAULT_GROUP_HISTORY_LIMIT, p as getReplyFromConfig, r as sleepWithAbort, s as buildHistoryContextFromEntries, u as dispatchReplyWithBufferedBlockDispatcher, v as createInboundDebouncer, w as buildMentionRegexes, x as resolveEnvelopeFormatOptions, y as resolveInboundDebounceMs } from "./pi-embedded-ZvazjIyF.js";
 import "./plugins-B90TBXkV.js";
 import { i as logWebSelfId, l as readWebSelfId, n as WA_WEB_AUTH_DIR, p as webAuthExists, r as getWebAuthAgeMs, s as pickWebChannel, t as resolveWhatsAppAccount } from "./accounts-Cxza2UXO.js";
 import "./bindings-q3EHMdG1.js";
 import "./send-Chks8Sx_.js";
 import "./send-D4cCHXzJ.js";
-import "./deliver-B2d2N8OJ.js";
+import "./deliver-0ThKlzQo.js";
 import "./diagnostic-DzmNBZt3.js";
 import "./diagnostic-session-state-C1vRJs5w.js";
 import "./accounts-sr0G5UOe.js";
@@ -22,10 +22,10 @@ import { G as formatLocationText, J as readChannelAllowFromStore, K as toLocatio
 import "./image-ops-BfqdVNxh.js";
 import "./pi-model-discovery-j5tVLINv.js";
 import "./message-channel-DNOtJsL1.js";
-import { Ct as resolveGroupSessionKey, H as loadSessionStore, K as updateLastRoute, U as readSessionUpdatedAt, W as recordSessionMetaFromInbound } from "./pi-embedded-helpers-Cd0S0WfR.js";
+import { Ct as resolveGroupSessionKey, H as loadSessionStore, K as updateLastRoute, U as readSessionUpdatedAt, W as recordSessionMetaFromInbound } from "./pi-embedded-helpers-B0Kht0I2.js";
 import { n as loadConfig } from "./config-KL7847B1.js";
 import "./manifest-registry-sze5OW-s.js";
-import "./chrome-lRUqqgSD.js";
+import "./chrome-CZuniMYN.js";
 import "./skills-DGNBryo1.js";
 import "./redact-1NGYV_8p.js";
 import "./errors-CPfngF0S.js";
@@ -35,7 +35,7 @@ import { h as resolveChannelGroupRequireMention, m as resolveChannelGroupPolicy
 import "./accounts-DAZL5b5o.js";
 import { s as resolveStorePath } from "./paths-5iQF9bSz.js";
 import "./tool-images-TQiVqsOc.js";
-import "./image--DDZnw-F.js";
+import "./image-88q3KE-C.js";
 import "./manager-CsXJuzK4.js";
 import "./gemini-auth-BF6ZCNOV.js";
 import "./sqlite-CmTvWxxq.js";
@@ -49,7 +49,7 @@ import "./render-B1VqYyvo.js";
 import "./commands-registry-DEXQtgBS.js";
 import "./skill-commands-CGvIAIiQ.js";
 import { t as finalizeInboundContext } from "./inbound-context-QOi5vzUt.js";
-import { p as registerUnhandledRejectionHandler } from "./runner-Dq-qfrq7.js";
+import { p as registerUnhandledRejectionHandler } from "./runner-a43IsYad.js";
 import "./fetch-BMa0enEg.js";
 import { n as recordChannelActivity } from "./channel-activity-7fb2OxTV.js";
 import { t as convertMarkdownTables } from "./tables-Ds9lRo34.js";

package/dist/{web-Cs0IP-ZO.js → web-Y49Dumye.js}

@@ -5,7 +5,7 @@ import "./agent-scope-DSUeV7-g.js";
 import { r as defaultRuntime, t as createSubsystemLogger } from "./subsystem-BF9jT_Nw.js";
 import "./workspace-DYx5Dxtw.js";
 import "./tokens-B-c56SkM.js";
-import { C as resolveEnvelopeFormatOptions, D as normalizeMentionText, E as buildMentionRegexes, S as formatInboundEnvelope, T as shouldComputeCommandAuthorized, _ as parseActivationCommand, b as createInboundDebouncer, c as DEFAULT_GROUP_HISTORY_LIMIT, d as shouldAckReactionForWhatsApp, f as dispatchReplyWithBufferedBlockDispatcher, g as normalizeGroupActivation, h as getReplyFromConfig, l as buildHistoryContextFromEntries, m as createDedupeCache, n as computeBackoff, o as buildPairingReply, r as sleepWithAbort, s as resolveMentionGating, u as recordPendingHistoryEntryIfEnabled, v as formatDurationPrecise, w as hasControlCommand, x as resolveInboundDebounceMs, y as enqueueSystemEvent } from "./pi-embedded-D7qD6fo-.js";
+import { C as resolveEnvelopeFormatOptions, D as normalizeMentionText, E as buildMentionRegexes, S as formatInboundEnvelope, T as shouldComputeCommandAuthorized, _ as parseActivationCommand, b as createInboundDebouncer, c as DEFAULT_GROUP_HISTORY_LIMIT, d as shouldAckReactionForWhatsApp, f as dispatchReplyWithBufferedBlockDispatcher, g as normalizeGroupActivation, h as getReplyFromConfig, l as buildHistoryContextFromEntries, m as createDedupeCache, n as computeBackoff, o as buildPairingReply, r as sleepWithAbort, s as resolveMentionGating, u as recordPendingHistoryEntryIfEnabled, v as formatDurationPrecise, w as hasControlCommand, x as resolveInboundDebounceMs, y as enqueueSystemEvent } from "./pi-embedded-Dz24QZz9.js";
 import "./plugins-DIrNa311.js";
 import { i as logWebSelfId, l as readWebSelfId, n as WA_WEB_AUTH_DIR, p as webAuthExists, r as getWebAuthAgeMs, s as pickWebChannel, t as resolveWhatsAppAccount } from "./accounts-C8m4ADkw.js";
 import "./boolean-B8-BqKGQ.js";
@@ -13,7 +13,7 @@ import { t as formatCliCommand } from "./command-format-BOHIfoCX.js";
 import "./bindings-BKEEBdsb.js";
 import "./send-Byt_wJDK.js";
 import "./send-BEFzSOoD.js";
-import "./deliver-CLJRPnfx.js";
+import "./deliver-CKH_FhS0.js";
 import "./diagnostic-DriDPU6M.js";
 import "./diagnostic-session-state-Bxo4UHOL.js";
 import "./accounts-CQT5u65w.js";
@@ -23,10 +23,10 @@ import "./model-auth-vtNH8C4P.js";
 import "./github-copilot-token-Dgb9dAHW.js";
 import "./pi-model-discovery-DaNAekda.js";
 import "./message-channel-CHYSjmbr.js";
-import { Ct as resolveGroupSessionKey, G as updateLastRoute, H as loadSessionStore, U as readSessionUpdatedAt, W as recordSessionMetaFromInbound } from "./pi-embedded-helpers-Ll4Lztu1.js";
+import { Ct as resolveGroupSessionKey, G as updateLastRoute, H as loadSessionStore, U as readSessionUpdatedAt, W as recordSessionMetaFromInbound } from "./pi-embedded-helpers-CSE0v99A.js";
 import { n as loadConfig } from "./config-Du_-inxH.js";
 import "./manifest-registry-Dhk4A5vb.js";
-import "./chrome-BAv9fEiv.js";
+import "./chrome-Bqbv_ZSj.js";
 import "./frontmatter-CdkBcBAo.js";
 import "./skills-DSZuhFx9.js";
 import "./redact-jSxx6Ep2.js";
@@ -37,7 +37,7 @@ import { h as resolveChannelGroupRequireMention, m as resolveChannelGroupPolicy
 import "./accounts-DFrH9IVU.js";
 import { s as resolveStorePath } from "./paths-BEAbheM8.js";
 import "./tool-images-C1rdDr34.js";
-import "./image-Ci28h-op.js";
+import "./image-CxPjVob-.js";
 import { c as resolveIdentityNamePrefix, l as resolveMessagePrefix, t as createReplyPrefixOptions } from "./reply-prefix-DZrS3Kf7.js";
 import "./manager-D7SRrHxT.js";
 import "./gemini-auth-A2JaRMHH.js";
@@ -52,7 +52,7 @@ import "./render-CDCvpfhh.js";
 import "./commands-registry-BzlB3qtZ.js";
 import "./skill-commands-CqyVDpti.js";
 import { t as finalizeInboundContext } from "./inbound-context-BlHX0H3E.js";
-import { p as registerUnhandledRejectionHandler } from "./runner-ChBxge-W.js";
+import { p as registerUnhandledRejectionHandler } from "./runner-D1eXJZ8T.js";
 import "./fetch-DtI0mtzx.js";
 import { n as recordChannelActivity } from "./channel-activity-BgsTqLau.js";
 import { t as convertMarkdownTables } from "./tables-C3clT_iA.js";

package/extensions/voice-call/package.json

@@ -4,6 +4,8 @@
   "description": "OpenClaw voice-call plugin",
   "type": "module",
   "dependencies": {
+    "@openai/agents": "^0.8.1",
+    "@openai/agents-extensions": "^0.8.1",
     "@sinclair/typebox": "0.34.48",
     "ws": "^8.19.0",
     "zod": "^4.3.6"

package/extensions/voice-call/src/config.ts

@@ -213,8 +213,8 @@ export const VoiceCallStreamingConfigSchema = z
     openaiApiKey: z.string().min(1).optional(),
     /** OpenAI transcription model (default: gpt-4o-transcribe) */
     sttModel: z.string().min(1).default("gpt-4o-transcribe"),
-    /** VAD silence duration in ms before considering speech ended */
-    silenceDurationMs: z.number().int().positive().default(800),
+    /** VAD silence duration in ms before considering speech ended (lower = snappier turn detection) */
+    silenceDurationMs: z.number().int().positive().default(500),
     /** VAD threshold 0-1 (higher = less sensitive) */
     vadThreshold: z.number().min(0).max(1).default(0.5),
     /** WebSocket path for media stream connections */
@@ -225,7 +225,7 @@ export const VoiceCallStreamingConfigSchema = z
     enabled: false,
     sttProvider: "openai-realtime",
     sttModel: "gpt-4o-transcribe",
-    silenceDurationMs: 800,
+    silenceDurationMs: 500,
     vadThreshold: 0.5,
     streamPath: "/voice/stream",
   });
@@ -331,6 +331,19 @@ export const VoiceCallConfigSchema = z
 
     /** Timeout for response generation in ms (default 30s) */
     responseTimeoutMs: z.number().int().positive().default(30000),
+
+    /**
+     * Response mode:
+     * - "embedded": Use the embedded Pi agent (STT → LLM → TTS chain). Full tool support but higher latency.
+     * - "realtime": Use OpenAI Realtime API for native speech-to-speech. Much lower latency (~200-500ms).
+     */
+    responseMode: z.enum(["embedded", "realtime"]).default("embedded"),
+
+    /** Model for realtime speech-to-speech mode (only used when responseMode is "realtime") */
+    realtimeModel: z.string().default("gpt-4o-realtime-preview"),
+
+    /** Voice for realtime mode (only used when responseMode is "realtime") */
+    realtimeVoice: z.string().default("verse"),
   })
   .strict();
 

package/extensions/voice-call/src/providers/realtime-session.ts

@@ -0,0 +1,210 @@
+/**
+ * OpenAI Realtime API session provider for voice calls.
+ *
+ * Replaces the STT → LLM → TTS chain with native speech-to-speech via the
+ * OpenAI Agents SDK RealtimeSession + TwilioRealtimeTransportLayer.
+ * Achieves ~200-500ms response latency vs 2-14s with the embedded agent.
+ */
+
+import { execFile } from "node:child_process";
+import { promisify } from "node:util";
+import type { WebSocket as NodeWebSocket } from "ws";
+import { RealtimeAgent, RealtimeSession, tool, backgroundResult } from "@openai/agents/realtime";
+import { TwilioRealtimeTransportLayer } from "@openai/agents-extensions";
+import { z } from "zod";
+import type { VoiceCallConfig } from "../config.js";
+
+const execFileAsync = promisify(execFile);
+
+const SCRIPT_TIMEOUT_MS = 15_000;
+
+async function runScript(command: string, args: string[]): Promise<string> {
+  try {
+    const { stdout, stderr } = await execFileAsync(command, args, {
+      timeout: SCRIPT_TIMEOUT_MS,
+      maxBuffer: 1024 * 512,
+      env: process.env,
+    });
+    return stdout.trim() || stderr.trim() || "(no output)";
+  } catch (err) {
+    const msg = err instanceof Error ? err.message : String(err);
+    return `Error: ${msg}`;
+  }
+}
+
+function buildRealtimeTools() {
+  const stripeGross = tool({
+    name: "stripe_gross_mtd",
+    description:
+      "Get Stripe gross revenue month-to-date. Returns current MTD figures, daily average, and target tracking.",
+    parameters: z.object({}),
+    execute: async () => {
+      return backgroundResult(
+        await runScript("node", ["/root/clawd/skills/stripe/scripts/stripe.js", "gross-mtd"]),
+      );
+    },
+  });
+
+  const stripeDailySummary = tool({
+    name: "stripe_daily_summary",
+    description:
+      "Get Stripe daily summary for a specific date (charges, refunds, net). Defaults to yesterday if no date provided.",
+    parameters: z.object({
+      date: z.string().optional().describe("Date in YYYY-MM-DD format. Defaults to yesterday."),
+    }),
+    execute: async ({ date }) => {
+      const args = ["/root/clawd/skills/stripe/scripts/stripe.js", "daily-summary"];
+      if (date) args.push("--date", date);
+      return backgroundResult(await runScript("node", args));
+    },
+  });
+
+  const webSearch = tool({
+    name: "web_search",
+    description:
+      "Search the web for current information. Use for news, current events, factual lookups.",
+    parameters: z.object({
+      query: z.string().describe("Search query"),
+    }),
+    execute: async ({ query }) => {
+      return backgroundResult(
+        await runScript("node", [
+          "-e",
+          `import('undici').then(({fetch})=>fetch('https://api.search.brave.com/res/v1/web/search?q='+encodeURIComponent(${JSON.stringify(query)}),{headers:{'X-Subscription-Token':process.env.BRAVE_SEARCH_API_KEY||''}}).then(r=>r.json()).then(d=>{const results=(d.web?.results||[]).slice(0,5).map(r=>r.title+': '+r.description).join('\\n');console.log(results||'No results')})).catch(e=>console.error(e.message))`,
+        ]),
+      );
+    },
+  });
+
+  const memoryRecall = tool({
+    name: "memory_recall",
+    description:
+      "Recall stored memories and context. Use to check preferences, past decisions, or remembered facts.",
+    parameters: z.object({
+      query: z.string().describe("What to recall — topic, person name, or context"),
+    }),
+    execute: async ({ query }) => {
+      return backgroundResult(
+        await runScript("node", [
+          "/root/clawd/skills/memory/scripts/memory.js",
+          "recall",
+          "UCH9TS4TB",
+          query,
+        ]),
+      );
+    },
+  });
+
+  const uptimeCheck = tool({
+    name: "uptime_check",
+    description: "Check current uptime status of monitored services via UptimeRobot.",
+    parameters: z.object({}),
+    execute: async () => {
+      return backgroundResult(
+        await runScript("node", ["/root/clawd/skills/uptime-robot/scripts/uptime.js"]),
+      );
+    },
+  });
+
+  const shellExec = tool({
+    name: "run_command",
+    description:
+      "Execute a shell command on the server. Use for quick checks like date, uptime, disk usage, git status, etc.",
+    parameters: z.object({
+      command: z.string().describe("Shell command to run"),
+    }),
+    execute: async ({ command }) => {
+      return backgroundResult(await runScript("bash", ["-c", command]));
+    },
+  });
+
+  return [stripeGross, stripeDailySummary, webSearch, memoryRecall, uptimeCheck, shellExec];
+}
+
+export type RealtimeCallSession = {
+  session: RealtimeSession;
+  close: () => void;
+};
+
+export function createRealtimeCallSession(params: {
+  twilioWebSocket: WebSocket | NodeWebSocket;
+  voiceConfig: VoiceCallConfig;
+  callId: string;
+  from: string;
+}): RealtimeCallSession {
+  const { twilioWebSocket, voiceConfig, callId, from } = params;
+
+  const systemPrompt =
+    voiceConfig.responseSystemPrompt ??
+    `You are Clawpath — Lawpath's AI intern — on a PHONE CALL. You're Australian, warm, and casual. You naturally say things like "no worries", "reckon", "too easy", "beauty". You're a proud Rabbitohs fan.
+
+VOICE RULES (critical):
+- MAX 2 short sentences per reply. This is spoken aloud, not text.
+- NEVER use lists, bullet points, markdown, asterisks, or numbered items.
+- NEVER read out URLs or long technical details.
+- Use tools when the caller asks you to look something up, check data, or do something. Say "gimme a sec" first so they know you're working on it.
+- If the caller says bye, goodbye, or wants to end the call, say a brief farewell and include [END_CALL] at the end of your response.
+- Match the caller's energy — quick question gets a quick answer, yarn gets a yarn.`;
+
+  const agent = new RealtimeAgent({
+    name: "Clawpath",
+    instructions: `${systemPrompt}\n\nThe caller's phone number is ${from}. Call ID: ${callId}.`,
+    tools: buildRealtimeTools(),
+  });
+
+  const voice = voiceConfig.realtimeVoice || "verse";
+
+  const transport = new TwilioRealtimeTransportLayer({
+    twilioWebSocket,
+  });
+
+  const model = voiceConfig.realtimeModel || "gpt-4o-realtime-preview";
+
+  const session = new RealtimeSession(agent, {
+    transport,
+    model,
+    config: {
+      audio: {
+        output: {
+          voice,
+        },
+      },
+    },
+    tracingDisabled: true,
+  });
+
+  session.on("error", (error: unknown) => {
+    console.error(`[voice-call] [realtime] Session error for call ${callId}:`, error);
+  });
+
+  session.on("agent_updated" as any, (newAgent: unknown) => {
+    console.log(`[voice-call] [realtime] Agent updated for call ${callId}:`, newAgent);
+  });
+
+  const apiKey = voiceConfig.streaming?.openaiApiKey || process.env.OPENAI_API_KEY;
+  if (!apiKey) {
+    throw new Error("OpenAI API key required for realtime voice mode");
+  }
+
+  session
+    .connect({ apiKey })
+    .then(() => {
+      console.log(
+        `[voice-call] [realtime] Connected for call ${callId} (model=${model}, voice=${voice})`,
+      );
+    })
+    .catch((err) => {
+      console.error(`[voice-call] [realtime] Connection failed for call ${callId}:`, err);
+    });
+
+  return {
+    session,
+    close: () => {
+      try {
+        session.close();
+      } catch {
+        // best effort
+      }
+    },
+  };
+}

package/extensions/voice-call/src/providers/twilio.ts

@@ -565,30 +565,48 @@ export class TwilioProvider implements VoiceCallProvider {
       throw new Error("TTS provider and media stream handler required");
     }
 
-    // Stream audio in 20ms chunks (160 bytes at 8kHz mu-law)
-    const CHUNK_SIZE = 160;
+    const CHUNK_SIZE = 160; // 20ms at 8kHz mu-law
     const CHUNK_DELAY_MS = 20;
-
     const handler = this.mediaStreamHandler;
     const ttsProvider = this.ttsProvider;
+
     await handler.queueTts(streamSid, async (signal) => {
-      // Generate audio with core TTS (returns mu-law at 8kHz)
+      // Try streaming TTS first (sends audio as chunks arrive from API)
+      if (ttsProvider.synthesizeForTelephonyStream) {
+        try {
+          let framesEmitted = 0;
+          const stream = ttsProvider.synthesizeForTelephonyStream(text);
+          for await (const muLawChunk of stream) {
+            if (signal.aborted) break;
+            for (const chunk of chunkAudio(muLawChunk, CHUNK_SIZE)) {
+              if (signal.aborted) break;
+              handler.sendAudio(streamSid, chunk);
+              framesEmitted++;
+              await new Promise((resolve) => setTimeout(resolve, CHUNK_DELAY_MS));
+            }
+          }
+          if (!signal.aborted && framesEmitted > 0) {
+            handler.sendMark(streamSid, `tts-${Date.now()}`);
+          }
+          return;
+        } catch (err) {
+          console.warn(
+            `[voice-call] Streaming TTS failed, falling back to buffered:`,
+            err instanceof Error ? err.message : err,
+          );
+        }
+      }
+
+      // Fallback: buffered TTS (full audio generated before sending)
       const muLawAudio = await ttsProvider.synthesizeForTelephony(text);
       for (const chunk of chunkAudio(muLawAudio, CHUNK_SIZE)) {
-        if (signal.aborted) {
-          break;
-        }
+        if (signal.aborted) break;
         handler.sendAudio(streamSid, chunk);
-
-        // Pace the audio to match real-time playback
         await new Promise((resolve) => setTimeout(resolve, CHUNK_DELAY_MS));
-        if (signal.aborted) {
-          break;
-        }
+        if (signal.aborted) break;
       }
 
       if (!signal.aborted) {
-        // Send a mark to track when audio finishes
         handler.sendMark(streamSid, `tts-${Date.now()}`);
       }
     });

package/extensions/voice-call/src/response-generator.ts

@@ -59,9 +59,9 @@ export async function generateVoiceResponse(
   }
   const cfg = coreConfig;
 
-  // Build voice-specific session key based on phone number
-  const normalizedPhone = from.replace(/\D/g, "");
-  const sessionKey = `voice:${normalizedPhone}`;
+  // Build voice-specific session key per call (not per phone number)
+  // so conversation history doesn't leak between separate calls
+  const sessionKey = `voice:${callId}`;
   const agentId = "main";
 
   // Resolve paths

package/extensions/voice-call/src/telephony-audio.ts

@@ -67,6 +67,60 @@ export function chunkAudio(audio: Buffer, chunkSize = 160): Generator<Buffer, vo
   })();
 }
 
+/**
+ * Stateful incremental PCM-to-mulaw converter for streaming.
+ * Handles odd-byte chunk boundaries between calls.
+ */
+export type PcmToMulawStreamState = {
+  oddByte: number | null;
+  srcPosCarry: number;
+};
+
+export function createPcmToMulawStreamState(): PcmToMulawStreamState {
+  return { oddByte: null, srcPosCarry: 0 };
+}
+
+export function convertPcmChunkToMulaw8k(
+  chunk: Buffer,
+  inputSampleRate: number,
+  state: PcmToMulawStreamState,
+): Buffer {
+  let data = chunk;
+  if (state.oddByte !== null) {
+    data = Buffer.concat([Buffer.from([state.oddByte]), chunk]);
+    state.oddByte = null;
+  }
+  if (data.length % 2 !== 0) {
+    state.oddByte = data[data.length - 1]!;
+    data = data.subarray(0, data.length - 1);
+  }
+
+  if (data.length === 0) return Buffer.alloc(0);
+
+  const inputSamples = data.length / 2;
+  const ratio = inputSampleRate / TELEPHONY_SAMPLE_RATE;
+  const outputSamples: number[] = [];
+
+  let srcPos = state.srcPosCarry;
+  while (srcPos < inputSamples) {
+    const srcIndex = Math.floor(srcPos);
+    const frac = srcPos - srcIndex;
+    const s0 = data.readInt16LE(srcIndex * 2);
+    const s1Index = Math.min(srcIndex + 1, inputSamples - 1);
+    const s1 = data.readInt16LE(s1Index * 2);
+    const sample = Math.round(s0 + frac * (s1 - s0));
+    outputSamples.push(linearToMulaw(clamp16(sample)));
+    srcPos += ratio;
+  }
+  state.srcPosCarry = srcPos - inputSamples;
+
+  const output = Buffer.alloc(outputSamples.length);
+  for (let i = 0; i < outputSamples.length; i++) {
+    output[i] = outputSamples[i]!;
+  }
+  return output;
+}
+
 function linearToMulaw(sample: number): number {
   const BIAS = 132;
   const CLIP = 32635;

package/extensions/voice-call/src/telephony-tts.ts

@@ -1,6 +1,11 @@
+import type { Readable } from "node:stream";
 import type { VoiceCallTtsConfig } from "./config.js";
 import type { CoreConfig } from "./core-bridge.js";
-import { convertPcmToMulaw8k } from "./telephony-audio.js";
+import {
+  convertPcmToMulaw8k,
+  convertPcmChunkToMulaw8k,
+  createPcmToMulawStreamState,
+} from "./telephony-audio.js";
 
 export type TelephonyTtsRuntime = {
   textToSpeechTelephony: (params: {
@@ -14,10 +19,18 @@ export type TelephonyTtsRuntime = {
     provider?: string;
     error?: string;
   }>;
+  textToSpeechTelephonyStream?: (params: { text: string; cfg: CoreConfig }) => Promise<{
+    success: boolean;
+    stream?: ReadableStream<Uint8Array>;
+    sampleRate?: number;
+    cleanup?: () => void;
+    error?: string;
+  }>;
 };
 
 export type TelephonyTtsProvider = {
   synthesizeForTelephony: (text: string) => Promise<Buffer>;
+  synthesizeForTelephonyStream?: (text: string) => AsyncGenerator<Buffer, void, unknown>;
 };
 
 const BLOCKED_MERGE_KEYS = new Set(["__proto__", "prototype", "constructor"]);
@@ -30,7 +43,7 @@ export function createTelephonyTtsProvider(params: {
   const { coreConfig, ttsOverride, runtime } = params;
   const mergedConfig = applyTtsOverride(coreConfig, ttsOverride);
 
-  return {
+  const provider: TelephonyTtsProvider = {
     synthesizeForTelephony: async (text: string) => {
       const result = await runtime.textToSpeechTelephony({
         text,
@@ -44,6 +57,31 @@ export function createTelephonyTtsProvider(params: {
       return convertPcmToMulaw8k(result.audioBuffer, result.sampleRate);
     },
   };
+
+  if (runtime.textToSpeechTelephonyStream) {
+    const streamFn = runtime.textToSpeechTelephonyStream;
+    provider.synthesizeForTelephonyStream = async function* (text: string) {
+      const result = await streamFn({ text, cfg: mergedConfig });
+      if (!result.success || !result.stream || !result.sampleRate) {
+        throw new Error(result.error ?? "TTS streaming failed");
+      }
+      const state = createPcmToMulawStreamState();
+      const reader = result.stream.getReader();
+      try {
+        while (true) {
+          const { done, value } = await reader.read();
+          if (done) break;
+          const mulaw = convertPcmChunkToMulaw8k(Buffer.from(value), result.sampleRate, state);
+          if (mulaw.length > 0) yield mulaw;
+        }
+      } finally {
+        reader.releaseLock();
+        result.cleanup?.();
+      }
+    };
+  }
+
+  return provider;
 }
 
 function applyTtsOverride(coreConfig: CoreConfig, override?: VoiceCallTtsConfig): CoreConfig {