@komaa/standin-sdk 0.1.0

package/dist/plugins/openclaw/realtime.js

@@ -0,0 +1,238 @@
+// Copyright (c) 2026 Komaa DigiTech
+// SPDX-License-Identifier: MIT
+/**
+ * One Microsoft Teams call, bridged to OpenClaw's realtime speech-to-speech session.
+ *
+ * This file is the reason the plugin cannot be a standalone worker. It does not
+ * drive a model: it CONSUMES the host's realtime session
+ * (`createRealtimeVoiceBridgeSession`), the host's provider registry and the
+ * host's config, all of which are in-process objects inside the OpenClaw gateway.
+ * Nothing here opens a provider socket, and nothing here knows which vendor is on
+ * the other end.
+ *
+ * What it does own is the audio boundary between two rates and two framings:
+ *
+ *     caller   16 kHz --> 24 kHz --> host realtime session
+ *     model    24 kHz --> 16 kHz --> session.sendAudio
+ *
+ * and the three things that make a spoken call feel right rather than merely
+ * work: the echo guard, the playout clock the guard keys on, and barge-in.
+ *
+ * Everything the SDK owns is deliberately absent here: the socket, the sequence
+ * number, the outbound timeline, the frame loop and the watchdogs.
+ */
+import { FrameAligner, REALTIME_SAMPLE_RATE_HZ, SAMPLE_RATE_HZ, frameDurationMs, resamplePcm16, } from "../../index.js";
+import { REALTIME_VOICE_AUDIO_FORMAT_PCM16_24KHZ, createRealtimeVoiceBridgeSession, } from "openclaw/plugin-sdk/realtime-voice";
+import { shouldSuppressEcho } from "../../echoGuard.js";
+import { isVerbalInterrupt } from "../../gate.js";
+/**
+ * Build the bridge for one call.
+ *
+ * `session` is the SDK's {@link CallSession}: the only route to the caller, and
+ * the owner of the sequence number and the outbound timeline, so nothing here
+ * tracks either.
+ */
+export function createRealtimeCall(params) {
+    const { session, deps } = params;
+    const { logger } = deps;
+    const callId = session.callId;
+    let closed = false;
+    /**
+     * Estimated epoch ms at which the audio already sent finishes PLAYING.
+     *
+     * The model generates faster than realtime and the service queues what we send,
+     * so send-time is not play-time. Summing each chunk's own duration tracks the
+     * playout clock instead, which is what the echo guard has to be keyed on: a
+     * guard armed off last-send disarms in the middle of the sentence it exists to
+     * cover.
+     */
+    let playbackEndAt = 0;
+    /**
+     * False until the caller has genuinely spoken once. See
+     * {@link EchoGuardOptions.allowBargeIn}: the opening greeting echoing off a
+     * speakerphone will clear the barge-in RMS, so the loudness exception has to
+     * stay off until there is a real caller turn to measure against.
+     */
+    let callerTurnStarted = false;
+    /**
+     * Microsoft Teams recording status, seeded from `session.start` and updated by context.
+     * Gates every media-derived path (audio in, DTMF, transcripts) when
+     * `requireRecordingStatus` is on.
+     */
+    let recordingActive = session.start.recordingStatus === "active";
+    const recordingGateBlocks = () => deps.requireRecordingStatus && !recordingActive;
+    /**
+     * Whole-frame alignment for the downlink.
+     *
+     * A 24 kHz delta resampled to 16 kHz does not divide evenly into the wire's
+     * 640-byte frame, and dropping the remainder clips a few milliseconds off every
+     * chunk seam - audible over a call as clipped word endings. The SDK owns this;
+     * do not hand-roll it.
+     */
+    const aligner = new FrameAligner();
+    /**
+     * The host's audio sink is SYNCHRONOUS and `session.sendAudio` is not, so the
+     * sends are chained rather than awaited. A chain, not fire-and-forget: two
+     * unordered sends would put the model's voice out of order on the wire, which
+     * is worse than late audio. The catch is inside the chain so one failed send
+     * cannot poison every send after it.
+     */
+    let sendChain = Promise.resolve();
+    const enqueue = (op) => {
+        sendChain = sendChain.then(op).catch((err) => {
+            logger?.debug?.(`standin-msteams: send failed on ${callId} - ${err instanceof Error ? err.message : String(err)}`);
+        });
+    };
+    /** Flush the held-back tail of a finished turn, zero-padded to a whole frame. */
+    const flushTail = () => {
+        const tail = aligner.flush();
+        if (tail)
+            enqueue(() => session.sendAudio(tail));
+    };
+    /**
+     * Barge-in. THE ORDER HERE IS THE WHOLE POINT.
+     *
+     * `cancelPlayback` is the only lever that un-sends audio the service has already
+     * buffered, so it goes FIRST: cancel the model first and the caller still hears
+     * the rest of the interrupted turn play out, which is exactly the "it kept
+     * talking over me" complaint barge-in exists to fix. The aligner's residual
+     * belongs to that same dead turn, so it is dropped rather than flushed.
+     */
+    const flushPlayback = () => {
+        aligner.reset();
+        enqueue(() => session.cancelPlayback());
+        // The flush stops playout immediately, so the playout estimate collapses to now.
+        playbackEndAt = Date.now();
+    };
+    const realtime = createRealtimeVoiceBridgeSession({
+        provider: deps.provider,
+        providerConfig: deps.providerConfig,
+        cfg: deps.cfg,
+        audioFormat: REALTIME_VOICE_AUDIO_FORMAT_PCM16_24KHZ,
+        instructions: deps.instructions,
+        initialGreetingInstructions: deps.greetingInstructions,
+        triggerGreetingOnReady: Boolean(deps.greetingInstructions),
+        autoRespondToAudio: true,
+        interruptResponseOnInputAudio: true,
+        audioSink: {
+            isOpen: () => !closed,
+            sendAudio: (pcm24k) => {
+                if (closed || pcm24k.length === 0) {
+                    return;
+                }
+                const pcm16k = resamplePcm16(pcm24k, REALTIME_SAMPLE_RATE_HZ, SAMPLE_RATE_HZ);
+                playbackEndAt =
+                    Math.max(playbackEndAt, Date.now()) + frameDurationMs(pcm16k);
+                for (const frame of aligner.push(pcm16k)) {
+                    enqueue(() => session.sendAudio(frame));
+                }
+            },
+            // CANCEL SITE 1: the model truncated its own turn because it heard the
+            // caller start speaking. Upstream is already cancelled - the model is the
+            // one telling us - so the only thing left to undo is the audio the service
+            // still has queued.
+            clearAudio: () => {
+                flushPlayback();
+            },
+        },
+        onTranscript: (role, text, isFinal) => {
+            // The caller's first real speech ends the opening echo-only window and
+            // restores normal RMS barge-in for the rest of the call.
+            if (role === "user" && text.trim().length > 0) {
+                callerTurnStarted = true;
+            }
+            if (role === "assistant" && isFinal) {
+                // End of turn: the residual is real audio, not a fragment to discard.
+                flushTail();
+            }
+            if (role !== "user" || !isFinal) {
+                return;
+            }
+            // CANCEL SITE 2: a deterministic verbal interrupt ("stop", "hold on").
+            // Handled here rather than left to the model because the model is
+            // mid-generation when it arrives; matching the phrase ourselves is what
+            // makes the cut feel instant. Only while we are actually still speaking -
+            // otherwise "stop" is just a word in a sentence.
+            if (Date.now() < playbackEndAt && isVerbalInterrupt(text)) {
+                logger?.debug?.(`standin-msteams: verbal interrupt on ${callId} - flushing playback`);
+                flushPlayback();
+            }
+        },
+        onError: (error) => {
+            logger?.warn?.(`standin-msteams: realtime session error on ${callId} - ${error.message}`);
+        },
+    });
+    return {
+        connect: () => realtime.connect(),
+        pushAudio: (pcm16k) => {
+            if (closed || pcm16k.length === 0) {
+                return;
+            }
+            // Recording gate: caller audio is call media, and the Microsoft Media Access
+            // API says a bot does not process it before recording status is active. A
+            // no-op when requireRecordingStatus is off.
+            if (recordingGateBlocks()) {
+                return;
+            }
+            if (shouldSuppressEcho(pcm16k, playbackEndAt, {
+                ...deps.echo,
+                allowBargeIn: callerTurnStarted,
+            })) {
+                return;
+            }
+            realtime.sendAudio(resamplePcm16(pcm16k, SAMPLE_RATE_HZ, REALTIME_SAMPLE_RATE_HZ));
+        },
+        pushContext: (text) => {
+            if (closed || !text.trim()) {
+                return;
+            }
+            // DTMF and participant counts are media-derived too, so they sit behind the
+            // same gate as audio. Recording-status context itself is handled by the
+            // handler before it reaches here - that one must never be gated on itself.
+            if (recordingGateBlocks()) {
+                return;
+            }
+            realtime.sendUserMessage(text);
+        },
+        setRecordingActive: (active) => {
+            recordingActive = active;
+        },
+        interrupt: () => {
+            if (closed)
+                return;
+            try {
+                realtime.handleBargeIn();
+            }
+            catch (err) {
+                // Non-fatal: handleBargeIn throws when nothing is speaking, which is a
+                // perfectly ordinary state to interrupt from.
+                logger?.debug?.(`standin-msteams: barge-in on ${callId} - ${err instanceof Error ? err.message : String(err)}`);
+            }
+        },
+        say: (text) => {
+            if (closed || !text.trim())
+                return;
+            try {
+                // Same path as the greeting: inject the line as an instruction and trigger
+                // a spoken response. There is no separate "speak this" surface.
+                realtime.triggerGreeting(text);
+            }
+            catch (err) {
+                logger?.warn?.(`standin-msteams: say failed on ${callId} - ${err instanceof Error ? err.message : String(err)}`);
+            }
+        },
+        close: () => {
+            if (closed)
+                return;
+            closed = true;
+            aligner.reset();
+            try {
+                realtime.close();
+            }
+            catch (err) {
+                logger?.debug?.(`standin-msteams: realtime close on ${callId} - ${err instanceof Error ? err.message : String(err)}`);
+            }
+        },
+    };
+}
+//# sourceMappingURL=realtime.js.map

package/dist/plugins/openclaw/realtime.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"realtime.js","sourceRoot":"","sources":["../../../src/plugins/openclaw/realtime.ts"],"names":[],"mappings":"AAAA,oCAAoC;AACpC,+BAA+B;AAE/B;;;;;;;;;;;;;;;;;;;;GAoBG;AAEH,OAAO,EACL,YAAY,EACZ,uBAAuB,EACvB,cAAc,EACd,eAAe,EACf,aAAa,GAEd,MAAM,gBAAgB,CAAC;AAExB,OAAO,EACL,uCAAuC,EACvC,gCAAgC,GAGjC,MAAM,oCAAoC,CAAC;AAE5C,OAAO,EAAE,kBAAkB,EAAyB,MAAM,oBAAoB,CAAC;AAC/E,OAAO,EAAE,iBAAiB,EAAE,MAAM,eAAe,CAAC;AAkDlD;;;;;;GAMG;AACH,MAAM,UAAU,kBAAkB,CAAC,MAGlC;IACC,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,GAAG,MAAM,CAAC;IACjC,MAAM,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;IACxB,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC;IAE9B,IAAI,MAAM,GAAG,KAAK,CAAC;IAEnB;;;;;;;;OAQG;IACH,IAAI,aAAa,GAAG,CAAC,CAAC;IAEtB;;;;;OAKG;IACH,IAAI,iBAAiB,GAAG,KAAK,CAAC;IAE9B;;;;OAIG;IACH,IAAI,eAAe,GAAG,OAAO,CAAC,KAAK,CAAC,eAAe,KAAK,QAAQ,CAAC;IACjE,MAAM,mBAAmB,GAAG,GAAY,EAAE,CACxC,IAAI,CAAC,sBAAsB,IAAI,CAAC,eAAe,CAAC;IAElD;;;;;;;OAOG;IACH,MAAM,OAAO,GAAG,IAAI,YAAY,EAAE,CAAC;IAEnC;;;;;;OAMG;IACH,IAAI,SAAS,GAAkB,OAAO,CAAC,OAAO,EAAE,CAAC;IACjD,MAAM,OAAO,GAAG,CAAC,EAAuB,EAAQ,EAAE;QAChD,SAAS,GAAG,SAAS,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,GAAY,EAAE,EAAE;YACpD,MAAM,EAAE,KAAK,EAAE,CACb,mCAAmC,MAAM,MAAM,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAClG,CAAC;QACJ,CAAC,CAAC,CAAC;IACL,CAAC,CAAC;IAEF,iFAAiF;IACjF,MAAM,SAAS,GAAG,GAAS,EAAE;QAC3B,MAAM,IAAI,GAAG,OAAO,CAAC,KAAK,EAAE,CAAC;QAC7B,IAAI,IAAI;YAAE,OAAO,CAAC,GAAG,EAAE,CAAC,OAAO,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;IACnD,CAAC,CAAC;IAEF;;;;;;;;OAQG;IACH,MAAM,aAAa,GAAG,GAAS,EAAE;QAC/B,OAAO,CAAC,KAAK,EAAE,CAAC;QAChB,OAAO,CAAC,GAAG,EAAE,CAAC,OAAO,CAAC,cAAc,EAAE,CAAC,CAAC;QACxC,iFAAiF;QACjF,aAAa,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAC7B,CAAC,CAAC;IAEF,MAAM,QAAQ,GAAG,gCAAgC,CAAC;QAChD,QAAQ,EAAE,IAAI,CAAC,QAAQ;QACvB,cAAc,EAAE,IAAI,CAAC,cAAc;QACnC,GAAG,EAAE,IAAI,CAAC,GAAG;QACb,WAAW,EAAE,uCAAuC;QACpD,YAAY,EAAE,IAAI,CAAC,YAAY;QAC/B,2BAA2B,EAAE,IAAI,CAAC,oBAAoB;QACtD,sBAAsB,EAAE,OAAO,CAAC,IAAI,CAAC,oBAAoB,CAAC;QAC1D,kBAAkB,EAAE,IAAI;QACxB,6BAA6B,EAAE,IAAI;QACnC,SAAS,EAAE;YACT,MAAM,EAAE,GAAG,EAAE,CAAC,CAAC,MAAM;YACrB,SAAS,EAAE,CAAC,MAAc,EAAE,EAAE;gBAC5B,IAAI,MAAM,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;oBAClC,OAAO;gBACT,CAAC;gBACD,MAAM,MAAM,GAAG,aAAa,CAC1B,MAAM,EACN,uBAAuB,EACvB,cAAc,CACf,CAAC;gBACF,aAAa;oBACX,IAAI,CAAC,GAAG,CAAC,aAAa,EAAE,IAAI,CAAC,GAAG,EAAE,CAAC,GAAG,eAAe,CAAC,MAAM,CAAC,CAAC;gBAChE,KAAK,MAAM,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;oBACzC,OAAO,CAAC,GAAG,EAAE,CAAC,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;gBAC1C,CAAC;YACH,CAAC;YACD,uEAAuE;YACvE,0EAA0E;YAC1E,2EAA2E;YAC3E,oBAAoB;YACpB,UAAU,EAAE,GAAG,EAAE;gBACf,aAAa,EAAE,CAAC;YAClB,CAAC;SACF;QACD,YAAY,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,OAAO,EAAE,EAAE;YACpC,uEAAuE;YACvE,yDAAyD;YACzD,IAAI,IAAI,KAAK,MAAM,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBAC9C,iBAAiB,GAAG,IAAI,CAAC;YAC3B,CAAC;YACD,IAAI,IAAI,KAAK,WAAW,IAAI,OAAO,EAAE,CAAC;gBACpC,sEAAsE;gBACtE,SAAS,EAAE,CAAC;YACd,CAAC;YACD,IAAI,IAAI,KAAK,MAAM,IAAI,CAAC,OAAO,EAAE,CAAC;gBAChC,OAAO;YACT,CAAC;YACD,uEAAuE;YACvE,kEAAkE;YAClE,wEAAwE;YACxE,0EAA0E;YAC1E,iDAAiD;YACjD,IAAI,IAAI,CAAC,GAAG,EAAE,GAAG,aAAa,IAAI,iBAAiB,CAAC,IAAI,CAAC,EAAE,CAAC;gBAC1D,MAAM,EAAE,KAAK,EAAE,CACb,wCAAwC,MAAM,sBAAsB,CACrE,CAAC;gBACF,aAAa,EAAE,CAAC;YAClB,CAAC;QACH,CAAC;QACD,OAAO,EAAE,CAAC,KAAY,EAAE,EAAE;YACxB,MAAM,EAAE,IAAI,EAAE,CACZ,8CAA8C,MAAM,MAAM,KAAK,CAAC,OAAO,EAAE,CAC1E,CAAC;QACJ,CAAC;KACF,CAAC,CAAC;IAEH,OAAO;QACL,OAAO,EAAE,GAAG,EAAE,CAAC,QAAQ,CAAC,OAAO,EAAE;QAEjC,SAAS,EAAE,CAAC,MAAc,EAAE,EAAE;YAC5B,IAAI,MAAM,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;gBAClC,OAAO;YACT,CAAC;YACD,6EAA6E;YAC7E,0EAA0E;YAC1E,4CAA4C;YAC5C,IAAI,mBAAmB,EAAE,EAAE,CAAC;gBAC1B,OAAO;YACT,CAAC;YACD,IACE,kBAAkB,CAAC,MAAM,EAAE,aAAa,EAAE;gBACxC,GAAG,IAAI,CAAC,IAAI;gBACZ,YAAY,EAAE,iBAAiB;aAChC,CAAC,EACF,CAAC;gBACD,OAAO;YACT,CAAC;YACD,QAAQ,CAAC,SAAS,CAChB,aAAa,CAAC,MAAM,EAAE,cAAc,EAAE,uBAAuB,CAAC,CAC/D,CAAC;QACJ,CAAC;QAED,WAAW,EAAE,CAAC,IAAY,EAAE,EAAE;YAC5B,IAAI,MAAM,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC;gBAC3B,OAAO;YACT,CAAC;YACD,4EAA4E;YAC5E,wEAAwE;YACxE,2EAA2E;YAC3E,IAAI,mBAAmB,EAAE,EAAE,CAAC;gBAC1B,OAAO;YACT,CAAC;YACD,QAAQ,CAAC,eAAe,CAAC,IAAI,CAAC,CAAC;QACjC,CAAC;QAED,kBAAkB,EAAE,CAAC,MAAe,EAAE,EAAE;YACtC,eAAe,GAAG,MAAM,CAAC;QAC3B,CAAC;QAED,SAAS,EAAE,GAAG,EAAE;YACd,IAAI,MAAM;gBAAE,OAAO;YACnB,IAAI,CAAC;gBACH,QAAQ,CAAC,aAAa,EAAE,CAAC;YAC3B,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,uEAAuE;gBACvE,8CAA8C;gBAC9C,MAAM,EAAE,KAAK,EAAE,CACb,gCAAgC,MAAM,MAAM,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAC/F,CAAC;YACJ,CAAC;QACH,CAAC;QAED,GAAG,EAAE,CAAC,IAAY,EAAE,EAAE;YACpB,IAAI,MAAM,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE;gBAAE,OAAO;YACnC,IAAI,CAAC;gBACH,2EAA2E;gBAC3E,gEAAgE;gBAChE,QAAQ,CAAC,eAAe,CAAC,IAAI,CAAC,CAAC;YACjC,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,MAAM,EAAE,IAAI,EAAE,CACZ,kCAAkC,MAAM,MAAM,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CACjG,CAAC;YACJ,CAAC;QACH,CAAC;QAED,KAAK,EAAE,GAAG,EAAE;YACV,IAAI,MAAM;gBAAE,OAAO;YACnB,MAAM,GAAG,IAAI,CAAC;YACd,OAAO,CAAC,KAAK,EAAE,CAAC;YAChB,IAAI,CAAC;gBACH,QAAQ,CAAC,KAAK,EAAE,CAAC;YACnB,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,MAAM,EAAE,KAAK,EAAE,CACb,sCAAsC,MAAM,MAAM,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CACrG,CAAC;YACJ,CAAC;QACH,CAAC;KACF,CAAC;AACJ,CAAC"}

package/dist/plugins/openclaw/runtime.d.ts

@@ -0,0 +1,19 @@
+import type { OpenClawPluginApi } from "openclaw/plugin-sdk/core";
+import type { ResolvedPluginConfig } from "./config.js";
+/** The service OpenClaw starts and stops. One per gateway process. */
+export declare class StandInCallRuntime {
+    #private;
+    constructor(api: OpenClawPluginApi, config: ResolvedPluginConfig);
+    /** Live calls right now. Useful to a health check, and to a test. */
+    get activeCalls(): number;
+    start(): Promise<void>;
+    /**
+     * Close live calls first, then the listener.
+     *
+     * That order is not cosmetic: the SDK's teardown frees a call's slot only once
+     * its close settles, so tearing the listener down over live calls is how a
+     * shutdown ends up waiting on sockets nobody is reading any more.
+     */
+    stop(): Promise<void>;
+}
+//# sourceMappingURL=runtime.d.ts.map

package/dist/plugins/openclaw/runtime.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"runtime.d.ts","sourceRoot":"","sources":["../../../src/plugins/openclaw/runtime.ts"],"names":[],"mappings":"AA2BA,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,0BAA0B,CAAC;AAGlE,OAAO,KAAK,EAAE,oBAAoB,EAAE,MAAM,aAAa,CAAC;AAkDxD,sEAAsE;AACtE,qBAAa,kBAAkB;;gBAOjB,GAAG,EAAE,iBAAiB,EAAE,MAAM,EAAE,oBAAoB;IAMhE,qEAAqE;IACrE,IAAI,WAAW,IAAI,MAAM,CAExB;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IA6C5B;;;;;;OAMG;IACG,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;CAmC5B"}

package/dist/plugins/openclaw/runtime.js

@@ -0,0 +1,160 @@
+// Copyright (c) 2026 Komaa DigiTech
+// SPDX-License-Identifier: MIT
+/**
+ * The plugin's background service: one call listener, shared by every call.
+ *
+ * OpenClaw starts this on boot and stops it on shutdown or reload, so it is the
+ * teardown hook. It is short because the WebSocket server, the handshake, the
+ * replay guard, the frame loop and the watchdogs all belong to the SDK's
+ * `CallServer`. What is left is the two things the SDK cannot know: which
+ * realtime provider the host resolved, and how many calls this operator agreed
+ * to take at once.
+ *
+ * Two frictions with the SDK's defaults are handled here on purpose, both by
+ * passing the value explicitly rather than letting the default win:
+ *
+ * - `CallServer` reads `STANDIN_SECRET` from the environment. The secret arrives
+ *   through `api.pluginConfig`, having been resolved by the host from a secret
+ *   reference, so it is passed in. An env var would silently override a portal
+ *   value the operator can actually see.
+ * - `CallServer` binds 0.0.0.0, which suits a worker behind an ingress but not a
+ *   gateway reached through a local tunnel. Neither is wrong, so `bindAddress` is
+ *   passed when the operator set it and the SDK's default stands when they did not.
+ */
+import { CallServer } from "../../index.js";
+import { resolveConfiguredRealtimeVoiceProvider } from "openclaw/plugin-sdk/realtime-voice";
+import { TeamsCallHandler, } from "./handler.js";
+/**
+ * Live calls, and the operator's concurrency cap.
+ *
+ * A slot is reserved BEFORE the model session is built, so two calls arriving
+ * together cannot both pass the check and then both open a provider socket. A
+ * reserved-but-unbound entry is a call still in `onStart`.
+ */
+class LiveCalls {
+    #max;
+    #calls = new Map();
+    constructor(max) {
+        this.#max = max;
+    }
+    get size() {
+        return this.#calls.size;
+    }
+    acquire(callId) {
+        if (this.#calls.size >= this.#max)
+            return false;
+        this.#calls.set(callId, undefined);
+        return true;
+    }
+    bind(callId, call) {
+        // Only if the slot is still ours: a teardown that raced onStart has already
+        // released it, and re-adding here would leak the slot for the worker's life.
+        if (this.#calls.has(callId))
+            this.#calls.set(callId, call);
+    }
+    release(callId) {
+        this.#calls.delete(callId);
+    }
+    /** Close every live call. Used by shutdown, before the listener goes away. */
+    closeAll() {
+        for (const call of this.#calls.values())
+            call?.close();
+        this.#calls.clear();
+    }
+}
+/** The service OpenClaw starts and stops. One per gateway process. */
+export class StandInCallRuntime {
+    #api;
+    #config;
+    #calls;
+    #server;
+    #realtime;
+    constructor(api, config) {
+        this.#api = api;
+        this.#config = config;
+        this.#calls = new LiveCalls(config.media.maxConcurrentCalls);
+    }
+    /** Live calls right now. Useful to a health check, and to a test. */
+    get activeCalls() {
+        return this.#calls.size;
+    }
+    async start() {
+        const log = this.#api.logger;
+        this.#realtime = this.#resolveRealtime();
+        // Fail LOUD, not closed. A gateway that boots healthy and then refuses every
+        // call with "realtime-unavailable" is the single most expensive way to
+        // discover a missing API key, so name the configured provider here where
+        // someone is reading the log, rather than only on the first call that fails.
+        if (!this.#realtime) {
+            const providerId = this.#config.voice.realtime.provider;
+            log.warn("standin-msteams: no realtime voice provider resolved" +
+                (providerId
+                    ? ` (configured provider "${providerId}" has no usable credentials)`
+                    : " (no realtime provider configured)") +
+                '. Every call will be refused with "realtime-unavailable". Set the provider API key.');
+        }
+        const server = new CallServer({
+            handlerFactory: () => new TeamsCallHandler({
+                config: this.#config,
+                realtime: this.#realtime,
+                cfg: this.#api.config,
+                registry: this.#calls,
+                logger: log,
+            }),
+            secret: this.#config.media.secret,
+            port: this.#config.media.port,
+            wsPath: this.#config.media.path,
+            maxConnections: this.#config.media.maxConcurrentCalls,
+            ...(this.#config.media.bindAddress
+                ? { host: this.#config.media.bindAddress }
+                : {}),
+        });
+        await server.start();
+        this.#server = server;
+        log.info(`standin-msteams: listening on ${this.#config.media.bindAddress ?? "0.0.0.0"}:${server.port}${server.wsPath}` +
+            ` (max ${this.#config.media.maxConcurrentCalls} concurrent)`);
+    }
+    /**
+     * Close live calls first, then the listener.
+     *
+     * That order is not cosmetic: the SDK's teardown frees a call's slot only once
+     * its close settles, so tearing the listener down over live calls is how a
+     * shutdown ends up waiting on sockets nobody is reading any more.
+     */
+    async stop() {
+        this.#calls.closeAll();
+        const server = this.#server;
+        this.#server = undefined;
+        if (server)
+            await server.aclose();
+        this.#api.logger.info("standin-msteams: stopped");
+    }
+    /**
+     * Ask the host which realtime provider is configured.
+     *
+     * Returns undefined rather than throwing on ANY failure, including a throw from
+     * the host resolver: a provider that cannot be resolved is a call that gets
+     * refused, not a gateway that will not boot. The plugin is one service among
+     * many in that process.
+     */
+    #resolveRealtime() {
+        try {
+            const resolved = resolveConfiguredRealtimeVoiceProvider({
+                configuredProviderId: this.#config.voice.realtime.provider,
+                providerConfigs: this.#config.voice.realtime.providers,
+                cfg: this.#api.config,
+            });
+            if (!resolved?.provider || !resolved.providerConfig)
+                return undefined;
+            return {
+                provider: resolved.provider,
+                providerConfig: resolved.providerConfig,
+            };
+        }
+        catch (err) {
+            this.#api.logger.warn(`standin-msteams: realtime provider resolution failed - ${err instanceof Error ? err.message : String(err)}`);
+            return undefined;
+        }
+    }
+}
+//# sourceMappingURL=runtime.js.map

package/dist/plugins/openclaw/runtime.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"runtime.js","sourceRoot":"","sources":["../../../src/plugins/openclaw/runtime.ts"],"names":[],"mappings":"AAAA,oCAAoC;AACpC,+BAA+B;AAE/B;;;;;;;;;;;;;;;;;;;;GAoBG;AAEH,OAAO,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAG5C,OAAO,EAAE,sCAAsC,EAAE,MAAM,oCAAoC,CAAC;AAG5F,OAAO,EACL,gBAAgB,GAGjB,MAAM,cAAc,CAAC;AAGtB;;;;;;GAMG;AACH,MAAM,SAAS;IACJ,IAAI,CAAS;IACb,MAAM,GAAG,IAAI,GAAG,EAAoC,CAAC;IAE9D,YAAY,GAAW;QACrB,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC;IAClB,CAAC;IAED,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;IAC1B,CAAC;IAED,OAAO,CAAC,MAAc;QACpB,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI;YAAE,OAAO,KAAK,CAAC;QAChD,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,MAAM,EAAE,SAAS,CAAC,CAAC;QACnC,OAAO,IAAI,CAAC;IACd,CAAC;IAED,IAAI,CAAC,MAAc,EAAE,IAAkB;QACrC,4EAA4E;QAC5E,6EAA6E;QAC7E,IAAI,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;YAAE,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;IAC7D,CAAC;IAED,OAAO,CAAC,MAAc;QACpB,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;IAC7B,CAAC;IAED,8EAA8E;IAC9E,QAAQ;QACN,KAAK,MAAM,IAAI,IAAI,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE;YAAE,IAAI,EAAE,KAAK,EAAE,CAAC;QACvD,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC;IACtB,CAAC;CACF;AAED,sEAAsE;AACtE,MAAM,OAAO,kBAAkB;IACpB,IAAI,CAAoB;IACxB,OAAO,CAAuB;IAC9B,MAAM,CAAY;IAC3B,OAAO,CAAyB;IAChC,SAAS,CAA+B;IAExC,YAAY,GAAsB,EAAE,MAA4B;QAC9D,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC;QAChB,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC;QACtB,IAAI,CAAC,MAAM,GAAG,IAAI,SAAS,CAAC,MAAM,CAAC,KAAK,CAAC,kBAAkB,CAAC,CAAC;IAC/D,CAAC;IAED,qEAAqE;IACrE,IAAI,WAAW;QACb,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;IAC1B,CAAC;IAED,KAAK,CAAC,KAAK;QACT,MAAM,GAAG,GAAG,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC;QAC7B,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,gBAAgB,EAAE,CAAC;QAEzC,6EAA6E;QAC7E,uEAAuE;QACvE,yEAAyE;QACzE,6EAA6E;QAC7E,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC;YACpB,MAAM,UAAU,GAAG,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC;YACxD,GAAG,CAAC,IAAI,CACN,sDAAsD;gBACpD,CAAC,UAAU;oBACT,CAAC,CAAC,0BAA0B,UAAU,8BAA8B;oBACpE,CAAC,CAAC,oCAAoC,CAAC;gBACzC,qFAAqF,CACxF,CAAC;QACJ,CAAC;QAED,MAAM,MAAM,GAAG,IAAI,UAAU,CAAC;YAC5B,cAAc,EAAE,GAAG,EAAE,CACnB,IAAI,gBAAgB,CAAC;gBACnB,MAAM,EAAE,IAAI,CAAC,OAAO;gBACpB,QAAQ,EAAE,IAAI,CAAC,SAAS;gBACxB,GAAG,EAAE,IAAI,CAAC,IAAI,CAAC,MAAmC;gBAClD,QAAQ,EAAE,IAAI,CAAC,MAAM;gBACrB,MAAM,EAAE,GAAG;aACZ,CAAC;YACJ,MAAM,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,MAAM;YACjC,IAAI,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI;YAC7B,MAAM,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI;YAC/B,cAAc,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,kBAAkB;YACrD,GAAG,CAAC,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,WAAW;gBAChC,CAAC,CAAC,EAAE,IAAI,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,WAAW,EAAE;gBAC1C,CAAC,CAAC,EAAE,CAAC;SACR,CAAC,CAAC;QAEH,MAAM,MAAM,CAAC,KAAK,EAAE,CAAC;QACrB,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC;QACtB,GAAG,CAAC,IAAI,CACN,iCAAiC,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,WAAW,IAAI,SAAS,IAAI,MAAM,CAAC,IAAI,GAAG,MAAM,CAAC,MAAM,EAAE;YAC3G,SAAS,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,kBAAkB,cAAc,CAC/D,CAAC;IACJ,CAAC;IAED;;;;;;OAMG;IACH,KAAK,CAAC,IAAI;QACR,IAAI,CAAC,MAAM,CAAC,QAAQ,EAAE,CAAC;QACvB,MAAM,MAAM,GAAG,IAAI,CAAC,OAAO,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,SAAS,CAAC;QACzB,IAAI,MAAM;YAAE,MAAM,MAAM,CAAC,MAAM,EAAE,CAAC;QAClC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,0BAA0B,CAAC,CAAC;IACpD,CAAC;IAED;;;;;;;OAOG;IACH,gBAAgB;QACd,IAAI,CAAC;YACH,MAAM,QAAQ,GAAG,sCAAsC,CAAC;gBACtD,oBAAoB,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,QAAQ;gBAC1D,eAAe,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,QAAQ,CAAC,SAAS;gBACtD,GAAG,EAAE,IAAI,CAAC,IAAI,CAAC,MAAmC;aACnD,CAA0C,CAAC;YAC5C,IAAI,CAAC,QAAQ,EAAE,QAAQ,IAAI,CAAC,QAAQ,CAAC,cAAc;gBAAE,OAAO,SAAS,CAAC;YACtE,OAAO;gBACL,QAAQ,EAAE,QAAQ,CAAC,QAAQ;gBAC3B,cAAc,EAAE,QAAQ,CAAC,cAAc;aACxC,CAAC;QACJ,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CACnB,0DAA0D,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAC7G,CAAC;YACF,OAAO,SAAS,CAAC;QACnB,CAAC;IACH,CAAC;CACF"}

package/dist/policy.d.ts

@@ -0,0 +1,38 @@
+/**
+ * Who is allowed to call the agent.
+ *
+ * StandIn authenticates the SOCKET, not the human: the HMAC handshake proves the
+ * call came from StandIn, and `session.start` then names a caller. Deciding
+ * whether that caller may be answered is the deployment's business, so it lives
+ * here rather than in the SDK.
+ *
+ * Framework-neutral on purpose: nothing here imports OpenClaw, so the policy can
+ * be read, tested and reused without a gateway in the process.
+ */
+/** Normalize a phone number to digits only. */
+export declare function normalizePhoneNumber(input?: string): string;
+/**
+ * True when the caller matches an allowlist entry - by phone number (digits
+ * only) OR by exact caller id, case-insensitive.
+ *
+ * The id match is what lets a Microsoft Teams caller be allowlisted at all: their id is an
+ * AAD object id, not a phone number, so phone normalization would reduce it to
+ * the empty string and match nothing.
+ */
+export declare function isAllowlistedCaller(from: string | undefined, allowFrom: string[] | undefined): boolean;
+/**
+ * Inbound-policy decision. `from` is the caller's AAD object id for a Microsoft Teams call.
+ *
+ * An unset or unknown policy REFUSES. Defaulting the other way would mean a
+ * config typo silently opens the agent to anyone who can reach the number.
+ */
+export declare function isInboundCallAllowed(inboundPolicy: "disabled" | "allowlist" | "pairing" | "open" | undefined, allowFrom: string[] | undefined, from: string | undefined): boolean;
+/**
+ * An actionable log line for a caller the policy just refused.
+ *
+ * "pairing" is enforced as a plain allowlist here: no pairing codes, expirations
+ * or approval prompts are issued for calls, so the operator's fix is the same as
+ * under "allowlist".
+ */
+export declare function describeInboundRejection(inboundPolicy: "disabled" | "allowlist" | "pairing" | "open" | undefined, from: string | undefined): string;
+//# sourceMappingURL=policy.d.ts.map

package/dist/policy.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"policy.d.ts","sourceRoot":"","sources":["../src/policy.ts"],"names":[],"mappings":"AAGA;;;;;;;;;;GAUG;AAEH,+CAA+C;AAC/C,wBAAgB,oBAAoB,CAAC,KAAK,CAAC,EAAE,MAAM,GAAG,MAAM,CAK3D;AAED;;;;;;;GAOG;AACH,wBAAgB,mBAAmB,CACjC,IAAI,EAAE,MAAM,GAAG,SAAS,EACxB,SAAS,EAAE,MAAM,EAAE,GAAG,SAAS,GAC9B,OAAO,CAwBT;AAED;;;;;GAKG;AACH,wBAAgB,oBAAoB,CAClC,aAAa,EAAE,UAAU,GAAG,WAAW,GAAG,SAAS,GAAG,MAAM,GAAG,SAAS,EACxE,SAAS,EAAE,MAAM,EAAE,GAAG,SAAS,EAC/B,IAAI,EAAE,MAAM,GAAG,SAAS,GACvB,OAAO,CAUT;AAED;;;;;;GAMG;AACH,wBAAgB,wBAAwB,CACtC,aAAa,EAAE,UAAU,GAAG,WAAW,GAAG,SAAS,GAAG,MAAM,GAAG,SAAS,EACxE,IAAI,EAAE,MAAM,GAAG,SAAS,GACvB,MAAM,CAqBR"}

package/dist/policy.js

@@ -0,0 +1,95 @@
+// Copyright (c) 2026 Komaa DigiTech
+// SPDX-License-Identifier: MIT
+/**
+ * Who is allowed to call the agent.
+ *
+ * StandIn authenticates the SOCKET, not the human: the HMAC handshake proves the
+ * call came from StandIn, and `session.start` then names a caller. Deciding
+ * whether that caller may be answered is the deployment's business, so it lives
+ * here rather than in the SDK.
+ *
+ * Framework-neutral on purpose: nothing here imports OpenClaw, so the policy can
+ * be read, tested and reused without a gateway in the process.
+ */
+/** Normalize a phone number to digits only. */
+export function normalizePhoneNumber(input) {
+    if (!input) {
+        return "";
+    }
+    return input.replace(/\D/g, "");
+}
+/**
+ * True when the caller matches an allowlist entry - by phone number (digits
+ * only) OR by exact caller id, case-insensitive.
+ *
+ * The id match is what lets a Microsoft Teams caller be allowlisted at all: their id is an
+ * AAD object id, not a phone number, so phone normalization would reduce it to
+ * the empty string and match nothing.
+ */
+export function isAllowlistedCaller(from, allowFrom) {
+    const raw = from?.trim();
+    if (!raw) {
+        return false;
+    }
+    const idFrom = raw.toLowerCase();
+    const normalizedFrom = normalizePhoneNumber(raw);
+    return (allowFrom ?? []).some((entry) => {
+        const trimmed = entry.trim();
+        if (!trimmed) {
+            return false;
+        }
+        // Exact caller-id match (e.g. a Microsoft Teams AAD object id), case-insensitive.
+        if (trimmed.toLowerCase() === idFrom) {
+            return true;
+        }
+        // Phone-number match (digits only).
+        const normalizedAllow = normalizePhoneNumber(trimmed);
+        return (normalizedAllow !== "" &&
+            normalizedFrom !== "" &&
+            normalizedAllow === normalizedFrom);
+    });
+}
+/**
+ * Inbound-policy decision. `from` is the caller's AAD object id for a Microsoft Teams call.
+ *
+ * An unset or unknown policy REFUSES. Defaulting the other way would mean a
+ * config typo silently opens the agent to anyone who can reach the number.
+ */
+export function isInboundCallAllowed(inboundPolicy, allowFrom, from) {
+    switch (inboundPolicy) {
+        case "open":
+            return true;
+        case "allowlist":
+        case "pairing":
+            return isAllowlistedCaller(from, allowFrom);
+        default:
+            return false;
+    }
+}
+/**
+ * An actionable log line for a caller the policy just refused.
+ *
+ * "pairing" is enforced as a plain allowlist here: no pairing codes, expirations
+ * or approval prompts are issued for calls, so the operator's fix is the same as
+ * under "allowlist".
+ */
+export function describeInboundRejection(inboundPolicy, from) {
+    const policy = inboundPolicy ?? "disabled";
+    const caller = from?.trim()
+        ? `caller "${from.trim()}"`
+        : "caller with no caller id";
+    if (policy === "disabled" || policy === "open") {
+        // "open" never rejects; kept for exhaustiveness if callers reuse this.
+        return (`inbound call rejected by policy "${policy}": inbound calling is ` +
+            (policy === "disabled"
+                ? 'disabled - set inboundPolicy to "allowlist" and add callers to allowFrom to accept calls'
+                : "open") +
+            ` (${caller})`);
+    }
+    const pairingHint = policy === "pairing"
+        ? ' Note: "pairing" currently enforces a plain allowlist (this plugin issues no pairing codes' +
+            " or approvals for calls) - add the caller's AAD object id to allowFrom."
+        : " Add the caller's AAD object id (or phone number) to allowFrom to accept them.";
+    return `inbound call rejected by policy "${policy}": ${caller} is not in allowFrom.${pairingHint}`;
+}
+//# sourceMappingURL=policy.js.map

package/dist/policy.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"policy.js","sourceRoot":"","sources":["../src/policy.ts"],"names":[],"mappings":"AAAA,oCAAoC;AACpC,+BAA+B;AAE/B;;;;;;;;;;GAUG;AAEH,+CAA+C;AAC/C,MAAM,UAAU,oBAAoB,CAAC,KAAc;IACjD,IAAI,CAAC,KAAK,EAAE,CAAC;QACX,OAAO,EAAE,CAAC;IACZ,CAAC;IACD,OAAO,KAAK,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;AAClC,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,mBAAmB,CACjC,IAAwB,EACxB,SAA+B;IAE/B,MAAM,GAAG,GAAG,IAAI,EAAE,IAAI,EAAE,CAAC;IACzB,IAAI,CAAC,GAAG,EAAE,CAAC;QACT,OAAO,KAAK,CAAC;IACf,CAAC;IACD,MAAM,MAAM,GAAG,GAAG,CAAC,WAAW,EAAE,CAAC;IACjC,MAAM,cAAc,GAAG,oBAAoB,CAAC,GAAG,CAAC,CAAC;IACjD,OAAO,CAAC,SAAS,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE;QACtC,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;QAC7B,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,OAAO,KAAK,CAAC;QACf,CAAC;QACD,kFAAkF;QAClF,IAAI,OAAO,CAAC,WAAW,EAAE,KAAK,MAAM,EAAE,CAAC;YACrC,OAAO,IAAI,CAAC;QACd,CAAC;QACD,oCAAoC;QACpC,MAAM,eAAe,GAAG,oBAAoB,CAAC,OAAO,CAAC,CAAC;QACtD,OAAO,CACL,eAAe,KAAK,EAAE;YACtB,cAAc,KAAK,EAAE;YACrB,eAAe,KAAK,cAAc,CACnC,CAAC;IACJ,CAAC,CAAC,CAAC;AACL,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,oBAAoB,CAClC,aAAwE,EACxE,SAA+B,EAC/B,IAAwB;IAExB,QAAQ,aAAa,EAAE,CAAC;QACtB,KAAK,MAAM;YACT,OAAO,IAAI,CAAC;QACd,KAAK,WAAW,CAAC;QACjB,KAAK,SAAS;YACZ,OAAO,mBAAmB,CAAC,IAAI,EAAE,SAAS,CAAC,CAAC;QAC9C;YACE,OAAO,KAAK,CAAC;IACjB,CAAC;AACH,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,wBAAwB,CACtC,aAAwE,EACxE,IAAwB;IAExB,MAAM,MAAM,GAAG,aAAa,IAAI,UAAU,CAAC;IAC3C,MAAM,MAAM,GAAG,IAAI,EAAE,IAAI,EAAE;QACzB,CAAC,CAAC,WAAW,IAAI,CAAC,IAAI,EAAE,GAAG;QAC3B,CAAC,CAAC,0BAA0B,CAAC;IAC/B,IAAI,MAAM,KAAK,UAAU,IAAI,MAAM,KAAK,MAAM,EAAE,CAAC;QAC/C,uEAAuE;QACvE,OAAO,CACL,oCAAoC,MAAM,wBAAwB;YAClE,CAAC,MAAM,KAAK,UAAU;gBACpB,CAAC,CAAC,0FAA0F;gBAC5F,CAAC,CAAC,MAAM,CAAC;YACX,KAAK,MAAM,GAAG,CACf,CAAC;IACJ,CAAC;IACD,MAAM,WAAW,GACf,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,4FAA4F;YAC5F,yEAAyE;QAC3E,CAAC,CAAC,gFAAgF,CAAC;IACvF,OAAO,oCAAoC,MAAM,MAAM,MAAM,wBAAwB,WAAW,EAAE,CAAC;AACrG,CAAC"}

package/dist/protocol.d.ts

@@ -0,0 +1,44 @@
+export { contextSentences, decodePcm, parseMessage } from "./protocolRuntime.js";
+export declare const SAMPLE_RATE_HZ = 16000;
+export declare const NUM_CHANNELS = 1;
+export declare const TYPE_SESSION_START = "session.start";
+export declare const TYPE_SESSION_END = "session.end";
+export declare const TYPE_RECORDING_STATUS = "recording.status";
+export declare const TYPE_AUDIO_FRAME = "audio.frame";
+export declare const TYPE_VIDEO_FRAME = "video.frame";
+export declare const TYPE_PARTICIPANTS = "participants";
+export declare const TYPE_DTMF = "dtmf";
+export declare const TYPE_PING = "ping";
+export declare const TYPE_ASSISTANT_SAY = "assistant.say";
+export declare const TYPE_ASSISTANT_CANCEL = "assistant.cancel";
+export declare const TYPE_EXPRESSION = "expression";
+export declare const TYPE_SPEECH_MARKS = "speech.marks";
+export declare const TYPE_DISPLAY_IMAGE = "display.image";
+export declare const TYPE_DISPLAY_FRAME = "display.frame";
+export declare const TYPE_PONG = "pong";
+/** Caller identity; blank or absent values normalize to undefined. */
+export interface Caller {
+    readonly aadId?: string;
+    readonly displayName?: string;
+    readonly tenantId?: string;
+}
+/** Call context with the SDK's compatible defaults. */
+export interface SessionStart {
+    readonly callId: string;
+    readonly threadId: string;
+    readonly caller: Caller;
+    readonly direction: "inbound" | "outbound";
+    readonly recordingStatus?: string;
+    readonly tenantId?: string;
+}
+/** Read call context; only callId lacks a safe default. */
+export declare function parseSessionStart(msg: Record<string, unknown>): SessionStart;
+/** Build an outbound `audio.frame` JSON frame. */
+export declare function audioFrame(seq: number, timestampMs: number, pcm: Buffer): string;
+/** Build an outbound `pong` JSON frame. */
+export declare function pong(ts: unknown): string;
+/** Build an outbound `assistant.cancel` JSON frame. */
+export declare function assistantCancel(turnId: number): string;
+/** Build an outbound `session.end` JSON frame. */
+export declare function sessionEnd(reason: string): string;
+//# sourceMappingURL=protocol.d.ts.map

package/dist/protocol.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"protocol.d.ts","sourceRoot":"","sources":["../src/protocol.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,gBAAgB,EAAE,SAAS,EAAE,YAAY,EAAE,MAAM,sBAAsB,CAAC;AAEjF,eAAO,MAAM,cAAc,QAAS,CAAC;AACrC,eAAO,MAAM,YAAY,IAAI,CAAC;AAE9B,eAAO,MAAM,kBAAkB,kBAAkB,CAAC;AAClD,eAAO,MAAM,gBAAgB,gBAAgB,CAAC;AAC9C,eAAO,MAAM,qBAAqB,qBAAqB,CAAC;AACxD,eAAO,MAAM,gBAAgB,gBAAgB,CAAC;AAC9C,eAAO,MAAM,gBAAgB,gBAAgB,CAAC;AAC9C,eAAO,MAAM,iBAAiB,iBAAiB,CAAC;AAChD,eAAO,MAAM,SAAS,SAAS,CAAC;AAChC,eAAO,MAAM,SAAS,SAAS,CAAC;AAChC,eAAO,MAAM,kBAAkB,kBAAkB,CAAC;AAClD,eAAO,MAAM,qBAAqB,qBAAqB,CAAC;AACxD,eAAO,MAAM,eAAe,eAAe,CAAC;AAC5C,eAAO,MAAM,iBAAiB,iBAAiB,CAAC;AAChD,eAAO,MAAM,kBAAkB,kBAAkB,CAAC;AAClD,eAAO,MAAM,kBAAkB,kBAAkB,CAAC;AAClD,eAAO,MAAM,SAAS,SAAS,CAAC;AAEhC,sEAAsE;AACtE,MAAM,WAAW,MAAM;IACrB,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,WAAW,CAAC,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,CAAC;CAC5B;AAED,uDAAuD;AACvD,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,SAAS,EAAE,SAAS,GAAG,UAAU,CAAC;IAC3C,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC;IAClC,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,CAAC;CAC5B;AAED,2DAA2D;AAC3D,wBAAgB,iBAAiB,CAAC,GAAG,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,YAAY,CAoB5E;AAED,kDAAkD;AAClD,wBAAgB,UAAU,CAAC,GAAG,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,EAAE,GAAG,EAAE,MAAM,GAAG,MAAM,CAOhF;AAED,2CAA2C;AAC3C,wBAAgB,IAAI,CAAC,EAAE,EAAE,OAAO,GAAG,MAAM,CAKxC;AAED,uDAAuD;AACvD,wBAAgB,eAAe,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAKtD;AAED,kDAAkD;AAClD,wBAAgB,UAAU,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAKjD"}