@alexkroman1/aai-ui 2.0.0 → 3.1.0

package/dist/_module-url-C_4gRVL0.js

@@ -0,0 +1,15 @@
+//#region worklets/_module-url.ts
+/**
+* Blob-URL module for an inlined AudioWorklet processor source.
+*
+* A blob URL rather than a data URI because the agent page's CSP allows
+* `script-src blob:` but not `data:` — a data-URI module fails `addModule`
+* with the opaque "Unable to load a worklet's module". Every inline worklet
+* in this package must load through this helper so that constraint lives in
+* one place.
+*/
+function workletModuleUrl(source) {
+	return URL.createObjectURL(new Blob([source], { type: "application/javascript" }));
+}
+//#endregion
+export { workletModuleUrl as t };

package/dist/audio.d.ts

@@ -1,12 +1,12 @@
 /**
- * Decode an audio file (any container/codec the browser can decode) and
- * resample it to mono PCM16 at `targetRate` — the format the server's STT
- * side expects on the wire. Returns the raw clip; any endpointing padding
- * is the caller's concern (the one-shot upload path needs none).
+ * Clamp-and-convert Float32 samples to PCM16. The one main-thread home of the
+ * asymmetric-rounding convention (negative × 0x8000, positive × 0x7fff) —
+ * the capture worklet's `accumulate` embeds the same math, which cannot
+ * import it (worklet source is a string).
  *
- * @throws If the browser cannot decode the payload.
+ * @public
  */
-export declare function decodeAudioToPcm16(data: ArrayBuffer, targetRate: number): Promise<Int16Array>;
+export declare function floatToPcm16(samples: Float32Array): Int16Array;
 /**
  * How much of one turn's playback was covered by concealment rather than
  * received audio — the playback worklet's underrun report, in the shape
@@ -78,6 +78,42 @@ export type VoiceIO = AsyncDisposable & {
     /** Release all audio resources (microphone, AudioContext, worklets). */
     close(): Promise<void>;
 };
+/**
+ * Throw unless the browser honored a requested context sample rate. The
+ * requested rates are never advisory: captured audio is tagged with the
+ * requested rate on the wire, so a context running at some other rate ships
+ * audio that only sounds like speech to the wrong decoder. Every capture
+ * path must call this after context creation.
+ */
+export declare function assertGranted(granted: number, requested: number, side: string): void;
+/**
+ * Release a microphone that was (or later gets) granted while another init
+ * step failed; if `getUserMedia` itself rejected, this is a no-op. Without
+ * it, a mic granted after a failed init keeps the browser's recording
+ * indicator lit with no way to turn it off.
+ */
+export declare function releaseStreamOnFailure(streamPromise: Promise<MediaStream>): void;
+/** Handle to one capture worklet node (`worklets/capture-processor.ts`). */
+export type CaptureNode = {
+    node: AudioWorkletNode;
+    /** Begin accumulating — the worklet gates capture on its start/stop protocol. */
+    start(): void;
+    /**
+     * Stop accumulating and wait (bounded by {@link CAPTURE_STOP_ACK_TIMEOUT_MS})
+     * for the 'stopped' ack that follows the worklet's final flush, so the tail
+     * of speech reaches `onChunk` before the node is torn down.
+     */
+    stop(): Promise<void>;
+};
+/**
+ * Wire one capture worklet node: node construction, the chunk/silent/stopped
+ * port protocol, and the stop→ack handshake — one spelling for both the
+ * WebSocket mic and the push-to-talk recorder. No `processorOptions`: the
+ * worklet reads the context rate from its global scope (callers assert the
+ * granted rate first) and owns its own batching default — re-spelling
+ * defaults caller-side is drift.
+ */
+export declare function createCaptureNode(ctx: AudioContext, onChunk: (pcm16: ArrayBuffer) => void, onSilent?: () => void): CaptureNode;
 /**
  * Create a {@link VoiceIO} instance that captures microphone audio and
  * plays back TTS audio using the Web Audio API.

package/dist/audio.js

@@ -1,50 +1,85 @@
-import { MIC_BUFFER_SECONDS, VOICE_CAPTURE_CONSTRAINTS } from "./types.js";
+import { CAPTURE_STOP_ACK_TIMEOUT_MS, PLAYBACK_DONE_MAX_WAIT_MS, PLAYBACK_DONE_POLL_MS, VOICE_CAPTURE_CONSTRAINTS } from "./types.js";
 //#region audio.ts
-/** How often {@link VoiceIO.done} checks that the AudioContext is still rendering. */
-const DONE_POLL_INTERVAL_MS = 1e3;
 /**
-* Hard cap on waiting for playback to drain. The playback worklet buffers up
-* to 60s of audio, so the longest legitimate drain is just under that — a
-* wait past this means the processor died without reporting 'stop'.
-*/
-const DONE_MAX_WAIT_MS = 65e3;
-/**
-* Bounded wait for the capture worklet's 'stopped' ack during close(). The
-* ack follows the final flush, so waiting for it keeps the tail of speech
-* from being dropped; the timeout covers a dead worklet.
-*/
-const CAPTURE_STOP_ACK_TIMEOUT_MS = 250;
-/**
-* Decode an audio file (any container/codec the browser can decode) and
-* resample it to mono PCM16 at `targetRate` — the format the server's STT
-* side expects on the wire. Returns the raw clip; any endpointing padding
-* is the caller's concern (the one-shot upload path needs none).
+* Clamp-and-convert Float32 samples to PCM16. The one main-thread home of the
+* asymmetric-rounding convention (negative × 0x8000, positive × 0x7fff) —
+* the capture worklet's `accumulate` embeds the same math, which cannot
+* import it (worklet source is a string).
 *
-* @throws If the browser cannot decode the payload.
+* @public
 */
-async function decodeAudioToPcm16(data, targetRate) {
-	const decoded = await new OfflineAudioContext(1, 1, targetRate).decodeAudioData(data);
-	const frames = Math.ceil(decoded.duration * targetRate);
-	const offline = new OfflineAudioContext(1, frames, targetRate);
-	const source = offline.createBufferSource();
-	source.buffer = decoded;
-	source.connect(offline.destination);
-	source.start();
-	const f32 = (await offline.startRendering()).getChannelData(0);
-	const pcm = new Int16Array(f32.length);
+function floatToPcm16(samples) {
+	const pcm = new Int16Array(samples.length);
 	let i = 0;
-	for (const sample of f32) {
+	for (const sample of samples) {
 		const s = Math.max(-1, Math.min(1, sample));
 		pcm[i++] = s < 0 ? s * 32768 : s * 32767;
 	}
 	return pcm;
 }
-/** Throw unless the browser honored a requested context sample rate. */
+/**
+* Throw unless the browser honored a requested context sample rate. The
+* requested rates are never advisory: captured audio is tagged with the
+* requested rate on the wire, so a context running at some other rate ships
+* audio that only sounds like speech to the wrong decoder. Every capture
+* path must call this after context creation.
+*/
 function assertGranted(granted, requested, side) {
 	if (granted === requested) return;
 	throw new Error(`Browser refused the ${side} sample rate: asked for ${requested} Hz, got ${granted} Hz`);
 }
 /**
+* Release a microphone that was (or later gets) granted while another init
+* step failed; if `getUserMedia` itself rejected, this is a no-op. Without
+* it, a mic granted after a failed init keeps the browser's recording
+* indicator lit with no way to turn it off.
+*/
+function releaseStreamOnFailure(streamPromise) {
+	streamPromise.then((s) => {
+		for (const t of s.getTracks()) t.stop();
+	}).catch(() => {});
+}
+/**
+* Wire one capture worklet node: node construction, the chunk/silent/stopped
+* port protocol, and the stop→ack handshake — one spelling for both the
+* WebSocket mic and the push-to-talk recorder. No `processorOptions`: the
+* worklet reads the context rate from its global scope (callers assert the
+* granted rate first) and owns its own batching default — re-spelling
+* defaults caller-side is drift.
+*/
+function createCaptureNode(ctx, onChunk, onSilent) {
+	const node = new AudioWorkletNode(ctx, "capture-processor", {
+		channelCount: 1,
+		channelCountMode: "explicit"
+	});
+	let onStopped = null;
+	node.port.onmessage = (e) => {
+		const d = e.data;
+		if (d.event === "chunk" && d.buffer) onChunk(d.buffer);
+		else if (d.event === "silent") onSilent?.();
+		else if (d.event === "stopped") {
+			onStopped?.();
+			onStopped = null;
+		}
+	};
+	return {
+		node,
+		start() {
+			node.port.postMessage({ event: "start" });
+		},
+		stop() {
+			return new Promise((resolve) => {
+				const cap = setTimeout(resolve, CAPTURE_STOP_ACK_TIMEOUT_MS);
+				onStopped = () => {
+					clearTimeout(cap);
+					resolve();
+				};
+				node.port.postMessage({ event: "stop" });
+			});
+		}
+	};
+}
+/**
 * Create a {@link VoiceIO} instance that captures microphone audio and
 * plays back TTS audio using the Web Audio API.
 *
@@ -92,43 +127,25 @@ async function createVoiceIO(opts) {
 		assertGranted(capCtx.sampleRate, sttSampleRate, "capture");
 		assertGranted(ctx.sampleRate, ttsSampleRate, "playback");
 	} catch (err) {
-		streamPromise.then((s) => {
-			for (const t of s.getTracks()) t.stop();
-		}).catch(() => {});
+		releaseStreamOnFailure(streamPromise);
 		await closeContexts();
 		throw err;
 	}
 	const mic = capCtx.createMediaStreamSource(stream);
-	const capNode = new AudioWorkletNode(capCtx, "capture-processor", {
-		channelCount: 1,
-		channelCountMode: "explicit",
-		processorOptions: {
-			sampleRate: sttSampleRate,
-			bufferSeconds: MIC_BUFFER_SECONDS
-		}
-	});
-	mic.connect(capNode);
-	capNode.onprocessorerror = () => {
+	const capture = createCaptureNode(capCtx, onMicData, onMicSilent);
+	mic.connect(capture.node);
+	capture.node.onprocessorerror = () => {
 		const err = /* @__PURE__ */ new Error("Audio capture worklet crashed");
 		console.error("[aai-ui]", err.message);
 		onError?.(err);
 	};
-	capNode.port.postMessage({ event: "start" });
-	let onCaptureStopped = null;
-	capNode.port.onmessage = (e) => {
-		if (e.data.event === "chunk") onMicData(e.data.buffer);
-		else if (e.data.event === "silent") onMicSilent?.();
-		else if (e.data.event === "stopped") {
-			onCaptureStopped?.();
-			onCaptureStopped = null;
-		}
-	};
+	capture.start();
 	let playNode = null;
 	let onPlaybackStop = null;
 	const lifecycle = new AbortController();
 	function ensurePlayNode() {
 		if (playNode) return playNode;
-		const node = new AudioWorkletNode(ctx, "playback-processor", { processorOptions: { sampleRate: ctx.sampleRate } });
+		const node = new AudioWorkletNode(ctx, "playback-processor");
 		node.connect(ctx.destination);
 		node.port.onmessage = (e) => {
 			if (e.data.event === "stop") {
@@ -160,6 +177,7 @@ async function createVoiceIO(opts) {
 		},
 		done() {
 			if (!playNode) return Promise.resolve();
+			playNode.port.postMessage({ event: "done" });
 			if (ctx.state !== "running") return Promise.resolve();
 			return new Promise((resolve) => {
 				onPlaybackStop?.();
@@ -171,10 +189,9 @@ async function createVoiceIO(opts) {
 				};
 				const poll = setInterval(() => {
 					if (ctx.state !== "running") settle();
-				}, DONE_POLL_INTERVAL_MS);
-				const cap = setTimeout(settle, DONE_MAX_WAIT_MS);
+				}, PLAYBACK_DONE_POLL_MS);
+				const cap = setTimeout(settle, PLAYBACK_DONE_MAX_WAIT_MS);
 				onPlaybackStop = settle;
-				playNode?.port.postMessage({ event: "done" });
 			});
 		},
 		flush() {
@@ -186,16 +203,9 @@ async function createVoiceIO(opts) {
 		async close() {
 			if (lifecycle.signal.aborted) return;
 			lifecycle.abort();
-			await new Promise((resolve) => {
-				const cap = setTimeout(resolve, CAPTURE_STOP_ACK_TIMEOUT_MS);
-				onCaptureStopped = () => {
-					clearTimeout(cap);
-					resolve();
-				};
-				capNode.port.postMessage({ event: "stop" });
-			});
+			await capture.stop();
 			mic.disconnect();
-			capNode.disconnect();
+			capture.node.disconnect();
 			if (playNode) playNode.disconnect();
 			for (const t of stream.getTracks()) t.stop();
 			await closeContexts();
@@ -207,4 +217,4 @@ async function createVoiceIO(opts) {
 	return io;
 }
 //#endregion
-export { createVoiceIO, decodeAudioToPcm16 };
+export { assertGranted, createCaptureNode, createVoiceIO, floatToPcm16, releaseStreamOnFailure };

package/dist/{chat-view-DKFhxMAT.js → chat-view-C1oJxsWz.js}

@@ -2,7 +2,7 @@ import { useSessionSelector, useTheme } from "./context.js";
 import { r as TEXT_FAINT, t as ERROR_COLOR } from "./_colors-DYX7XRTr.js";
 import { t as AaiLogo } from "./aai-logo-B8lDmsut.js";
 import { t as Eyebrow } from "./eyebrow-C6ZFuiz6.js";
-import { t as Controls } from "./controls-4OJoekj6.js";
+import { t as Controls } from "./controls-DV368uhb.js";
 import { MessageList } from "./components/message-list.js";
 import clsx from "clsx";
 import { jsx, jsxs } from "react/jsx-runtime";
@@ -24,8 +24,8 @@ function stateColor(state, primary) {
 	}
 }
 /**
-* The design-system "console" chrome shared by the chat shell and the
-* workflow run surface: a 760px column on the cream page with a header
+* The design-system "console" chrome for the chat shell:
+* a 760px column on the cream page with a header
 * (logo + live-status eyebrow), an optional error banner, the main content
 * on a raised white card, and a footer row beneath it.
 *
@@ -126,4 +126,4 @@ function ChatView({ icon, title, className }) {
 	});
 }
 //#endregion
-export { ConsoleShell as n, ChatView as t };
+export { ChatView as t };

package/dist/client-config.d.ts

@@ -2,12 +2,11 @@
  * Pre-connection client-config lookup.
  *
  * `GET client-config` (relative to the agent's base URL — see
- * `sdk/client-config.ts` in `@alexkroman1/aai`) tells the default client
- * what kind of app the agent is before any connection exists — a
- * conversational agent (WebSocket chat shell) or a workflow (one-shot run
- * surface). Every failure path — network error, 404 from an older server,
- * malformed body — degrades to the agent default, so this lookup can never
- * break an existing agent.
+ * `sdk/client-config.ts` in `@alexkroman1/aai`) gives the default client the
+ * agent's display name and greeting before any connection exists. Every
+ * failure path — network error, 404 from an older server, malformed body —
+ * degrades to the empty default, so this lookup can never break an existing
+ * agent.
  */
 import { type ClientConfigResponse } from "@alexkroman1/aai/protocol";
 export type { ClientConfigResponse } from "@alexkroman1/aai/protocol";

package/dist/components/chat-view.js

@@ -1,3 +1,3 @@
 import "../context.js";
-import { t as ChatView } from "../chat-view-DKFhxMAT.js";
+import { t as ChatView } from "../chat-view-C1oJxsWz.js";
 export { ChatView };

package/dist/components/console-shell.d.ts

@@ -7,8 +7,8 @@ import type { AgentState } from "../types.ts";
  */
 export declare function stateColor(state: AgentState, primary: string): string;
 /**
- * The design-system "console" chrome shared by the chat shell and the
- * workflow run surface: a 760px column on the cream page with a header
+ * The design-system "console" chrome for the chat shell:
+ * a 760px column on the cream page with a header
  * (logo + live-status eyebrow), an optional error banner, the main content
  * on a raised white card, and a footer row beneath it.
  *

package/dist/components/controls.js

@@ -1,3 +1,3 @@
 import "../context.js";
-import { t as Controls } from "../controls-4OJoekj6.js";
+import { t as Controls } from "../controls-DV368uhb.js";
 export { Controls };

package/dist/components/url-chips.d.ts

@@ -4,9 +4,6 @@
  * The label is what makes a pair of these readable — on its own a bare URL
  * leaves you guessing whether it's the page or the socket.
  *
- * Exported for the sync-transport shell, which labels its HTTP endpoint
- * without a session snapshot to read from.
- *
  * @internal
  */
 export declare function UrlChip({ label, url, hint, testId, className, }: {

package/dist/context.d.ts

@@ -7,7 +7,7 @@ export declare function SessionProvider({ value, children }: {
 }): import("react").FunctionComponentElement<import("react").ProviderProps<SessionCore | null>>;
 /** The session snapshot merged with the core's control methods. Method
  *  signatures come from {@link SessionCore} — one source of truth. */
-export type Session = SessionSnapshot & Pick<SessionCore, "start" | "cancel" | "resetState" | "reset" | "disconnect" | "toggle" | "startRecording" | "stopRecording" | "sendAudioFile">;
+export type Session = SessionSnapshot & Pick<SessionCore, "start" | "cancel" | "resetState" | "reset" | "disconnect" | "toggle">;
 /**
  * Return the raw {@link SessionCore} from context without subscribing to
  * snapshot changes. Useful for accessing stable methods (`start`, `toggle`,

package/dist/context.js

@@ -36,10 +36,7 @@ function useSession() {
 		resetState: core.resetState,
 		reset: core.reset,
 		disconnect: core.disconnect,
-		toggle: core.toggle,
-		startRecording: core.startRecording,
-		stopRecording: core.stopRecording,
-		sendAudioFile: core.sendAudioFile
+		toggle: core.toggle
 	}), [snapshot, core]);
 }
 /**

package/dist/{controls-4OJoekj6.js → controls-DV368uhb.js}

@@ -14,9 +14,6 @@ const COPIED_FEEDBACK_MS = 1500;
 * The label is what makes a pair of these readable — on its own a bare URL
 * leaves you guessing whether it's the page or the socket.
 *
-* Exported for the sync-transport shell, which labels its HTTP endpoint
-* without a session snapshot to read from.
-*
 * @internal
 */
 function UrlChip({ label, url, hint, testId, className }) {
@@ -140,4 +137,4 @@ const Controls = memo(function Controls({ className }) {
 	});
 });
 //#endregion
-export { UrlChip as a, UiUrlChip as i, ApiUrlChip as n, SessionUrlChips as r, Controls as t };
+export { UiUrlChip as i, ApiUrlChip as n, SessionUrlChips as r, Controls as t };

package/dist/default-client/assets/_module-url-BX0RuRU2.js

@@ -0,0 +1 @@
+function e(e){return URL.createObjectURL(new Blob([e],{type:`application/javascript`}))}export{e as t};

package/dist/default-client/assets/audio-BGHiiDY_.js

@@ -0,0 +1 @@
+import{a as e,o as t,t as n}from"./index-DXf7-M0Q.js";function r(e){let t=new Int16Array(e.length),n=0;for(let r of e){let e=Math.max(-1,Math.min(1,r));t[n++]=e<0?e*32768:e*32767}return t}function i(e,t,n){if(e!==t)throw Error(`Browser refused the ${n} sample rate: asked for ${t} Hz, got ${e} Hz`)}function a(e){e.then(e=>{for(let t of e.getTracks())t.stop()}).catch(()=>{})}function o(e,t,n){let r=new AudioWorkletNode(e,`capture-processor`,{channelCount:1,channelCountMode:`explicit`}),i=null;return r.port.onmessage=e=>{let r=e.data;r.event===`chunk`&&r.buffer?t(r.buffer):r.event===`silent`?n?.():r.event===`stopped`&&(i?.(),i=null)},{node:r,start(){r.port.postMessage({event:`start`})},stop(){return new Promise(e=>{let t=setTimeout(e,250);i=()=>{clearTimeout(t),e()},r.port.postMessage({event:`stop`})})}}}async function s(r){let{sttSampleRate:s,ttsSampleRate:c,captureWorkletSrc:l,playbackWorkletSrc:u,onMicData:d,onError:f,onPlaybackStats:p,onMicSilent:m}=r,h=new AudioContext({sampleRate:c,latencyHint:`playback`}),g=s===c,_=g?h:new AudioContext({sampleRate:s,latencyHint:`interactive`});async function v(){let e=g?[h]:[h,_];await Promise.all(e.map(e=>e.close().catch(e=>{console.warn(`AudioContext close failed:`,e)})))}let y=navigator.mediaDevices.getUserMedia({audio:{deviceId:{ideal:`default`},...n}}),b;try{[b]=await Promise.all([y,h.resume(),_.resume(),_.audioWorklet.addModule(l),h.audioWorklet.addModule(u)]),i(_.sampleRate,s,`capture`),i(h.sampleRate,c,`playback`)}catch(e){throw a(y),await v(),e}let x=_.createMediaStreamSource(b),S=o(_,d,m);x.connect(S.node),S.node.onprocessorerror=()=>{let e=Error(`Audio capture worklet crashed`);console.error(`[aai-ui]`,e.message),f?.(e)},S.start();let C=null,w=null,T=new AbortController;function E(){if(C)return C;let e=new AudioWorkletNode(h,`playback-processor`);return e.connect(h.destination),e.port.onmessage=e=>{if(e.data.event===`stop`){let t=e.data.stats;if(t&&t.concealedSamples>0&&p?.(t),e.data.reason===`interrupt`)return;w?.(),w=null}},e.onprocessorerror=()=>{let e=Error(`Audio playback worklet crashed`);console.error(`[aai-ui]`,e.message),w?.(),w=null,f?.(e)},C=e,e}let D={enqueue(e){T.signal.aborted||e.byteLength!==0&&E().port.postMessage({event:`write`,buffer:new Uint8Array(e)},[e])},done(){return!C||(C.port.postMessage({event:`done`}),h.state!==`running`)?Promise.resolve():new Promise(n=>{w?.();let r=()=>{clearInterval(i),clearTimeout(a),w===r&&(w=null),n()},i=setInterval(()=>{h.state!==`running`&&r()},t),a=setTimeout(r,e);w=r})},flush(){C&&(w?.(),w=null,C.port.postMessage({event:`interrupt`}))},async close(){if(!T.signal.aborted){T.abort(),await S.stop(),x.disconnect(),S.node.disconnect(),C&&C.disconnect();for(let e of b.getTracks())e.stop();await v()}},async[Symbol.asyncDispose](){await D.close()}};return D}export{i as assertGranted,o as createCaptureNode,s as createVoiceIO,r as floatToPcm16,a as releaseStreamOnFailure};

package/dist/default-client/assets/{capture-processor-DLHxAIfT.js → capture-processor-BLPsqnGl.js}

@@ -1,4 +1,4 @@
-import{p as e}from"./types-Bpg3ZIZK.js";var t=`
+import{n as e,r as t}from"./index-DXf7-M0Q.js";import{t as n}from"./_module-url-BX0RuRU2.js";var r=n(`
 class CaptureProcessor extends AudioWorkletProcessor {
   constructor(options) {
     super();
@@ -10,14 +10,14 @@ class CaptureProcessor extends AudioWorkletProcessor {
     // Int16 accumulation buffer: flushed to the main thread as one transferred
     // ArrayBuffer once ~bufferSeconds of samples are batched. Sized 2x the
     // flush target so a whole render quantum always fits before flushing.
-    this.targetSamples = Math.max(1, Math.round(this.rate * (opts.bufferSeconds || 0.1)));
+    this.targetSamples = Math.max(1, Math.round(this.rate * (opts.bufferSeconds || ${e})));
     this.pending = new Int16Array(this.targetSamples * 2);
     this.pendingLen = 0;
     // Dead-mic probe: samples left to inspect before concluding the device
     // delivers nothing but digital silence. Only consumed while recording, so
     // the cost disappears after the window (or after the first real sample).
     this.probeSamplesLeft = Math.round(
-      (this.rate * (opts.silenceProbeMs ?? ${e})) / 1000,
+      (this.rate * (opts.silenceProbeMs ?? ${t})) / 1000,
     );
     this.port.onmessage = (e) => {
       if (e.data.event === 'start') this.recording = true;
@@ -87,4 +87,4 @@ class CaptureProcessor extends AudioWorkletProcessor {
 }
 
 registerProcessor('capture-processor', CaptureProcessor);
-`,n=new Blob([t],{type:`application/javascript`}),r=URL.createObjectURL(n);export{r as default};
+`);export{r as default};