@elizaos/ui 2.0.3-beta.2 → 2.0.3-beta.3

package/voice/voice-provider-defaults.js

@@ -5,15 +5,18 @@
  * settings "advanced mode" picker design:
  *
  *   - Desktop running a local agent → on-device models
- *     (TTS: `local-inference` / OmniVoice, ASR: `local-inference` / Qwen3-ASR).
+ *     (TTS: `local-inference` / OmniVoice, ASR: `local-inference` / Gemma ASR).
  *   - Mobile running a local agent → on-device Kokoro TTS
  *     (TTS: `local-inference`; Kokoro is ~82M params and runs comfortably on
  *     phones — see `selectVoiceBackend({ mobile: true })`). ASR still routes
  *     to Eliza Cloud (`eliza-cloud`) because on-device speech recognition is
  *     heavier than TTS.
- *   - Cloud agents (any device) → always Eliza Cloud.
- *   - Remote-controller surfaces (UI hitting a remote API base) → Eliza
- *     Cloud, same rationale as cloud agents.
+ *   - Cloud agents (any device) → fast free Microsoft Edge neural TTS
+ *     (`edge`) for speech, Eliza Cloud (`eliza-cloud`) for ASR. ElevenLabs is
+ *     no longer a default — it is slow and key-gated; users can still opt into
+ *     it from the advanced voice picker.
+ *   - Remote-controller surfaces (UI hitting a remote API base) → same as
+ *     cloud agents (`edge` TTS, Eliza Cloud ASR).
  *
  * The picker is intentionally a pure function so it can be unit-tested
  * exhaustively. The React hook wrapper lives in
@@ -26,14 +29,16 @@
  */
 export function pickDefaultVoiceProvider(input) {
     const { platform, runtimeMode } = input;
-    // Cloud / remote: cloud everything, regardless of which device drives
-    // the UI. The agent isn't on this machine; we always route audio to the
-    // server's cloud-backed pipelines.
+    // Cloud / remote: the agent isn't on this machine, so speech can't run
+    // on-device. Default to the fast, free Microsoft Edge neural voices
+    // (`edge`) rather than the slow, key-gated ElevenLabs path, and route ASR
+    // to Eliza Cloud. The user can still opt back into ElevenLabs in advanced
+    // settings.
     if (runtimeMode === "cloud" || runtimeMode === "remote") {
-        return { tts: "elevenlabs", asr: "eliza-cloud" };
+        return { tts: "edge", asr: "eliza-cloud" };
     }
     // Local / local-only: split by platform. Desktop has the CPU/GPU budget
-    // for OmniVoice + Qwen3-ASR. Mobile runs on-device Kokoro for TTS (small +
+    // for OmniVoice + Gemma ASR. Mobile runs on-device Kokoro for TTS (small +
     // fast) but offloads the heavier ASR pipeline to Eliza Cloud. A web shell
     // hosting a local agent can't run on-device audio, so it stays on Cloud.
     if (platform === "desktop") {
@@ -42,5 +47,7 @@ export function pickDefaultVoiceProvider(input) {
     if (platform === "mobile") {
         return { tts: "local-inference", asr: "eliza-cloud" };
     }
-    return { tts: "elevenlabs", asr: "eliza-cloud" };
+    // Web shell hosting a local agent: no on-device audio runtime, so use the
+    // fast free Edge neural voices for TTS and Eliza Cloud for ASR.
+    return { tts: "edge", asr: "eliza-cloud" };
 }

package/voice/voice-selftest/VoiceSelfTestShell.d.ts

@@ -12,6 +12,17 @@
  *   - a DOM-mirrored JSON report at [data-testid="voice-selftest-report"]
  * so an automated runner can scrape the verdict with no human in the loop.
  */
+import { type VoiceSelfTestMode, type VoiceSelfTestReport } from "./voice-selftest-harness.js";
+declare global {
+    interface Window {
+        /** Legacy vendor-prefixed AudioContext (Safari / older WebKit). */
+        webkitAudioContext?: typeof AudioContext;
+        /** e2e automation hook — runs the self-test and returns its report. */
+        __voiceSelfTest?: (opts?: {
+            mode?: VoiceSelfTestMode;
+        }) => Promise<VoiceSelfTestReport>;
+    }
+}
 export declare function VoiceSelfTestShell(): import("react/jsx-runtime").JSX.Element;
 export default VoiceSelfTestShell;
 //# sourceMappingURL=VoiceSelfTestShell.d.ts.map

package/voice/voice-selftest/VoiceSelfTestShell.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"VoiceSelfTestShell.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/VoiceSelfTestShell.tsx"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;GAaG;AAgDH,wBAAgB,kBAAkB,4CA8JjC;AAED,eAAe,kBAAkB,CAAC"}
+{"version":3,"file":"VoiceSelfTestShell.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/VoiceSelfTestShell.tsx"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;GAaG;AAUH,OAAO,EAEL,KAAK,iBAAiB,EAEtB,KAAK,mBAAmB,EACzB,MAAM,0BAA0B,CAAC;AAElC,OAAO,CAAC,MAAM,CAAC;IACb,UAAU,MAAM;QACd,mEAAmE;QACnE,kBAAkB,CAAC,EAAE,OAAO,YAAY,CAAC;QACzC,uEAAuE;QACvE,eAAe,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE;YACxB,IAAI,CAAC,EAAE,iBAAiB,CAAC;SAC1B,KAAK,OAAO,CAAC,mBAAmB,CAAC,CAAC;KACpC;CACF;AA6BD,wBAAgB,kBAAkB,4CAwJjC;AAED,eAAe,kBAAkB,CAAC"}

package/voice/voice-selftest/VoiceSelfTestShell.js

@@ -33,10 +33,7 @@ function resolveTtsRoute(platform) {
     return platform === "web" ? "/api/tts/cloud" : "/api/tts/local-inference";
 }
 function getAudioCtx() {
-    const Ctor = window
-        .AudioContext ??
-        window
-            .webkitAudioContext;
+    const Ctor = window.AudioContext ?? window.webkitAudioContext;
     if (!Ctor)
         throw new Error("AudioContext unavailable");
     return new Ctor();

package/voice/voice-selftest/VoiceWorkbenchShell.d.ts

@@ -0,0 +1,29 @@
+/**
+ * Multi-turn voice SCENARIO player screen (?shellMode=voice-workbench).
+ *
+ * The headful half of the Voice Workbench (#8785). Renders OUTSIDE the app
+ * chrome / onboarding gate (mounted directly by App.tsx like the self-test
+ * shell), so it is reachable deterministically by a single URL param on web
+ * (Vite), desktop (Electrobun renderer) and Android (Capacitor WebView) — the
+ * SAME bundle covers all three platforms.
+ *
+ * It runs {@link runVoiceWorkbench} against the REAL production functions (no
+ * mocks here) for whatever {@link WorkbenchScenario} the automation passes, shows
+ * a per-turn PASS/FAIL/SKIPPED, and exposes:
+ *   - `window.__voiceWorkbench(scenario)` -> Promise<VoiceWorkbenchReport> for e2e
+ *   - a per-turn DOM mirror at [data-testid="voice-workbench-turn-<i>"] and an
+ *     overall verdict at [data-testid="voice-workbench-overall"]
+ * so an automated runner can scrape the verdict with no human in the loop.
+ */
+import { type VoiceWorkbenchReport, type WorkbenchScenario } from "./voice-workbench-player.js";
+declare global {
+    interface Window {
+        /** Legacy vendor-prefixed AudioContext (Safari / older WebKit). */
+        webkitAudioContext?: typeof AudioContext;
+        /** e2e automation hook — drives a WorkbenchScenario and returns its report. */
+        __voiceWorkbench?: (scenario: WorkbenchScenario) => Promise<VoiceWorkbenchReport>;
+    }
+}
+export declare function VoiceWorkbenchShell(): import("react/jsx-runtime").JSX.Element;
+export default VoiceWorkbenchShell;
+//# sourceMappingURL=VoiceWorkbenchShell.d.ts.map

package/voice/voice-selftest/VoiceWorkbenchShell.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"VoiceWorkbenchShell.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/VoiceWorkbenchShell.tsx"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;GAgBG;AAQH,OAAO,EAGL,KAAK,oBAAoB,EACzB,KAAK,iBAAiB,EAEvB,MAAM,0BAA0B,CAAC;AAElC,OAAO,CAAC,MAAM,CAAC;IACb,UAAU,MAAM;QACd,mEAAmE;QACnE,kBAAkB,CAAC,EAAE,OAAO,YAAY,CAAC;QACzC,+EAA+E;QAC/E,gBAAgB,CAAC,EAAE,CACjB,QAAQ,EAAE,iBAAiB,KACxB,OAAO,CAAC,oBAAoB,CAAC,CAAC;KACpC;CACF;AA8CD,wBAAgB,mBAAmB,4CAwIlC;AAED,eAAe,mBAAmB,CAAC"}

package/voice/voice-selftest/VoiceWorkbenchShell.js

@@ -0,0 +1,133 @@
+import { jsx as _jsx, jsxs as _jsxs } from "react/jsx-runtime";
+/**
+ * Multi-turn voice SCENARIO player screen (?shellMode=voice-workbench).
+ *
+ * The headful half of the Voice Workbench (#8785). Renders OUTSIDE the app
+ * chrome / onboarding gate (mounted directly by App.tsx like the self-test
+ * shell), so it is reachable deterministically by a single URL param on web
+ * (Vite), desktop (Electrobun renderer) and Android (Capacitor WebView) — the
+ * SAME bundle covers all three platforms.
+ *
+ * It runs {@link runVoiceWorkbench} against the REAL production functions (no
+ * mocks here) for whatever {@link WorkbenchScenario} the automation passes, shows
+ * a per-turn PASS/FAIL/SKIPPED, and exposes:
+ *   - `window.__voiceWorkbench(scenario)` -> Promise<VoiceWorkbenchReport> for e2e
+ *   - a per-turn DOM mirror at [data-testid="voice-workbench-turn-<i>"] and an
+ *     overall verdict at [data-testid="voice-workbench-overall"]
+ * so an automated runner can scrape the verdict with no human in the loop.
+ */
+import { useCallback, useEffect, useMemo, useRef, useState } from "react";
+import { ElizaClient } from "../../api/client-base.js";
+import { fetchWithCsrf } from "../../api/csrf-client.js";
+import { isElectrobunRuntime } from "../../bridge/electrobun-runtime.js";
+import { isAndroid } from "../../platform/init.js";
+import { resolveApiUrl } from "../../utils/index.js";
+import { runVoiceWorkbench, } from "./voice-workbench-player.js";
+function detectPlatform() {
+    if (isAndroid)
+        return "android";
+    if (isElectrobunRuntime())
+        return "desktop";
+    return "web";
+}
+function resolveTtsRoute(platform) {
+    // Desktop and Android run the on-device fused omnivoice TTS; only the web
+    // build (no on-device inference engine) falls back to cloud TTS.
+    return platform === "web" ? "/api/tts/cloud" : "/api/tts/local-inference";
+}
+function getAudioCtx() {
+    const Ctor = window.AudioContext ?? window.webkitAudioContext;
+    if (!Ctor)
+        throw new Error("AudioContext unavailable");
+    return new Ctor();
+}
+/**
+ * Corpus clip location for a turn: an explicit `audioRef` (relative to the
+ * scenario corpus root) or a deterministic per-turn path. The corpus generator
+ * writes `voice-corpus/<scenarioId>/turn-<i>.wav`; the e2e lanes route-mock
+ * this path. A missing clip makes the fetch throw → the turn is `skipped`.
+ */
+function turnWavUrl(scenarioId, turn, index) {
+    const ref = turn.audioRef?.trim();
+    const path = ref
+        ? `/voice-corpus/${scenarioId}/${ref}`
+        : `/voice-corpus/${scenarioId}/turn-${index}.wav`;
+    return resolveApiUrl(path);
+}
+const STATUS_COLOR = {
+    pass: "#2ec27e",
+    fail: "#e5484d",
+    skipped: "#9b9b9b",
+};
+export function VoiceWorkbenchShell() {
+    const platform = useMemo(detectPlatform, []);
+    const ttsRoute = useMemo(() => resolveTtsRoute(platform), [platform]);
+    const clientRef = useRef(null);
+    const audioRef = useRef(null);
+    const [report, setReport] = useState(null);
+    const [running, setRunning] = useState(false);
+    const run = useCallback(async (scenario) => {
+        setRunning(true);
+        try {
+            clientRef.current ??= new ElizaClient();
+            audioRef.current ??= getAudioCtx();
+            if (audioRef.current.state === "suspended") {
+                await audioRef.current.resume().catch(() => { });
+            }
+            const result = await runVoiceWorkbench({
+                scenario,
+                platform,
+                ttsRoute,
+                ttsExtraBody: ttsRoute === "/api/tts/cloud"
+                    ? {
+                        voiceId: "21m00Tcm4TlvDq8ikWAM",
+                        modelId: "eleven_turbo_v2_5",
+                    }
+                    : undefined,
+                resolveTurnWav: async (turn, index) => {
+                    const res = await fetchWithCsrf(turnWavUrl(scenario.id, turn, index), { method: "GET", headers: { Accept: "audio/*" } });
+                    if (!res.ok) {
+                        throw new Error(`corpus clip ${turn.audioRef ?? `turn-${index}.wav`} ${res.status}`);
+                    }
+                    return new Uint8Array(await res.arrayBuffer());
+                },
+                client: clientRef.current,
+                audioCtx: audioRef.current,
+            });
+            setReport(result);
+            return result;
+        }
+        finally {
+            setRunning(false);
+        }
+    }, [platform, ttsRoute]);
+    // Expose the player to automation. There is no default scenario — the runner
+    // (or the e2e lane) supplies the WorkbenchScenario to drive.
+    useEffect(() => {
+        window.__voiceWorkbench = (scenario) => run(scenario);
+    }, [run]);
+    return (_jsxs("div", { "data-testid": "voice-workbench-shell", "data-overall": report?.overall ?? "pending", style: {
+            position: "fixed",
+            inset: 0,
+            background: "#0b0b0b",
+            color: "#e8e8e8",
+            font: "14px ui-monospace, monospace",
+            padding: 24,
+            overflow: "auto",
+        }, children: [_jsx("h1", { style: { fontSize: 18, marginBottom: 4 }, children: "Voice workbench" }), _jsxs("div", { style: { color: "#9b9b9b", marginBottom: 16 }, children: ["platform=", platform, " \u00B7 ttsRoute=", ttsRoute, report ? ` · scenario=${report.scenarioId}` : "", report ? ` · classes=${report.classes.join(",")}` : ""] }), _jsxs("div", { "data-testid": "voice-workbench-overall", "data-overall": report?.overall ?? "pending", "data-diarization-status": report?.diarization.status ?? "pending", "data-der": report?.diarization.der ?? "", "data-max-der": report?.diarization.maxDer ?? "", "data-running": running ? "1" : "0", style: {
+                    fontSize: 16,
+                    fontWeight: 700,
+                    marginBottom: 12,
+                    color: report
+                        ? (STATUS_COLOR[report.overall] ?? "#e8e8e8")
+                        : "#9b9b9b",
+                }, children: ["overall: ", report?.overall ?? "pending"] }), _jsx("ul", { style: { listStyle: "none", padding: 0, margin: 0 }, children: (report?.turns ?? []).map((t) => (_jsxs("li", { "data-testid": `voice-workbench-turn-${t.index}`, "data-status": t.status, "data-speaker": t.speaker, "data-predicted-speaker-label": t.predictedSpeakerLabel ?? "", "data-expected-speaker-label": t.expectedSpeakerLabel, "data-responded": t.responded ? "1" : "0", "data-expect-respond": t.expectRespond ? "1" : "0", style: { marginBottom: 6 }, children: [_jsxs("span", { style: { color: STATUS_COLOR[t.status] ?? "#e8e8e8" }, children: ["[", t.status, "]"] }), " ", "turn ", t.index, " \u00B7 ", t.speaker, " (", t.durationMs, "ms)", t.error ? ` — ${t.error}` : ""] }, t.index))) }), _jsx("pre", { "data-testid": "voice-workbench-report", style: {
+                    marginTop: 16,
+                    padding: 12,
+                    background: "#141414",
+                    borderRadius: 6,
+                    whiteSpace: "pre-wrap",
+                    wordBreak: "break-word",
+                }, children: report ? JSON.stringify(report, null, 2) : "{}" })] }));
+}
+export default VoiceWorkbenchShell;

package/voice/voice-selftest/timing.d.ts

@@ -0,0 +1,8 @@
+/**
+ * Timing helpers shared by the voice self-test harness and the workbench
+ * player. `now()` reads a monotonic clock when available; `sleep()` awaits a
+ * fixed delay.
+ */
+export declare const now: () => number;
+export declare const sleep: (ms: number) => Promise<void>;
+//# sourceMappingURL=timing.d.ts.map

package/voice/voice-selftest/timing.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"timing.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/timing.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,eAAO,MAAM,GAAG,QAAO,MACqC,CAAC;AAE7D,eAAO,MAAM,KAAK,GAAI,IAAI,MAAM,KAAG,OAAO,CAAC,IAAI,CACI,CAAC"}

package/voice/voice-selftest/timing.js

@@ -0,0 +1,7 @@
+/**
+ * Timing helpers shared by the voice self-test harness and the workbench
+ * player. `now()` reads a monotonic clock when available; `sleep()` awaits a
+ * fixed delay.
+ */
+export const now = () => typeof performance !== "undefined" ? performance.now() : 0;
+export const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

package/voice/voice-selftest/voice-selftest-harness.d.ts

@@ -16,7 +16,7 @@
  * false-greens.
  */
 import { wordErrorRate } from "@elizaos/shared/voice-wer";
-import type { ElizaClient } from "../../api/client-base";
+import type { ElizaClient } from "../../api/client-base.js";
 /** Re-exported from the single source of truth (`@elizaos/shared/voice-wer`). */
 export { wordErrorRate };
 export type StageStatus = "pass" | "fail" | "skipped";

package/voice/voice-selftest/voice-selftest-harness.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"voice-selftest-harness.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/voice-selftest-harness.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;GAgBG;AAEH,OAAO,EAAE,aAAa,EAAE,MAAM,2BAA2B,CAAC;AAC1D,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uBAAuB,CAAC;AASzD,iFAAiF;AACjF,OAAO,EAAE,aAAa,EAAE,CAAC;AAEzB,MAAM,MAAM,WAAW,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;AACtD,MAAM,MAAM,iBAAiB,GACzB,YAAY,GACZ,aAAa,GACb,mBAAmB,CAAC;AACxB,MAAM,MAAM,qBAAqB,GAAG,KAAK,GAAG,SAAS,GAAG,SAAS,CAAC;AAElE,MAAM,WAAW,kBAAkB;IACjC,KAAK,EAAE,KAAK,GAAG,MAAM,GAAG,KAAK,CAAC;IAC9B,MAAM,EAAE,WAAW,CAAC;IACpB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC,CAAC;IAClD,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,mBAAmB;IAClC,aAAa,EAAE,CAAC,CAAC;IACjB,OAAO,EAAE,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;IACrC,QAAQ,EAAE,qBAAqB,CAAC;IAChC,IAAI,EAAE,iBAAiB,CAAC;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,cAAc,EAAE,MAAM,CAAC;IACvB,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;IACd,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,kBAAkB,EAAE,CAAC;CAC9B;AAED,MAAM,WAAW,oBAAoB;IACnC,QAAQ,EAAE,qBAAqB,CAAC;IAChC,8EAA8E;IAC9E,IAAI,CAAC,EAAE,iBAAiB,CAAC;IACzB,mDAAmD;IACnD,UAAU,EAAE,MAAM,CAAC;IACnB,oDAAoD;IACpD,cAAc,EAAE,MAAM,CAAC;IACvB,4EAA4E;IAC5E,QAAQ,EAAE,0BAA0B,GAAG,gBAAgB,CAAC;IACxD,wEAAwE;IACxE,YAAY,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACvC,2EAA2E;IAC3E,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,kEAAkE;IAClE,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,MAAM,EAAE,WAAW,CAAC;IACpB,QAAQ,EAAE,YAAY,CAAC;IACvB,MAAM,CAAC,EAAE,WAAW,CAAC;CACtB;AA+GD,wBAAsB,gBAAgB,CACpC,IAAI,EAAE,oBAAoB,GACzB,OAAO,CAAC,mBAAmB,CAAC,CAiN9B"}
+{"version":3,"file":"voice-selftest-harness.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/voice-selftest-harness.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;GAgBG;AAEH,OAAO,EAAE,aAAa,EAAE,MAAM,2BAA2B,CAAC;AAC1D,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uBAAuB,CAAC;AAUzD,iFAAiF;AACjF,OAAO,EAAE,aAAa,EAAE,CAAC;AAEzB,MAAM,MAAM,WAAW,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;AACtD,MAAM,MAAM,iBAAiB,GACzB,YAAY,GACZ,aAAa,GACb,mBAAmB,CAAC;AACxB,MAAM,MAAM,qBAAqB,GAAG,KAAK,GAAG,SAAS,GAAG,SAAS,CAAC;AAElE,MAAM,WAAW,kBAAkB;IACjC,KAAK,EAAE,KAAK,GAAG,MAAM,GAAG,KAAK,CAAC;IAC9B,MAAM,EAAE,WAAW,CAAC;IACpB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC,CAAC;IAClD,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,mBAAmB;IAClC,aAAa,EAAE,CAAC,CAAC;IACjB,OAAO,EAAE,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;IACrC,QAAQ,EAAE,qBAAqB,CAAC;IAChC,IAAI,EAAE,iBAAiB,CAAC;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,cAAc,EAAE,MAAM,CAAC;IACvB,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;IACd,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,kBAAkB,EAAE,CAAC;CAC9B;AAED,MAAM,WAAW,oBAAoB;IACnC,QAAQ,EAAE,qBAAqB,CAAC;IAChC,8EAA8E;IAC9E,IAAI,CAAC,EAAE,iBAAiB,CAAC;IACzB,mDAAmD;IACnD,UAAU,EAAE,MAAM,CAAC;IACnB,oDAAoD;IACpD,cAAc,EAAE,MAAM,CAAC;IACvB,4EAA4E;IAC5E,QAAQ,EAAE,0BAA0B,GAAG,gBAAgB,CAAC;IACxD,wEAAwE;IACxE,YAAY,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACvC,2EAA2E;IAC3E,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,kEAAkE;IAClE,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,MAAM,EAAE,WAAW,CAAC;IACpB,QAAQ,EAAE,YAAY,CAAC;IACvB,MAAM,CAAC,EAAE,WAAW,CAAC;CACtB;AAyGD,wBAAsB,gBAAgB,CACpC,IAAI,EAAE,oBAAoB,GACzB,OAAO,CAAC,mBAAmB,CAAC,CAiN9B"}

package/voice/voice-selftest/voice-selftest-harness.js

@@ -20,6 +20,7 @@ import { fetchWithCsrf } from "../../api/csrf-client.js";
 import { resolveApiUrl } from "../../utils/index.js";
 import { startLocalAsrRecorder } from "../local-asr-capture.js";
 import { isLocalInferenceAsrReady, transcribeLocalInferenceWav, } from "../local-asr-transcribe.js";
+import { now, sleep } from "./timing.js";
 /** Re-exported from the single source of truth (`@elizaos/shared/voice-wer`). */
 export { wordErrorRate };
 /**
@@ -49,8 +50,6 @@ async function captureMicWav(signal) {
         });
     });
 }
-const now = () => typeof performance !== "undefined" ? performance.now() : 0;
-const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));
 /**
  * Peak + RMS amplitude across every channel of a decoded buffer. A buffer of
  * pure silence decodes fine and reports a positive `duration`, so duration

package/voice/voice-selftest/voice-workbench-player.d.ts

@@ -0,0 +1,139 @@
+/**
+ * Multi-turn voice SCENARIO player — the headful half of the Voice Workbench
+ * (#8785). Where {@link runVoiceSelfTest} drives ONE phrase through the real
+ * client pipeline, this drives an ordered {@link VoiceScenario} turn-by-turn
+ * through the SAME real production functions (transcribeLocalInferenceWav,
+ * ElizaClient.sendConversationMessageStream, a real TTS fetch +
+ * AudioContext.decodeAudioData), honouring injected pauses and scoring every
+ * turn against the scenario's ground truth (expected transcript, respond
+ * decision, speaker label, entity).
+ *
+ * Reused by:
+ *   - the in-app voice workbench screen (?shellMode=voice-workbench)
+ *   - the web / android / desktop e2e lanes (they navigate to that screen and
+ *     scrape `window.__voiceWorkbench(scenario)` / the DOM-mirrored report).
+ *
+ * Honesty contract (identical to the self-test): a turn whose backend / corpus
+ * artifact is genuinely absent on this host reports `skipped` — NEVER `pass` —
+ * so CI can tell "can't run here" from "verified working" and never false-green.
+ *
+ * Scenario type: the canonical `VoiceScenario` schema lives in
+ * `@elizaos/plugin-local-inference` (`src/services/voice/voice-scenario.ts`).
+ * That plugin depends inward on `@elizaos/ui`; importing it back here would
+ * invert the dependency direction, so we mirror the exact transport shape as a
+ * local structural type. The plugin's `validateVoiceScenario` remains the single
+ * source of truth for cross-field validity; the player only reads the fields it
+ * drives and never re-implements scenario validation.
+ */
+import type { ElizaClient } from "../../api/client-base.js";
+export type TurnStatus = "pass" | "fail" | "skipped";
+/** Structural mirror of `VoiceScenarioParticipant` (@elizaos/plugin-local-inference). */
+export interface WorkbenchParticipant {
+    label: string;
+    ttsVoiceId?: string;
+    entityId?: string;
+    isOwner?: boolean;
+}
+/** Structural mirror of `VoiceScenarioTurn` (@elizaos/plugin-local-inference). */
+export interface WorkbenchTurn {
+    /** Ground-truth speaker identity that synthesized this turn's audio. */
+    speaker: string;
+    text?: string;
+    audioRef?: string;
+    ttsVoiceId?: string;
+    pausesMs?: number[];
+    expectRespond: boolean;
+    expectedTranscript?: string;
+    expectedSpeakerLabel?: string;
+    expectedEntity?: string;
+}
+/** Structural mirror of `VoiceScenario` (@elizaos/plugin-local-inference). */
+export interface WorkbenchScenario {
+    id: string;
+    description?: string;
+    classes: string[];
+    participants: WorkbenchParticipant[];
+    turns: WorkbenchTurn[];
+    agents?: string[];
+}
+export type VoiceWorkbenchPlatform = "web" | "android" | "desktop";
+/** A single scored turn of the scenario. */
+export interface VoiceWorkbenchTurnReport {
+    index: number;
+    speaker: string;
+    expectedSpeakerLabel: string;
+    predictedSpeakerLabel: string | null;
+    status: TurnStatus;
+    /** Did the real client pipeline produce a reply for this turn? */
+    responded: boolean;
+    /** Ground-truth respond decision for this turn. */
+    expectRespond: boolean;
+    /** ASR transcript for this turn (empty when ASR was skipped). */
+    transcript: string;
+    /** Expected ASR reference (explicit `expectedTranscript` or the turn text). */
+    expectedTranscript: string;
+    /** Agent reply text ("" when the agent did not respond / send was skipped). */
+    reply: string;
+    durationMs: number;
+    detail: Record<string, string | number | boolean>;
+    error?: string;
+}
+export interface VoiceWorkbenchReport {
+    schemaVersion: 1;
+    overall: "pass" | "fail" | "skipped";
+    scenarioId: string;
+    classes: string[];
+    platform: VoiceWorkbenchPlatform;
+    ttsRoute: string;
+    startedAt: string;
+    finishedAt: string;
+    turns: VoiceWorkbenchTurnReport[];
+    diarization: {
+        status: TurnStatus;
+        /** Turns with a real attribution output (the DER denominator). */
+        total: number;
+        der: number;
+        confusions: number;
+        /** Turns with no attribution output — no diarizer ran; excluded from DER. */
+        unattributed: number;
+        maxDer: number;
+        /** Did any turn carry a real attribution output? When false the gate is
+         * SKIPPED (no diarizer on this host), distinct from a real DER failure. */
+        evaluated: boolean;
+        passed: boolean;
+        reason?: string;
+    };
+}
+export interface VoiceWorkbenchOptions {
+    scenario: WorkbenchScenario;
+    platform: VoiceWorkbenchPlatform;
+    /**
+     * Resolves a turn to a decodable WAV (`Uint8Array`). The corpus generator
+     * synthesizes one clip per turn; the player fetches it by `audioRef` or by a
+     * deterministic per-turn url. When a turn's clip is absent the resolver
+     * throws and the turn is reported `skipped`, never `pass`.
+     */
+    resolveTurnWav: (turn: WorkbenchTurn, index: number) => Promise<Uint8Array>;
+    /**
+     * Resolve the PREDICTED speaker label for a turn from an actual
+     * diarization/speaker-attribution output. When provided, its result — NOT the
+     * scenario's ground-truth `speaker` — drives the diarization gate, so the gate
+     * fails on a real misattribution. Return `null` when the model can't attribute
+     * the turn (unattributed: excluded from DER, the gate reports skipped, never a
+     * pass). When absent, the player records no predicted speaker label; it never
+     * falls back to `speaker`.
+     */
+    resolvePredictedSpeakerLabel?: (turn: WorkbenchTurn, index: number, wav: Uint8Array) => Promise<string | null> | string | null;
+    /** TTS route to exercise. local for desktop/android, cloud for web. */
+    ttsRoute: "/api/tts/local-inference" | "/api/tts/cloud";
+    /** Extra TTS body fields (e.g. voiceId/modelId for the cloud route). */
+    ttsExtraBody?: Record<string, unknown>;
+    /** Max word-error-rate a turn's ASR transcript may have vs its reference. */
+    werTolerance?: number;
+    client: ElizaClient;
+    audioCtx: AudioContext;
+    signal?: AbortSignal;
+}
+export declare function scoreWorkbenchDiarization(turns: ReadonlyArray<VoiceWorkbenchTurnReport>, maxDer?: number): VoiceWorkbenchReport["diarization"];
+export declare function runVoiceWorkbench(opts: VoiceWorkbenchOptions): Promise<VoiceWorkbenchReport>;
+//# sourceMappingURL=voice-workbench-player.d.ts.map

package/voice/voice-selftest/voice-workbench-player.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"voice-workbench-player.d.ts","sourceRoot":"","sources":["../../../../../../src/voice/voice-selftest/voice-workbench-player.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAGH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uBAAuB,CAAC;AASzD,MAAM,MAAM,UAAU,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;AAErD,yFAAyF;AACzF,MAAM,WAAW,oBAAoB;IACnC,KAAK,EAAE,MAAM,CAAC;IACd,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,OAAO,CAAC,EAAE,OAAO,CAAC;CACnB;AAED,kFAAkF;AAClF,MAAM,WAAW,aAAa;IAC5B,wEAAwE;IACxE,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;IACpB,aAAa,EAAE,OAAO,CAAC;IACvB,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,8EAA8E;AAC9E,MAAM,WAAW,iBAAiB;IAChC,EAAE,EAAE,MAAM,CAAC;IACX,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,YAAY,EAAE,oBAAoB,EAAE,CAAC;IACrC,KAAK,EAAE,aAAa,EAAE,CAAC;IACvB,MAAM,CAAC,EAAE,MAAM,EAAE,CAAC;CACnB;AAED,MAAM,MAAM,sBAAsB,GAAG,KAAK,GAAG,SAAS,GAAG,SAAS,CAAC;AAEnE,4CAA4C;AAC5C,MAAM,WAAW,wBAAwB;IACvC,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,EAAE,MAAM,CAAC;IAChB,oBAAoB,EAAE,MAAM,CAAC;IAC7B,qBAAqB,EAAE,MAAM,GAAG,IAAI,CAAC;IACrC,MAAM,EAAE,UAAU,CAAC;IACnB,kEAAkE;IAClE,SAAS,EAAE,OAAO,CAAC;IACnB,mDAAmD;IACnD,aAAa,EAAE,OAAO,CAAC;IACvB,iEAAiE;IACjE,UAAU,EAAE,MAAM,CAAC;IACnB,+EAA+E;IAC/E,kBAAkB,EAAE,MAAM,CAAC;IAC3B,+EAA+E;IAC/E,KAAK,EAAE,MAAM,CAAC;IACd,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC,CAAC;IAClD,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,oBAAoB;IACnC,aAAa,EAAE,CAAC,CAAC;IACjB,OAAO,EAAE,MAAM,GAAG,MAAM,GAAG,SAAS,CAAC;IACrC,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,QAAQ,EAAE,sBAAsB,CAAC;IACjC,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,wBAAwB,EAAE,CAAC;IAClC,WAAW,EAAE;QACX,MAAM,EAAE,UAAU,CAAC;QACnB,kEAAkE;QAClE,KAAK,EAAE,MAAM,CAAC;QACd,GAAG,EAAE,MAAM,CAAC;QACZ,UAAU,EAAE,MAAM,CAAC;QACnB,6EAA6E;QAC7E,YAAY,EAAE,MAAM,CAAC;QACrB,MAAM,EAAE,MAAM,CAAC;QACf;mFAC2E;QAC3E,SAAS,EAAE,OAAO,CAAC;QACnB,MAAM,EAAE,OAAO,CAAC;QAChB,MAAM,CAAC,EAAE,MAAM,CAAC;KACjB,CAAC;CACH;AAED,MAAM,WAAW,qBAAqB;IACpC,QAAQ,EAAE,iBAAiB,CAAC;IAC5B,QAAQ,EAAE,sBAAsB,CAAC;IACjC;;;;;OAKG;IACH,cAAc,EAAE,CAAC,IAAI,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,KAAK,OAAO,CAAC,UAAU,CAAC,CAAC;IAC5E;;;;;;;;OAQG;IACH,4BAA4B,CAAC,EAAE,CAC7B,IAAI,EAAE,aAAa,EACnB,KAAK,EAAE,MAAM,EACb,GAAG,EAAE,UAAU,KACZ,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,GAAG,MAAM,GAAG,IAAI,CAAC;IAC5C,uEAAuE;IACvE,QAAQ,EAAE,0BAA0B,GAAG,gBAAgB,CAAC;IACxD,wEAAwE;IACxE,YAAY,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACvC,6EAA6E;IAC7E,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,MAAM,EAAE,WAAW,CAAC;IACpB,QAAQ,EAAE,YAAY,CAAC;IACvB,MAAM,CAAC,EAAE,WAAW,CAAC;CACtB;AA0SD,wBAAgB,yBAAyB,CACvC,KAAK,EAAE,aAAa,CAAC,wBAAwB,CAAC,EAC9C,MAAM,SAAM,GACX,oBAAoB,CAAC,aAAa,CAAC,CAmCrC;AAED,wBAAsB,iBAAiB,CACrC,IAAI,EAAE,qBAAqB,GAC1B,OAAO,CAAC,oBAAoB,CAAC,CAyE/B"}