bosun 0.37.1 → 0.37.2

package/setup-web-server.mjs

@@ -442,6 +442,9 @@ function buildStableSetupDefaults({
     voiceTurnDetection: "semantic_vad",
     voiceFallbackMode: "browser",
     voiceDelegateExecutor: "codex-sdk",
+    voiceTranscriptionEnabled: true,
+    voiceTranscriptionModel: "gpt-4o-transcribe",
+    voiceAzureTranscriptionEnabled: false,
     openaiRealtimeApiKey: "",
     azureOpenaiRealtimeEndpoint: "",
     azureOpenaiRealtimeApiKey: "",
@@ -914,6 +917,32 @@ function applyNonBlockingSetupEnvDefaults(envMap, env = {}, sourceEnv = process.
     ["codex-sdk", "copilot-sdk", "claude-sdk", "gemini-sdk", "opencode-sdk"],
     "codex-sdk",
   );
+  envMap.VOICE_TRANSCRIPTION_ENABLED = toBooleanEnvString(
+    pickNonEmptyValue(
+      env.voiceTranscriptionEnabled,
+      env.VOICE_TRANSCRIPTION_ENABLED,
+      envMap.VOICE_TRANSCRIPTION_ENABLED,
+      sourceEnv.VOICE_TRANSCRIPTION_ENABLED,
+    ),
+    true,
+  );
+  envMap.VOICE_TRANSCRIPTION_MODEL = String(
+    pickNonEmptyValue(
+      env.voiceTranscriptionModel,
+      env.VOICE_TRANSCRIPTION_MODEL,
+      envMap.VOICE_TRANSCRIPTION_MODEL,
+      sourceEnv.VOICE_TRANSCRIPTION_MODEL,
+    ) || "gpt-4o-transcribe",
+  ).trim() || "gpt-4o-transcribe";
+  envMap.VOICE_AZURE_TRANSCRIPTION_ENABLED = toBooleanEnvString(
+    pickNonEmptyValue(
+      env.voiceAzureTranscriptionEnabled,
+      env.VOICE_AZURE_TRANSCRIPTION_ENABLED,
+      envMap.VOICE_AZURE_TRANSCRIPTION_ENABLED,
+      sourceEnv.VOICE_AZURE_TRANSCRIPTION_ENABLED,
+    ),
+    false,
+  );
 
   const openaiRealtimeApiKey = pickNonEmptyValue(
     env.openaiRealtimeApiKey,
@@ -1481,14 +1510,12 @@ async function handleVoiceEndpointTest(body) {
       // Strip path suffix so users can paste full URLs without double-path 404s.
       let base = String(azureEndpoint).replace(/\/+$/, "");
       try { const u = new URL(base); base = `${u.protocol}//${u.host}`; } catch { /* keep as-is */ }
-      // Single-deployment GET only requires Cognitive Services User role.
-      // Use the GA api-version (2024-10-21) for broad compatibility across
-      // classic Azure OpenAI and Azure AI Foundry resources.
-      const dep = String(deployment || "").trim();
-      testUrl = dep
-        ? `${base}/openai/deployments/${encodeURIComponent(dep)}?api-version=2024-10-21`
-        : `${base}/openai/models?api-version=2024-10-21`;
+      // Prefer /openai/models as a credential check — it works on both classic
+      // Azure OpenAI resources AND Azure AI Foundry "Global Standard" deployments.
+      // If a deployment name is provided, we verify it separately after confirming
+      // the endpoint + key are valid.
       headers["api-key"] = apiKey;
+      testUrl = `${base}/openai/models?api-version=2024-10-21`;
     } else if (normalizedProvider === "claude") {
       testUrl = "https://api.anthropic.com/v1/models";
       headers["anthropic-version"] = "2023-06-01";
@@ -1525,6 +1552,36 @@ async function handleVoiceEndpointTest(body) {
     clearTimeout(timer);
     const latencyMs = Date.now() - start;
     if (resp.ok || resp.status === 200) {
+      // For Azure with a deployment name, do a secondary check to verify the
+      // deployment exists. We try chat/completions with max_tokens=1 — realtime
+      // deployments return 400 (wrong endpoint type) which still confirms existence.
+      if (normalizedProvider === "azure" && deployment) {
+        const dep = String(deployment).trim();
+        try {
+          let base = String(azureEndpoint).replace(/\/+$/, "");
+          try { const u = new URL(base); base = `${u.protocol}//${u.host}`; } catch { /* keep */ }
+          const depUrl = `${base}/openai/deployments/${encodeURIComponent(dep)}/chat/completions?api-version=2024-10-21`;
+          const depCtrl = new AbortController();
+          const depTimer = setTimeout(() => depCtrl.abort(), 8_000);
+          const depResp = await fetch(depUrl, {
+            method: "POST",
+            headers: { ...headers, "Content-Type": "application/json" },
+            body: JSON.stringify({ messages: [{ role: "user", content: "test" }], max_tokens: 1 }),
+            signal: depCtrl.signal,
+          });
+          clearTimeout(depTimer);
+          // 200 = chat model works, 400 = realtime model (expected), both confirm deployment exists
+          if (depResp.ok || depResp.status === 400) {
+            return { ok: true, latencyMs, deployment: dep };
+          }
+          if (depResp.status === 404) {
+            return { ok: false, error: `Credentials valid but deployment "${dep}" not found — check the deployment name in Azure AI Foundry`, latencyMs };
+          }
+          return { ok: true, latencyMs, warning: `Credentials valid. Could not verify deployment "${dep}" (HTTP ${depResp.status})` };
+        } catch {
+          return { ok: true, latencyMs, warning: `Credentials valid. Could not verify deployment "${dep}" (timeout)` };
+        }
+      }
       return { ok: true, latencyMs };
     }
     const text = await resp.text().catch(() => "");
@@ -1535,9 +1592,13 @@ async function handleVoiceEndpointTest(body) {
     } catch {
       // Keep generic HTTP status message.
     }
-    // Friendly message when the deployment name itself is not found (key is fine)
-    if (resp.status === 404 && deployment) {
-      error = `Deployment "${deployment}" not found — check deployment name in Azure AI Foundry`;
+    // Azure-specific: provide helpful messages for common errors
+    if (normalizedProvider === "azure") {
+      if (resp.status === 401 || resp.status === 403) {
+        error = `Authentication failed (HTTP ${resp.status}) — check API key and endpoint URL`;
+      } else if (resp.status === 404) {
+        error = `Endpoint not found (HTTP 404) — check the Azure endpoint URL. Use https://<resource>.openai.azure.com`;
+      }
     }
     return { ok: false, error, latencyMs };
   } catch (err) {

package/ui/app.js

@@ -70,6 +70,7 @@ const VOICE_LAUNCH_QUERY_KEYS = [
   "executor",
   "mode",
   "model",
+  "voiceAgentId",
   "vision",
   "source",
   "chat_id",
@@ -127,6 +128,7 @@ function parseVoiceLaunchFromUrl() {
       executor: String(params.get("executor") || "").trim() || null,
       mode: String(params.get("mode") || "").trim() || null,
       model: String(params.get("model") || "").trim() || null,
+      voiceAgentId: String(params.get("voiceAgentId") || "").trim() || null,
     },
   };
 }
@@ -166,11 +168,13 @@ function buildBrowserFollowUrl(detail = {}) {
   const executor = String(detail?.executor || "").trim();
   const mode = String(detail?.mode || "").trim();
   const model = String(detail?.model || "").trim();
+  const voiceAgentId = String(detail?.voiceAgentId || "").trim();
   const vision = String(detail?.initialVisionSource || "").trim();
   if (sessionId) target.searchParams.set("sessionId", sessionId);
   if (executor) target.searchParams.set("executor", executor);
   if (mode) target.searchParams.set("mode", mode);
   if (model) target.searchParams.set("model", model);
+  if (voiceAgentId) target.searchParams.set("voiceAgentId", voiceAgentId);
   if (vision) target.searchParams.set("vision", vision);
   return target.toString();
 }
@@ -1464,12 +1468,14 @@ function App() {
   const [voiceExecutor, setVoiceExecutor] = useState(null);
   const [voiceAgentMode, setVoiceAgentMode] = useState(null);
   const [voiceModel, setVoiceModel] = useState(null);
+  const [voiceAgentId, setVoiceAgentId] = useState(null);
   const [voiceCallType, setVoiceCallType] = useState("voice");
   const [voiceInitialVisionSource, setVoiceInitialVisionSource] = useState(
     null,
   );
   const followWindowMode = isFollowWindowFromUrl();
   const followOverlayOpenedRef = useRef(false);
+  const externalizeInFlightRef = useRef(false);
   const [floatingCallState, setFloatingCallState] = useState(() =>
     readFloatingCallState(),
   );
@@ -1805,6 +1811,9 @@ function App() {
         const currentModel =
           String(event?.detail?.model || selectedModel.value || "").trim() ||
           null;
+        const currentVoiceAgentId =
+          String(event?.detail?.voiceAgentId || voiceAgentId || "").trim() ||
+          null;
         const explicitSessionId =
           String(event?.detail?.sessionId || "").trim() || null;
         let currentSessionId =
@@ -1836,6 +1845,7 @@ function App() {
         setVoiceExecutor(currentExecutor);
         setVoiceAgentMode(currentMode);
         setVoiceModel(currentModel);
+        setVoiceAgentId(currentVoiceAgentId);
         setVoiceCallType(requestedCallType);
         setVoiceInitialVisionSource(requestedVisionSource);
 
@@ -1859,6 +1869,7 @@ function App() {
             executor: currentExecutor || undefined,
             mode: currentMode || undefined,
             model: currentModel || undefined,
+            voiceAgentId: currentVoiceAgentId || undefined,
           });
           if (followResult?.ok) {
             const nextFloatingState = {
@@ -1868,6 +1879,7 @@ function App() {
               executor: currentExecutor,
               mode: currentMode,
               model: currentModel,
+              voiceAgentId: currentVoiceAgentId,
               initialVisionSource: requestedVisionSource,
             };
             setFloatingCallState(nextFloatingState);
@@ -1890,7 +1902,7 @@ function App() {
     globalThis.addEventListener?.("ve:open-voice-mode", handleOpenVoiceMode);
     return () =>
       globalThis.removeEventListener?.("ve:open-voice-mode", handleOpenVoiceMode);
-  }, [followWindowMode]);
+  }, [followWindowMode, voiceAgentId]);
 
   useEffect(() => {
     const onStorage = (event) => {
@@ -1912,6 +1924,7 @@ function App() {
       executor: voiceExecutor,
       mode: voiceAgentMode,
       model: voiceModel,
+      voiceAgentId,
       initialVisionSource: voiceInitialVisionSource,
     };
     setFloatingCallState(nextFloatingState);
@@ -1924,6 +1937,7 @@ function App() {
     voiceExecutor,
     voiceAgentMode,
     voiceModel,
+    voiceAgentId,
     voiceInitialVisionSource,
   ]);
 
@@ -1937,6 +1951,7 @@ function App() {
         executor: voiceExecutor,
         mode: voiceAgentMode,
         model: voiceModel,
+        voiceAgentId,
         initialVisionSource: voiceInitialVisionSource,
       };
       setFloatingCallState(nextFloatingState);
@@ -1951,6 +1966,7 @@ function App() {
     voiceExecutor,
     voiceAgentMode,
     voiceModel,
+    voiceAgentId,
     voiceInitialVisionSource,
   ]);
 
@@ -2002,7 +2018,10 @@ function App() {
           navigateTo("chat", { replace: true, skipGuard: true });
         }
       }
-      await new Promise((resolve) => setTimeout(resolve, 60));
+      // Wait for UI components to mount before dispatching the voice launch
+      // event.  60 ms was too aggressive for cold-start Electron windows where
+      // JS bundles are still being parsed; 200 ms is reliably sufficient.
+      await new Promise((resolve) => setTimeout(resolve, 200));
       if (cancelled) return;
       globalThis.dispatchEvent?.(
         new CustomEvent("ve:open-voice-mode", { detail: launch.detail }),
@@ -2341,6 +2360,7 @@ function App() {
                   executor: floatingCallState?.executor,
                   mode: floatingCallState?.mode,
                   model: floatingCallState?.model,
+                  voiceAgentId: floatingCallState?.voiceAgentId,
                 });
                 if (!popupResult.ok) {
                   showToast(
@@ -2366,6 +2386,10 @@ function App() {
       onDismiss=${(detail = {}) => {
         const reason = String(detail?.reason || "").trim().toLowerCase();
         if (!followWindowMode && reason === "externalize") {
+          if (externalizeInFlightRef.current) {
+            return;
+          }
+          externalizeInFlightRef.current = true;
           const followDetail = {
             call: voiceCallType,
             sessionId: voiceSessionId,
@@ -2373,6 +2397,7 @@ function App() {
             executor: voiceExecutor,
             mode: voiceAgentMode,
             model: voiceModel,
+            voiceAgentId,
           };
           const desktopFollowApi = globalThis?.veDesktop?.follow;
           if (typeof desktopFollowApi?.open === "function") {
@@ -2390,13 +2415,17 @@ function App() {
                   executor: followDetail.executor,
                   mode: followDetail.mode,
                   model: followDetail.model,
+                  voiceAgentId: followDetail.voiceAgentId,
                   initialVisionSource: followDetail.initialVisionSource,
                 };
                 setFloatingCallState(nextFloatingState);
                 writeFloatingCallState(nextFloatingState);
                 setVoiceOverlayOpen(false);
               })
-              .catch(() => showToast("Could not open floating call window.", "error"));
+              .catch(() => showToast("Could not open floating call window.", "error"))
+              .finally(() => {
+                externalizeInFlightRef.current = false;
+              });
             return;
           }
           const popupResult = openBrowserFollowWindow(followDetail);
@@ -2405,6 +2434,7 @@ function App() {
               popupResult.reason || "Could not open floating browser call window.",
               "error",
             );
+            externalizeInFlightRef.current = false;
             return;
           }
           const nextFloatingState = {
@@ -2414,13 +2444,16 @@ function App() {
             executor: followDetail.executor,
             mode: followDetail.mode,
             model: followDetail.model,
+            voiceAgentId: followDetail.voiceAgentId,
             initialVisionSource: followDetail.initialVisionSource,
           };
           setFloatingCallState(nextFloatingState);
           writeFloatingCallState(nextFloatingState);
           setVoiceOverlayOpen(false);
+          externalizeInFlightRef.current = false;
           return;
         }
+        externalizeInFlightRef.current = false;
         if (followWindowMode && globalThis?.veDesktop?.follow?.hide) {
           globalThis.veDesktop.follow.hide().catch(() => {});
           return;
@@ -2432,6 +2465,10 @@ function App() {
       executor=${voiceExecutor}
       mode=${voiceAgentMode}
       model=${voiceModel}
+      voiceAgentId=${voiceAgentId}
+      onVoiceAgentChange=${(nextAgentId) => {
+        setVoiceAgentId(String(nextAgentId || "").trim() || null);
+      }}
       callType=${voiceCallType}
       initialVisionSource=${voiceInitialVisionSource}
       compact=${followWindowMode}

package/ui/components/session-list.js

@@ -36,9 +36,19 @@ function sessionPath(id, action = "") {
 
 /* ─── Data loaders ─── */
 export async function loadSessions(filter = {}) {
-  _lastLoadFilter = filter;
+  const normalizedFilter = {
+    ...(filter && typeof filter === "object" ? filter : {}),
+  };
+  if (!Object.prototype.hasOwnProperty.call(normalizedFilter, "workspace")) {
+    normalizedFilter.workspace = "active";
+  }
+  _lastLoadFilter = normalizedFilter;
   try {
-    const params = new URLSearchParams(filter);
+    const params = new URLSearchParams();
+    for (const [key, value] of Object.entries(normalizedFilter)) {
+      if (value == null || value === "") continue;
+      params.set(key, String(value));
+    }
     const res = await apiFetch(`/api/sessions?${params}`, { _silent: true });
     if (res?.sessions) sessionsData.value = res.sessions;
     sessionsError.value = null;
@@ -317,11 +327,19 @@ export function initSessionWsListener() {
   if (_wsListenerReady) return;
   _wsListenerReady = true;
   onWsMessage((msg) => {
-    if (msg?.type !== "session-message") return;
-    const payload = msg.payload || {};
-    const sessionId = payload.sessionId || payload.taskId;
-    if (!sessionId) return;
-    appendSessionMessage(sessionId, payload.message, payload.session);
+    if (msg?.type === "session-message") {
+      const payload = msg.payload || {};
+      const sessionId = payload.sessionId || payload.taskId;
+      if (!sessionId) return;
+      appendSessionMessage(sessionId, payload.message, payload.session);
+      return;
+    }
+    if (msg?.type === "invalidate") {
+      const channels = Array.isArray(msg.channels) ? msg.channels : [];
+      if (channels.includes("*") || channels.includes("sessions")) {
+        loadSessions(_lastLoadFilter).catch(() => {});
+      }
+    }
   });
 }
 

package/ui/components/workspace-switcher.js

@@ -8,7 +8,7 @@ import { h } from "preact";
 import { useState, useEffect, useCallback } from "preact/hooks";
 import { signal } from "@preact/signals";
 import htm from "htm";
-import { apiFetch } from "../modules/api.js";
+import { apiFetch, onWsMessage } from "../modules/api.js";
 import { haptic } from "../modules/telegram.js";
 import { Modal } from "./shared.js";
 import { iconText, resolveIcon } from "../modules/icon-utils.js";
@@ -53,6 +53,24 @@ export async function switchWorkspace(wsId) {
     }
     activeWorkspaceId.value = String(res.activeId || wsId);
     await loadWorkspaces();
+    try {
+      globalThis.dispatchEvent?.(
+        new CustomEvent("ve:workspace-switched", {
+          detail: { workspaceId: activeWorkspaceId.value || String(wsId || "") },
+        }),
+      );
+    } catch {
+      // no-op
+    }
+    try {
+      const { refreshTab } = await import("../modules/state.js");
+      await Promise.allSettled([
+        refreshTab("tasks", { background: true, manual: false }),
+        refreshTab("dashboard", { background: true, manual: false }),
+      ]);
+    } catch {
+      // best effort
+    }
     return true;
   } catch (err) {
     console.warn("[workspace-switcher] Failed to switch workspace:", err);
@@ -461,6 +479,35 @@ export function WorkspaceSwitcher() {
     loadWorkspaces();
   }, []);
 
+  // Keep selector state in sync when workspace is switched externally
+  // (for example via Electron menu or another client).
+  useEffect(() => {
+    const unsubscribe = onWsMessage((msg) => {
+      if (msg?.type !== "invalidate") return;
+      const channels = Array.isArray(msg.channels) ? msg.channels : [];
+      if (channels.includes("*") || channels.includes("workspaces")) {
+        loadWorkspaces().catch(() => {});
+      }
+    });
+    return unsubscribe;
+  }, []);
+
+  // Desktop fallback when WS is unavailable: refresh workspace state whenever
+  // the window regains focus/visibility.
+  useEffect(() => {
+    const sync = () => loadWorkspaces().catch(() => {});
+    const onFocus = () => sync();
+    const onVisibility = () => {
+      if (document.visibilityState === "visible") sync();
+    };
+    globalThis.addEventListener?.("focus", onFocus);
+    document.addEventListener?.("visibilitychange", onVisibility);
+    return () => {
+      globalThis.removeEventListener?.("focus", onFocus);
+      document.removeEventListener?.("visibilitychange", onVisibility);
+    };
+  }, []);
+
   const wsList = workspaces.value;
   const currentId = activeWorkspaceId.value;
 

package/ui/demo.html

@@ -2723,6 +2723,55 @@
         return { data: STATE.executors.map((e) => ({ ...e, status: e.enabled ? 'active' : 'disabled' })) };
       if (route === '/api/telemetry/alerts')
         return { data: [] };
+      if (route === '/api/analytics/usage') {
+        const daysParam = Number(params.get('days') || '30');
+        // Build mock usage analytics from demo STATE
+        const now = Date.now();
+        const dayMs = 86400000;
+        const numDays = Math.min(daysParam || 30, 30);
+        // Generate synthetic daily trend data
+        const dates = [];
+        const agentDaily = { codex: [], copilot: [], claude: [] };
+        const skillDaily = { 'background-task-execution': [], 'pr-workflow': [], 'error-recovery': [] };
+        const mcpDaily = { execute_bash: [], read_file: [], write_file: [], list_directory: [] };
+        for (let i = numDays - 1; i >= 0; i--) {
+          const d = new Date(now - i * dayMs);
+          dates.push(d.toISOString().slice(0, 10));
+          const active = (i < (numDays * 0.6)) ? 1 : 0;
+          agentDaily.codex.push(active ? Math.floor(Math.random() * 8 + 2) : 0);
+          agentDaily.copilot.push(active ? Math.floor(Math.random() * 5 + 1) : 0);
+          agentDaily.claude.push(active ? Math.floor(Math.random() * 4) : 0);
+          skillDaily['background-task-execution'].push(active ? Math.floor(Math.random() * 6 + 1) : 0);
+          skillDaily['pr-workflow'].push(active ? Math.floor(Math.random() * 4) : 0);
+          skillDaily['error-recovery'].push(active ? Math.floor(Math.random() * 3) : 0);
+          mcpDaily.execute_bash.push(active ? Math.floor(Math.random() * 12 + 2) : 0);
+          mcpDaily.read_file.push(active ? Math.floor(Math.random() * 8 + 1) : 0);
+          mcpDaily.write_file.push(active ? Math.floor(Math.random() * 6) : 0);
+          mcpDaily.list_directory.push(active ? Math.floor(Math.random() * 5) : 0);
+        }
+        const sum = (arr) => arr.reduce((a, b) => a + b, 0);
+        const agentRuns = sum(agentDaily.codex) + sum(agentDaily.copilot) + sum(agentDaily.claude);
+        const skillInvocations = Object.values(skillDaily).reduce((t, a) => t + sum(a), 0);
+        const mcpToolCalls = Object.values(mcpDaily).reduce((t, a) => t + sum(a), 0);
+        return { ok: true, data: {
+          agentRuns, skillInvocations, mcpToolCalls,
+          avgPerDay: Math.round((agentRuns + skillInvocations + mcpToolCalls) / (numDays || 1)),
+          lastActiveAt: new Date(now - 7200000).toISOString(),
+          sinceAt: new Date(now - numDays * dayMs).toISOString(),
+          topAgents: [
+            { name: 'codex', count: sum(agentDaily.codex) },
+            { name: 'copilot', count: sum(agentDaily.copilot) },
+            { name: 'claude', count: sum(agentDaily.claude) },
+          ].filter(a => a.count > 0).sort((a, b) => b.count - a.count),
+          topSkills: Object.entries(skillDaily)
+            .map(([name, v]) => ({ name, count: sum(v) }))
+            .filter(s => s.count > 0).sort((a, b) => b.count - a.count),
+          topMcpTools: Object.entries(mcpDaily)
+            .map(([name, v]) => ({ name, count: sum(v) }))
+            .filter(t => t.count > 0).sort((a, b) => b.count - a.count),
+          trend: { dates, agents: agentDaily, skills: skillDaily, mcpTools: mcpDaily },
+        }};
+      }
       if (route === '/api/executor/pause') {
         STATE.paused = true; addLog('info', 'executor', 'Executor paused');
         return { ok: true, paused: true };
@@ -3543,6 +3592,67 @@
         return { data: buildProjectSnapshot() };
 
       // ── Voice ──
+      if (route === '/api/voice/agents' && method === 'GET') {
+        const fromLibrary = (STATE.libraryEntries || [])
+          .filter((entry) => {
+            if (!entry || entry.type !== 'agent') return false;
+            const id = String(entry.id || '').toLowerCase();
+            const tags = Array.isArray(entry.tags) ? entry.tags.map((t) => String(t || '').toLowerCase()) : [];
+            return id.includes('voice-agent') || tags.includes('voice') || tags.includes('audio-agent');
+          })
+          .map((entry) => ({
+            id: entry.id,
+            name: entry.name || entry.id,
+            description: entry.description || '',
+            tags: Array.isArray(entry.tags) ? entry.tags : [],
+            model: null,
+            voicePersona: String(entry.id || '').includes('female')
+              ? 'female'
+              : (String(entry.id || '').includes('male') ? 'male' : 'neutral'),
+            voiceInstructions: '',
+            skills: [],
+            promptOverride: null,
+          }));
+
+        const defaults = [
+          {
+            id: 'voice-agent-female',
+            name: 'Voice Agent (Female)',
+            description: 'Conversational voice specialist with concise guidance and call-friendly pacing.',
+            tags: ['voice', 'audio-agent', 'female', 'realtime'],
+            model: null,
+            voicePersona: 'female',
+            voiceInstructions: 'You are Nova, a concise and practical female voice agent.',
+            skills: ['concise-voice-guidance', 'conversation-memory'],
+            promptOverride: null,
+          },
+          {
+            id: 'voice-agent-male',
+            name: 'Voice Agent (Male)',
+            description: 'Operational voice specialist focused on diagnostics and execution.',
+            tags: ['voice', 'audio-agent', 'male', 'realtime'],
+            model: null,
+            voicePersona: 'male',
+            voiceInstructions: 'You are Atlas, a direct and execution-oriented male voice agent.',
+            skills: ['ops-diagnostics', 'task-execution'],
+            promptOverride: null,
+          },
+        ];
+
+        const seen = new Set();
+        const agents = [...fromLibrary, ...defaults].filter((agent) => {
+          const id = String(agent?.id || '').trim();
+          if (!id || seen.has(id)) return false;
+          seen.add(id);
+          return true;
+        });
+        return {
+          ok: true,
+          agents,
+          defaultAgentId: agents[0]?.id || 'voice-agent-female',
+        };
+      }
+
       if (route === '/api/voice/audio/respond' && method === 'POST') {
         const inputText = String(body?.inputText || body?.text || '').trim();
         if (!inputText) {

package/ui/modules/mic-track-registry.js

@@ -0,0 +1,83 @@
+/**
+ * mic-track-registry.js
+ *
+ * Tracks microphone input streams obtained via getUserMedia and provides a
+ * hard-stop primitive used by voice teardown to prevent lingering "mic in use"
+ * indicators after a call is closed.
+ */
+
+const trackedStreams = new Set();
+let patched = false;
+
+function isMediaStreamLike(stream) {
+  return Boolean(stream && typeof stream.getTracks === "function");
+}
+
+function getAudioTracks(stream) {
+  if (!isMediaStreamLike(stream)) return [];
+  try {
+    return (stream.getAudioTracks?.() || [])
+      .filter((track) => String(track?.kind || "").toLowerCase() === "audio");
+  } catch {
+    return [];
+  }
+}
+
+function pruneInactiveStreams() {
+  for (const stream of trackedStreams) {
+    const tracks = getAudioTracks(stream);
+    if (!tracks.length) {
+      trackedStreams.delete(stream);
+      continue;
+    }
+    const hasLive = tracks.some((track) => String(track?.readyState || "live").toLowerCase() !== "ended");
+    if (!hasLive) trackedStreams.delete(stream);
+  }
+}
+
+export function registerMicStream(stream) {
+  if (!isMediaStreamLike(stream)) return;
+  trackedStreams.add(stream);
+  const tracks = getAudioTracks(stream);
+  for (const track of tracks) {
+    try {
+      track.addEventListener?.("ended", () => {
+        pruneInactiveStreams();
+      }, { once: true });
+    } catch {
+      // no-op
+    }
+  }
+}
+
+export function ensureMicTrackingPatched() {
+  if (patched) return;
+  const mediaDevices = globalThis?.navigator?.mediaDevices;
+  if (!mediaDevices || typeof mediaDevices.getUserMedia !== "function") return;
+  const original = mediaDevices.getUserMedia.bind(mediaDevices);
+  mediaDevices.getUserMedia = async (...args) => {
+    const stream = await original(...args);
+    registerMicStream(stream);
+    return stream;
+  };
+  patched = true;
+}
+
+export function stopTrackedMicStreams() {
+  for (const stream of trackedStreams) {
+    const tracks = getAudioTracks(stream);
+    for (const track of tracks) {
+      try {
+        track.stop();
+      } catch {
+        // no-op
+      }
+    }
+  }
+  pruneInactiveStreams();
+}
+
+export function _resetMicTrackRegistryForTests() {
+  trackedStreams.clear();
+  patched = false;
+}

package/ui/modules/settings-schema.js

@@ -133,6 +133,9 @@ export const SETTINGS_SCHEMA = [
   { key: "AZURE_OPENAI_REALTIME_DEPLOYMENT", label: "Azure Deployment (legacy)",   category: "voice", type: "select", defaultVal: "gpt-audio-1.5", options: ["gpt-audio-1.5", "gpt-realtime-1.5", "gpt-4o-realtime-preview", "custom"], description: "Legacy fallback: Azure deployment name. Use the Voice Endpoints card above. GA models (gpt-realtime-1.5) auto-use /openai/v1/ paths." },
   { key: "VOICE_ID",                       label: "Voice",                      category: "voice", type: "select", defaultVal: "alloy", options: ["alloy", "ash", "ballad", "coral", "echo", "fable", "onyx", "nova", "sage", "shimmer", "verse"], description: "Voice personality for text-to-speech output." },
   { key: "VOICE_TURN_DETECTION",           label: "Turn Detection",             category: "voice", type: "select", defaultVal: "semantic_vad", options: ["server_vad", "semantic_vad", "none"], description: "How the model detects when you stop speaking. 'semantic_vad' is more intelligent but higher latency." },
+  { key: "VOICE_TRANSCRIPTION_ENABLED",    label: "Input Transcription Enabled", category: "voice", type: "boolean", defaultVal: true, description: "Enable per-turn input audio transcription for OpenAI-compatible realtime sessions." },
+  { key: "VOICE_TRANSCRIPTION_MODEL",      label: "Input Transcription Model",  category: "voice", type: "string", defaultVal: "gpt-4o-transcribe", description: "Model used for input audio transcription when transcription is enabled." },
+  { key: "VOICE_AZURE_TRANSCRIPTION_ENABLED", label: "Azure Input Transcription", category: "voice", type: "boolean", defaultVal: false, description: "Enable input transcription specifically for Azure realtime sessions. Disabled by default to avoid Azure per-item transcription failures." },
   { key: "VOICE_DELEGATE_EXECUTOR",        label: "Delegate Executor",          category: "voice", type: "select", defaultVal: "codex-sdk", options: ["codex-sdk", "copilot-sdk", "claude-sdk", "gemini-sdk", "opencode-sdk"], description: "Which agent executor voice tool calls delegate to for complex tasks." },
   { key: "VOICE_FALLBACK_MODE",            label: "Fallback Mode",              category: "voice", type: "select", defaultVal: "browser", options: ["browser", "disabled"], description: "When Tier 1 (Realtime API) is unavailable, use browser speech APIs as fallback." },