dsh-agent-voice 0.1.2 → 0.1.4

package/lib/client.js

@@ -39,7 +39,6 @@ window.__ModuleLoader__.load({
     function AgentVoice(props) {
       const [voices, setVoices] = React.useState(FALLBACK_VOICES);
       const [status, setStatus] = React.useState("");
-      const [lastPlayed, setLastPlayed] = React.useState(null);
 
       const action = React.useMemo(() => useAction(props.action), [props.action]);
       const cfgSnap = useSnapshot(props.config);
@@ -51,20 +50,21 @@ window.__ModuleLoader__.load({
       const announceAttention = cfg.announceAttention !== false;
       const volume = cfg.volume ?? 1.0;
 
-      // Live audio playback: watch the status namespace for a new announcement.
+      // Live status feed: show the last announcement (playback happens in the
+      // always-on overlay player, so it works on every tab).
       const statusSnap = useSnapshot(props.status);
       const statusJson = (statusSnap && statusSnap.value && statusSnap.value.json) || null;
+      const [lastText, setLastText] = React.useState(null);
+      const lastIdRef = React.useRef(null);
       React.useEffect(() => {
         if (!statusJson) return;
         let p = null;
         try { p = JSON.parse(statusJson); } catch { return; }
-        if (!p || !p.dataUrl || !p.id) return;
-        if (p.id === lastPlayed) return;
-        setLastPlayed(p.id);
-        const audio = new Audio(p.dataUrl);
-        audio.volume = Math.max(0, Math.min(1, Number(p.volume ?? 1)));
-        audio.play().catch(() => {});
-      }, [statusJson, lastPlayed]);
+        if (!p || !p.id) return;
+        if (p.id === lastIdRef.current) return;
+        lastIdRef.current = p.id;
+        setLastText(p.text || "");
+      }, [statusJson]);
 
       React.useEffect(() => {
         action({ action: "voices" }).then((r) => {
@@ -123,7 +123,10 @@ window.__ModuleLoader__.load({
           React.createElement("div", { className: "av-row" },
             React.createElement("button", { className: "av-btn av-primary", onClick: test }, "▶ Vyzkoušet hlas"),
             React.createElement("button", { className: "av-btn", onClick: attention }, "🔔 Výzva k pozornosti"),
-            status ? React.createElement("span", { className: "av-badge" }, status) : null)),
+            status ? React.createElement("span", { className: "av-badge" }, status) : null),
+          lastText ? React.createElement("div", { className: "av-row" },
+            React.createElement("span", { className: "av-tag" }, "Naposledy hlášeno:"),
+            React.createElement("span", { className: "av-sub" }, lastText)) : null),
 
         React.createElement("p", { className: "av-sub" },
           "Doporučené použití: agent zavolá dsh_agent_voice action='speak' po dokončení úkolu a action='attention', " +
@@ -139,6 +142,61 @@ window.__ModuleLoader__.load({
       const action = ctx.settingsScope.bind({ namespace: "agent-voice-action" });
       const config = ctx.settingsScope.bind({ namespace: "agent-voice" });
       const status = ctx.settingsScope.bind({ namespace: "agent-voice-status" });
+
+      // Global always-on audio player: plays announcements on EVERY tab, not
+      // only when the Agent Voice tab is active. Lives in the root-scoped
+      // overlay seat, so it stays mounted across view switches.
+      function Player(props) {
+        const statusSnap = useSnapshot(props.status);
+        const statusJson = (statusSnap && statusSnap.value && statusSnap.value.json) || null;
+        const cfgSnap = useSnapshot(props.config);
+        const cfg = (cfgSnap && cfgSnap.value) || {};
+        const [lastPlayed, setLastPlayed] = React.useState(null);
+        const [toast, setToast] = React.useState(null);
+        const toastTimer = React.useRef(null);
+
+        React.useEffect(() => {
+          if (!statusJson) return;
+          let p = null;
+          try { p = JSON.parse(statusJson); } catch { return; }
+          if (!p || !p.dataUrl || !p.id) return;
+          if (p.id === lastPlayed) return;
+          setLastPlayed(p.id);
+          const audio = new Audio(p.dataUrl);
+          audio.volume = Math.max(0, Math.min(1, Number(p.volume ?? 1)));
+          const kindLabel = p.kind === "attention" ? "🔔 Pozor: " : (p.kind === "task" ? "✅ " : "🔊 ");
+          setToast(kindLabel + p.text);
+          if (toastTimer.current) clearTimeout(toastTimer.current);
+          toastTimer.current = setTimeout(() => setToast(null), 6000);
+          audio.play().catch(() => {});
+        }, [statusJson, lastPlayed]);
+
+        // Disabled = silent (config read from the same source the host uses).
+        if (cfg.enabled === false) return null;
+
+        return React.createElement("div", {
+          style: {
+            position: "fixed", bottom: 16, right: 16, zIndex: 9999,
+            background: "var(--dsw-alias-bg-overlay)",
+            border: "1px solid var(--dsw-alias-border-l2)",
+            borderRadius: 12, padding: "10px 14px",
+            boxShadow: "0 8px 30px rgba(0,0,0,.35)",
+            color: "var(--dsw-alias-label-primary)",
+            font: "13px -apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif",
+            maxWidth: 320, display: toast ? "block" : "none",
+          },
+        }, toast);
+      }
+
+      ctx.slots.inject("shell.overlay", () => ctx.slots.register({
+        name: "shell.overlay",
+        id: "agent-voice-player",
+        order: 100,
+        label: () => "🔊 Agent Voice přehrávač",
+        inject: () => ({ config, status }),
+        children: {},
+      }, Player));
+
       ctx.slots.inject("conversation.view", () => ctx.slots.register({
         name: "conversation.view",
         id: "studio-agent-voice",

package/lib/index.js

@@ -111,6 +111,7 @@ function apply(ctx) {
   }), { base: {} });
 
   async function synthesize(text, voice, language) {
+    await ensureWorker();
     const res = await postWorker("/tts", { text, voice, language });
     const bytes = await readFile(res.audio);
     return "data:audio/wav;base64," + Buffer.from(bytes).toString("base64");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-agent-voice",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "description": "DeepSeek Harness plugin: gives the agent a voice — it reads completed tasks and attention requests aloud (never reasoning or interim reports) via local Piper / XTTS v2, with a settings panel and a curated set of voices.",
   "type": "module",
   "main": "lib/index.js",

package/worker/agent_voice_worker.log

@@ -1,3 +1,4 @@
 [2026-08-31 22:08:19] agent-voice worker listening on http://127.0.0.1:7863
 [2026-08-31 22:08:23] error: 'PiperVoice' object has no attribute 'synthesize_stream_raw'
 [2026-08-31 22:08:46] agent-voice worker listening on http://127.0.0.1:7863
+[2026-09-01 13:40:42] agent-voice worker listening on http://127.0.0.1:7863