@agentprojectcontext/apx 1.28.0 → 1.30.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@agentprojectcontext/apx",
-  "version": "1.28.0",
+  "version": "1.30.0",
   "description": "APX — unified CLI + daemon for the Agent Project Context (APC) standard.",
   "publishConfig": {
     "access": "public"

package/src/interfaces/desktop/renderer.js

@@ -36,6 +36,14 @@
   // doesn't talk into the dead gap before the recorder starts.
   let micReady = false;
 
+  // Dead-mic detection: track the loudest RMS seen this session. If it stays
+  // near zero for DEAD_MIC_MS the stream is silent (no permission / muted /
+  // wrong device) and we surface a notice instead of hanging in "listening".
+  let listenStartTs = 0;
+  let micPeakRms = 0;
+  const DEAD_MIC_MS = 3500;
+  const DEAD_MIC_RMS = 0.004;
+
   // Silence auto-send: once speech has been heard, SILENCE_MS of quiet
   // auto-commits the recording. RMS (time-domain) is the voice/silence gate.
   // Both are overridable from config.json (desktop.silence_ms / voice_rms).
@@ -389,17 +397,28 @@
         <div class="bubble-user">${escapeHtml(m.text)}${viaIcon}</div>
       `;
     } else {
-      t.innerHTML = `
+      // Consecutive agent messages (intro + post-tool answer …) read as one
+      // continued reply: only the FIRST shows the "Roby" header — the rest skip
+      // it so a tool turn isn't a stack of repeated "Roby" labels. A new header
+      // only appears when something (a user message) breaks the run.
+      const idx = messages.indexOf(m);
+      const prevMsg = idx > 0 ? messages[idx - 1] : null;
+      const agentCont = !!(prevMsg && prevMsg.role === "agent");
+      if (agentCont) t.classList.add("cont");
+      const header = agentCont ? "" : `
         <div class="role agent">
           <span class="ava sa"><img src="assets/superagent.png" alt=""/></span>
           <span class="who">${escapeHtml(agentName)}</span>
           <span class="time">${m.t || ""}</span>
-        </div>
-        <div class="msg-agent">${formatWordsHtml(m.text)}</div>
-        ${m.audio ? "" /* scrubber added separately */ : ""}
+        </div>`;
+      // Copy is an inline icon at the end of the text, hover-only, so it never
+      // reserves an empty row. Regenerate lives in turn-actions and CSS shows it
+      // only on the last turn.
+      t.innerHTML = `
+        ${header}
+        <div class="msg-agent">${formatWordsHtml(m.text)}<button class="btn-copy" aria-label="Copiar" title="Copiar">${ICON.copy()}</button></div>
         <div class="turn-actions">
           <button class="chip btn-regen">${ICON.refresh()} Regenerar</button>
-          <button class="chip btn-copy">${ICON.copy()} Copiar</button>
         </div>
       `;
       if (m.audio && m.dur) {
@@ -409,13 +428,13 @@
         actions.insertAdjacentHTML("beforebegin", scrubberHtml);
         wireScrubber(t, m);
       }
-      // copy
+      // copy (inline icon → swaps to a check briefly)
       t.querySelector(".btn-copy")?.addEventListener("click", (e) => {
         navigator.clipboard?.writeText(m.text).catch(() => {});
         const btn = e.currentTarget;
         btn.classList.add("done");
-        btn.innerHTML = `${ICON.check()} Copiado`;
-        setTimeout(() => { btn.classList.remove("done"); btn.innerHTML = `${ICON.copy()} Copiar`; }, 1400);
+        btn.innerHTML = ICON.check();
+        setTimeout(() => { btn.classList.remove("done"); btn.innerHTML = ICON.copy(); }, 1400);
       });
       // regen: only the LAST agent turn can be regenerated. Past turns
       // can't because we'd have to re-issue the user prompt that came right
@@ -749,6 +768,8 @@
     micReady = false;      // show "Cargando…" until the recorder is actually live
     speechSeen = false;
     lastVoiceTs = 0;
+    listenStartTs = 0;
+    micPeakRms = 0;
     pausePreviewed = false;
     reuseLiveOnStop = false;
     livePromise = null;
@@ -869,13 +890,42 @@
       // won't auto-send (speechSeen gates that).
       micReady = true;
       lastVoiceTs = Date.now();
+      listenStartTs = Date.now();
+      micPeakRms = 0;
       if (mode === "listening") render();
     } catch (e) {
+      // getUserMedia failed → classify and tell the user, instead of silently
+      // bailing to idle (which looks like "it just doesn't work").
       console.error("desktop renderer: mic error", e);
+      micReady = false;
       mode = "idle";
+      const name = e?.name || "";
+      let notice;
+      if (name === "NotAllowedError" || name === "SecurityError" || /permission|denied/i.test(e?.message || "")) {
+        notice = "Roby no tiene permiso para el micrófono. Activalo en Ajustes del sistema › Privacidad y seguridad › Micrófono, y volvé a intentar.";
+      } else if (name === "NotFoundError" || name === "DevicesNotFoundError" || name === "OverconstrainedError") {
+        notice = "No encontré ningún micrófono conectado. Revisá el dispositivo de entrada y reintentá.";
+      } else {
+        notice = "No pude abrir el micrófono (" + (name || e?.message || "error") + "). Reintentá o revisá los permisos.";
+      }
+      showMicNotice(notice);
       render();
     }
   }
+  // Visual-only system notice in the conversation (not a message, not history).
+  function showMicNotice(text) {
+    ensureConv();
+    if ($convScroll) {
+      // Replace any prior notice so they don't stack.
+      $convScroll.querySelector(".sys-notice")?.remove();
+      const el = document.createElement("div");
+      el.className = "sys-notice";
+      el.innerHTML = `<span class="ic">${ICON.mic()}</span><span>${escapeHtml(text)}</span>`;
+      $convScroll.appendChild(el);
+      scrollConvToBottom();
+    }
+    requestWindowResize();
+  }
   function stopMic() {
     try { mediaRecorder?.stop(); } catch {}
     try { audioStream?.getTracks().forEach((t) => t.stop()); } catch {}
@@ -915,6 +965,19 @@
         }
         const rms = Math.sqrt(sumSq / timeData.length);
         const now = Date.now();
+        if (rms > micPeakRms) micPeakRms = rms;
+        // Dead-mic guard: a real mic always has a noise floor (rms ≳ 0.005). If
+        // after DEAD_MIC_MS the signal is essentially flat zero, the stream is
+        // muted / wrong device / OS-blocked — tell the user instead of sitting
+        // in "listening" forever waiting for speech that can't arrive.
+        if (!speechSeen && listenStartTs && now - listenStartTs > DEAD_MIC_MS && micPeakRms < DEAD_MIC_RMS) {
+          waveRaf = null;
+          stopMic();
+          mode = "idle";
+          showMicNotice("No me está llegando audio del micrófono. Revisá que Roby tenga permiso (Ajustes del sistema › Privacidad y seguridad › Micrófono) y que esté seleccionado el micrófono correcto.");
+          render();
+          return;
+        }
         if (rms > VOICE_RMS) {
           speechSeen = true;
           lastVoiceTs = now;

package/src/interfaces/desktop/style.css

@@ -328,19 +328,30 @@ button { font-family: inherit; }
 .wavebar i.cur { transform: scaleY(1.25); }
 .audio .dur { font-size: 11px; color: var(--ink-3); font-variant-numeric: tabular-nums; min-width: 30px; text-align: right; }
 
-/* per-turn actions */
+/* Continued agent messages (no repeated "Roby" header) hug the previous one. */
+.turn.cont { padding-top: 0; }
+.turn.cont .msg-agent { margin-top: 1px; }
+
+/* Inline copy icon at the end of an agent message — hover-only, so it never
+   reserves an empty row of vertical space. */
+.msg-agent .btn-copy {
+  display: inline-flex; align-items: center; vertical-align: -3px;
+  margin-left: 6px; padding: 1px; border: none; background: transparent;
+  cursor: pointer; color: var(--ink-3); opacity: 0; transition: opacity .15s ease, color .15s ease;
+}
+.turn:hover .msg-agent .btn-copy { opacity: .55; }
+.msg-agent .btn-copy:hover { opacity: 1; color: var(--ink); }
+.msg-agent .btn-copy.done { opacity: 1; color: oklch(0.6 0.15 150); }
+
+/* per-turn actions — only Regenerate now, and only on the LAST agent turn.
+   Regenerating a past turn would replay the most-recent user prompt (not the
+   one that produced that reply) and silently break the flow, so it's hidden
+   everywhere else and takes no space. */
 .turn-actions {
-  margin: 7px 0 0 24px; display: flex; gap: 4px;
+  margin: 7px 0 0 24px; display: none; gap: 4px;
   opacity: 0; transition: opacity .2s ease;
 }
-.turn:hover .turn-actions, .turn.last .turn-actions { opacity: 1; }
-
-/* Regenerate is only meaningful on the LAST agent turn — regenerating a
-   past one would replay the most-recent user prompt (not the one that
-   produced this reply) and silently break the conversation flow. Copy
-   stays available on every turn so users can grab old replies. */
-.turn .btn-regen { display: none; }
-.turn.last .btn-regen { display: inline-flex; }
+.turn.last .turn-actions { display: flex; opacity: 1; }
 .chip {
   display: inline-flex; align-items: center; gap: 5px; cursor: pointer;
   padding: 4px 8px; border-radius: 9px; border: 1px solid var(--glass-hairline);
@@ -350,6 +361,16 @@ button { font-family: inherit; }
 .chip:hover { background: var(--glass-hairline); color: var(--ink); border-color: transparent; }
 .chip.done  { color: oklch(0.6 0.15 150); }
 
+/* system notice (e.g. no mic permission / dead mic) — not a chat message */
+.sys-notice {
+  margin: 8px 0 4px; padding: 9px 11px; border-radius: 11px;
+  display: flex; align-items: flex-start; gap: 8px;
+  background: color-mix(in oklch, oklch(0.72 0.17 55) 12%, transparent);
+  border: 1px solid color-mix(in oklch, oklch(0.72 0.17 55) 35%, transparent);
+  color: var(--ink); font-size: 12.5px; line-height: 1.4;
+}
+.sys-notice .ic { flex: none; color: oklch(0.62 0.18 45); display: inline-flex; margin-top: 1px; }
+
 /* tool pill (when the agent runs a tool) */
 .tool-pill {
   display: inline-flex; align-items: center; gap: 6px;