agent.libx.js 0.92.8 → 0.92.9

package/cli/cli.ts

@@ -990,14 +990,21 @@ async function repl(args: Args, ai: ChatLike, cfg: Partial<AgentConfig>, cwd: st
   const duplexAsk = async (call: ToolUse): Promise<{ decision: 'allow' | 'deny' }> => {
     if (args.voice && dx) {
       const hint = summarizeCall(call.name, call.args).slice(0, 80);
+      // 'menu' mode: approve like a normal session — suspend the editor, pop the picker.
+      if ((cfg as any).voiceAskUi === 'menu') {
+        editorRef?.suspend();
+        const v = await selectMenu(process.stderr, { title: `? background worker asks to run ${call.name} ${hint}`, items: [{ label: 'Allow', value: 'y' }, { label: 'Deny', value: 'n' }], current: 'n' });
+        editorRef?.resume();
+        editorRef?.redrawNow();
+        return { decision: v === 'y' ? 'allow' : 'deny' };
+      }
       // NB: perm asks are keyed perm-N (PermissionPolicy.ask carries no task identity), so a
       // cancelled task can't clean its parked perm question — bounded by askTimeoutMs → deny.
+      // (Chrome prints once via the task_ask notify below — no extra line here.)
       const id = `perm-${++permSeq}`;
-      err('\r\x1b[0J' + yellow(`  ? worker asks to run ${call.name}`) + dim(` ${hint} — say yes/no (or it auto-denies)\n`));
-      editorRef?.redrawNow();
-      const a = await dx.parkQuestion(id, `Permission: may the background worker run ${call.name}${hint ? ` (${hint})` : ''}? Yes or no.`);
+      const a = await dx.parkQuestion(id, `Permission: may the background worker run ${call.name}${hint ? ` (${hint})` : ''}? Answer yes or no (you can also type it).`);
       const allow = /^\s*(y(es|ep|eah)?|sure|ok(ay)?|allow|go|approved?|do it)\b/i.test(a);
-      err('\r\x1b[0J' + dim(`  ${allow ? '✓ allowed' : '⊘ denied'} ${call.name} (${a.trim() || 'no answer'})\n`));
+      err('\r\x1b[0J' + (allow ? green(`  ✓ allowed ${call.name}`) : yellow(`  ⊘ denied ${call.name}`)) + dim(` (${a.trim() || 'no answer'})\n`));
       editorRef?.redrawNow();
       return { decision: allow ? 'allow' : 'deny' };
     }
@@ -1060,7 +1067,8 @@ async function repl(args: Args, ai: ChatLike, cfg: Partial<AgentConfig>, cwd: st
         // residue glued to 'agentx › '). task_done returned above; everything else falls through.
         if (typeof e.kind === 'string' && e.kind.startsWith('task_')) {
           spinner.stop();
-          err('\r\x1b[0J' + dim(`  · ${e.message}\n`));
+          // asks are decisions, not chatter — make them stand out (the voice also speaks them)
+          err('\r\x1b[0J' + (e.kind === 'task_ask' ? yellow(`  ? ${e.message} — answer by voice or type yes/no\n`) : dim(`  · ${e.message}\n`)));
           editorRef?.redrawNow();
           return;
         }

package/dist/cli.js

@@ -3589,6 +3589,7 @@ ${recent}` : brief;
     let steps = 0;
     let inflight = null;
     const due = () => {
+      if (this.pendingAsks.size) return void 0;
       const rec = this.tasks.get(id);
       return rec && rec.status === "running" && Date.now() - lastAt >= this.options.progressIntervalMs ? rec : void 0;
     };
@@ -3872,7 +3873,7 @@ var VoiceEngineOptions = class {
    *  vocabulary) resume from the precise sample and are dropped. false disables. */
   overlapPause = true;
   /** sustained overlap ≥ this → cede the turn */
-  overlapSustainMs = 350;
+  overlapSustainMs = 450;
   /** quiet for this long while paused → resume, drop the interjection */
   overlapResumeMs = 700;
   /** energy floor for "overlap candidate" — must sit ABOVE typical room ambient (~110 rms measured;
@@ -3914,6 +3915,8 @@ var VoiceEngine = class {
   // loud chunks since pause (sustain must be real sound, not two clicks)
   overlapLastLoudAt = 0;
   // continuity guard: a gap re-arms the onset (sparse noise ≠ sustained speech)
+  loudTimes = [];
+  // recent loud-chunk timestamps (sliding onset window)
   resumeTimer = null;
   constructor(options) {
     this.options = { ...new VoiceEngineOptions(), ...options };
@@ -4133,15 +4136,18 @@ var VoiceEngine = class {
     if (rms < o.overlapRms) return;
     const t = now();
     if (!this.pausedAt) {
-      this.overlapLoud = t - this.overlapLastLoudAt <= 60 ? this.overlapLoud + 1 : 1;
-      this.overlapLastLoudAt = t;
-      if (this.overlapLoud < 3) return;
+      this.loudTimes = this.loudTimes.filter((x) => t - x < 400);
+      this.loudTimes.push(t);
+      if (this.loudTimes.length < 2) return;
+      this.loudTimes = [];
       this.pausedAt = t;
+      this.overlapLoud = 2;
+      this.overlapLastLoudAt = t;
       this.player.pause();
       this.armResume();
       return;
     }
-    if (t - this.overlapLastLoudAt > 300) {
+    if (t - this.overlapLastLoudAt > 450) {
       this.pausedAt = t;
       this.overlapLoud = 1;
       this.overlapLastLoudAt = t;
@@ -4150,7 +4156,7 @@ var VoiceEngine = class {
     }
     this.overlapLastLoudAt = t;
     this.overlapLoud++;
-    if (t - this.pausedAt >= o.overlapSustainMs && this.overlapLoud >= 4) {
+    if (t - this.pausedAt >= o.overlapSustainMs && this.overlapLoud >= 3) {
       const phase = this.ctxOpen ? "speaking" : "drain";
       this.interrupt();
       this.options.onBargeIn(phase);
@@ -4174,6 +4180,7 @@ var VoiceEngine = class {
     if (this.pausedAt && resume) this.player.resume?.();
     this.pausedAt = 0;
     this.overlapLoud = 0;
+    this.loudTimes = [];
   }
   /** energy two-stage barge-in (heuristic tier only): spike over echo baseline → pause + confirm via STT */
   handleLevel(rms) {
@@ -7966,13 +7973,17 @@ async function repl(args, ai, cfg, cwd) {
   const duplexAsk = async (call) => {
     if (args.voice && dx) {
       const hint = summarizeCall(call.name, call.args).slice(0, 80);
+      if (cfg.voiceAskUi === "menu") {
+        editorRef?.suspend();
+        const v = await selectMenu(process.stderr, { title: `? background worker asks to run ${call.name} ${hint}`, items: [{ label: "Allow", value: "y" }, { label: "Deny", value: "n" }], current: "n" });
+        editorRef?.resume();
+        editorRef?.redrawNow();
+        return { decision: v === "y" ? "allow" : "deny" };
+      }
       const id = `perm-${++permSeq}`;
-      err("\r\x1B[0J" + yellow(`  ? worker asks to run ${call.name}`) + dim(` ${hint} \u2014 say yes/no (or it auto-denies)
-`));
-      editorRef?.redrawNow();
-      const a = await dx.parkQuestion(id, `Permission: may the background worker run ${call.name}${hint ? ` (${hint})` : ""}? Yes or no.`);
+      const a = await dx.parkQuestion(id, `Permission: may the background worker run ${call.name}${hint ? ` (${hint})` : ""}? Answer yes or no (you can also type it).`);
       const allow = /^\s*(y(es|ep|eah)?|sure|ok(ay)?|allow|go|approved?|do it)\b/i.test(a);
-      err("\r\x1B[0J" + dim(`  ${allow ? "\u2713 allowed" : "\u2298 denied"} ${call.name} (${a.trim() || "no answer"})
+      err("\r\x1B[0J" + (allow ? green(`  \u2713 allowed ${call.name}`) : yellow(`  \u2298 denied ${call.name}`)) + dim(` (${a.trim() || "no answer"})
 `));
       editorRef?.redrawNow();
       return { decision: allow ? "allow" : "deny" };
@@ -8022,8 +8033,9 @@ async function repl(args, ai, cfg, cwd) {
         }
         if (typeof e.kind === "string" && e.kind.startsWith("task_")) {
           spinner.stop();
-          err("\r\x1B[0J" + dim(`  \xB7 ${e.message}
-`));
+          err("\r\x1B[0J" + (e.kind === "task_ask" ? yellow(`  ? ${e.message} \u2014 answer by voice or type yes/no
+`) : dim(`  \xB7 ${e.message}
+`)));
           editorRef?.redrawNow();
           return;
         }