omnius 1.0.425 → 1.0.427

package/dist/index.js

@@ -29383,9 +29383,19 @@ function _extractSponsorFromMeshMessage(msg) {
   if (payload.type === 'sponsor.announce') return payload;
   return null;
 }
+var _profilePublishInFlight = false;
+var _profilePublishLastAt = 0;
 async function _publishAgentProfileSnapshot() {
   try {
     if (!connected || !nexus.network || !nexus.network.dht || !nexus.network.dht.registry) return;
+    // DHT publishes on congested networks time out (observed as repeated
+    // [dht:registry] TimeoutError storms). Promise.race below does NOT cancel
+    // the underlying query, so overlapping calls stack concurrent DHT walks
+    // and burn CPU. Single-flight + a 60s floor keeps one publish at a time.
+    var _ppNow = Date.now();
+    if (_profilePublishInFlight || _ppNow - _profilePublishLastAt < 60000) return;
+    _profilePublishInFlight = true;
+    _profilePublishLastAt = _ppNow;
     var node = nexus.network.node;
     var caps = typeof nexus.getRegisteredCapabilities === 'function' ? nexus.getRegisteredCapabilities() : [];
     var profile = {
@@ -29407,6 +29417,8 @@ async function _publishAgentProfileSnapshot() {
     ]);
   } catch (err) {
     dlog('profile publish failed: ' + (err.message || err));
+  } finally {
+    _profilePublishInFlight = false;
   }
 }
 async function _publishCapabilityRecord(name, details) {
@@ -31286,9 +31298,11 @@ async function handleCmd(cmd) {
 
               // Wait for input data to arrive (invoke protocol sends data after accept)
               var waitMs = 0;
+              // 50ms poll (was a 10ms spin) — bounded to 5s; input latency is
+              // network-dominated so the coarser tick is invisible.
               while (!inputDone && dataChunks.length === 0 && waitMs < 5000) {
-                await new Promise(function(r) { setTimeout(r, 10); });
-                waitMs += 10;
+                await new Promise(function(r) { setTimeout(r, 50); });
+                waitMs += 50;
               }
               prompt = dataChunks.join('');
               dlog('expose: received ' + dataChunks.length + ' chunks, prompt_len=' + prompt.length + ' inputDone=' + inputDone);
@@ -32228,8 +32242,11 @@ async function handleCmd(cmd) {
   }
 }
 
-// Command polling loop — check for cmd.json every 50ms (fast IPC)
-// fs.existsSync + readFileSync costs ~0.01ms per call, negligible CPU at 50ms interval.
+// Command intake: fs.watch on the nexus dir delivers cmd.json changes
+// instantly (registered below); this interval is only a fallback for
+// platforms/filesystems where fs.watch drops events. 50ms polling here was
+// measured pinning CPU on deployed boxes — 1s fallback costs nothing since
+// the watcher provides the low-latency path.
 let lastCmdId = '';
 function checkCmd() {
   try {
@@ -32243,7 +32260,7 @@ function checkCmd() {
     });
   } catch {}
 }
-setInterval(checkCmd, 50);
+setInterval(checkCmd, 1000);
 // Also watch for cmd.json changes for instant notification (best-effort)
 try {
   fsWatch(nexusDir, { persistent: false }, function(evType, filename) {
@@ -32629,18 +32646,35 @@ process.on('unhandledRejection', (reason) => {
     } catch {}
     try {
       if (nexus.nats && typeof nexus.nats.subscribe === 'function') {
+        var _peersFileLastWrite = 0;
         nexus.nats.subscribe(function(announcement) {
           if (!announcement || !announcement.peerId) return;
           if (announcement.peerId === nexus.peerId) return; // skip self
-          discoveredPeers[announcement.peerId] = {
+          // Peers re-announce every 10-30s. Persisting the full JSON map and
+          // logging on EVERY announcement turned steady-state discovery into
+          // constant file I/O + log churn. Only react to NEW peers or
+          // material changes; lastSeen bumps persist on a 30s throttle.
+          var prev = discoveredPeers[announcement.peerId];
+          var next = {
             peerId: announcement.peerId,
             agentName: announcement.agentName || '',
             capabilities: announcement.capabilities || [],
             multiaddrs: announcement.multiaddrs || [],
             lastSeen: Date.now(),
           };
-          try { writeFileSync(discoveredPeersFile, JSON.stringify(discoveredPeers, null, 2)); } catch {}
-          dlog('NATS peer discovered: ' + String(announcement.peerId).slice(0, 20) + ' caps=' + (announcement.capabilities || []).length);
+          var changed = !prev
+            || prev.agentName !== next.agentName
+            || JSON.stringify(prev.capabilities) !== JSON.stringify(next.capabilities)
+            || JSON.stringify(prev.multiaddrs) !== JSON.stringify(next.multiaddrs);
+          discoveredPeers[announcement.peerId] = next;
+          var nowMs = Date.now();
+          if (changed || nowMs - _peersFileLastWrite >= 30000) {
+            _peersFileLastWrite = nowMs;
+            try { writeFileSync(discoveredPeersFile, JSON.stringify(discoveredPeers, null, 2)); } catch {}
+          }
+          if (changed) {
+            dlog('NATS peer discovered: ' + String(announcement.peerId).slice(0, 20) + ' caps=' + (announcement.capabilities || []).length);
+          }
         });
         dlog('NATS peer discovery subscription active');
       }
@@ -603585,7 +603619,9 @@ var init_listen = __esm({
                 case "transcript":
                   this.emit("transcript", {
                     text: evt.text,
-                    isFinal: evt.isFinal ?? false
+                    isFinal: evt.isFinal ?? false,
+                    consensus: typeof evt.consensus === "boolean" ? evt.consensus : void 0,
+                    doa: typeof evt.doa === "number" ? evt.doa : void 0
                   });
                   break;
                 case "consensus_rejected":
@@ -603982,7 +604018,7 @@ var init_listen = __esm({
               this.lastTranscriptTime = Date.now();
               this.pendingText = evt.text.trim();
               updateListenLiveState({ lastTranscript: this.pendingText, lastTranscriptAt: this.lastTranscriptTime });
-              this.emit("transcript", this.pendingText, evt.isFinal);
+              this.emit("transcript", this.pendingText, evt.isFinal, { consensus: evt.consensus, doa: evt.doa });
               if (this.config.mode === "auto") this.resetSilenceTimer();
             });
             fallback.on("error", (err) => {
@@ -617195,6 +617231,25 @@ function rms(arr) {
   return Math.sqrt(sum / arr.length);
 }
 
+// Linear resampler: browsers frequently IGNORE the requested AudioContext
+// sample rate and run at the hardware rate (typically 48kHz). Sending those
+// frames as-if-16kHz made the audio unpack 3x slow/pitch-dropped server-side,
+// wrecking ASR. Always resample the actual context rate down to 16kHz.
+function resampleTo16k(input, srcRate) {
+  if (srcRate === 16000) return input;
+  const ratio = srcRate / 16000;
+  const outLen = Math.floor(input.length / ratio);
+  const out = new Float32Array(outLen);
+  for (let i = 0; i < outLen; i++) {
+    const pos = i * ratio;
+    const i0 = Math.floor(pos);
+    const i1 = Math.min(input.length - 1, i0 + 1);
+    const frac = pos - i0;
+    out[i] = input[i0] * (1 - frac) + input[i1] * frac;
+  }
+  return out;
+}
+
 async function toggleMic() {
   if (micActive) { stopMic(); return; }
   try {
@@ -617203,17 +617258,19 @@ async function toggleMic() {
       audio: { sampleRate: 16000, channelCount: 1, echoCancellation: true, noiseSuppression: true }
     });
     const source = micCtx.createMediaStreamSource(micStream);
+    const actualRate = micCtx.sampleRate; // may be 48000 regardless of the request
     scriptProcessor = micCtx.createScriptProcessor(4096, 1, 1);
     scriptProcessor.onaudioprocess = (e) => {
       if (!micActive || !ws || ws.readyState !== 1) return;
-      const input = e.inputBuffer.getChannelData(0);
+      const raw = e.inputBuffer.getChannelData(0);
+      const input = resampleTo16k(raw, actualRate);
       const int16 = new Int16Array(input.length);
       for (let i = 0; i < input.length; i++) {
         const s = Math.max(-1, Math.min(1, input[i]));
         int16[i] = s < 0 ? s * 0x8000 : s * 0x7FFF;
       }
       ws.send(int16.buffer);
-      micLevel = Math.min(1, rms(input) * 5);
+      micLevel = Math.min(1, rms(raw) * 5);
     };
     source.connect(scriptProcessor);
     scriptProcessor.connect(micCtx.destination);
@@ -617222,7 +617279,8 @@ async function toggleMic() {
     micBtn.textContent = 'stop';
     micBtn.classList.add('active');
     hintEl.textContent = 'listening...';
-    ws.send(JSON.stringify({ type: 'mic_start', username: 'web-user' }));
+    // Declare the post-resample rate (and the raw context rate for debugging)
+    ws.send(JSON.stringify({ type: 'mic_start', username: 'web-user', sampleRate: 16000, contextRate: actualRate }));
   } catch (err) {
     hintEl.textContent = 'mic error: ' + err.message;
   }
@@ -617595,6 +617653,23 @@ connectPP();
 </body>
 </html>`;
 }
+function resamplePcm16Mono(pcm, srcRate, dstRate) {
+  if (srcRate === dstRate || srcRate <= 0 || dstRate <= 0) return pcm;
+  const srcSamples = Math.floor(pcm.length / 2);
+  if (srcSamples === 0) return pcm;
+  const ratio = srcRate / dstRate;
+  const outSamples = Math.max(1, Math.floor(srcSamples / ratio));
+  const out = Buffer.allocUnsafe(outSamples * 2);
+  for (let i2 = 0; i2 < outSamples; i2++) {
+    const pos = i2 * ratio;
+    const i0 = Math.min(srcSamples - 1, Math.floor(pos));
+    const i1 = Math.min(srcSamples - 1, i0 + 1);
+    const frac = pos - i0;
+    const sample = pcm.readInt16LE(i0 * 2) * (1 - frac) + pcm.readInt16LE(i1 * 2) * frac;
+    out.writeInt16LE(Math.max(-32768, Math.min(32767, Math.round(sample))), i2 * 2);
+  }
+  return out;
+}
 function renderVoiceSessionStart(tunnelUrl) {
   renderBoxedBlock({
     title: "☁ Live Voice Session",
@@ -617902,17 +617977,26 @@ var init_voice_session = __esm({
             clearInterval(keepaliveInterval);
           }
         }, 1e4);
+        let clientPcmRate = 16e3;
         ws.on("message", (data, isBinary) => {
           if (isBinary) {
             if (!this.ttsSpeaking) {
-              this.onUserAudio?.(Buffer.isBuffer(data) ? data : Buffer.from(data), clientId);
+              let pcm = Buffer.isBuffer(data) ? data : Buffer.from(data);
+              if (clientPcmRate !== 16e3 && clientPcmRate > 0) {
+                pcm = resamplePcm16Mono(pcm, clientPcmRate, 16e3);
+              }
+              this.onUserAudio?.(pcm, clientId);
             }
           } else {
             try {
               const msg = JSON.parse(data.toString());
-              if (msg.type === "mic_start" && msg.username) {
-                const entry = this.state.connectedUsers.get(clientId);
-                if (entry) entry.username = msg.username;
+              if (msg.type === "mic_start") {
+                if (msg.username) {
+                  const entry = this.state.connectedUsers.get(clientId);
+                  if (entry) entry.username = msg.username;
+                }
+                const declared = Number(msg.sampleRate);
+                clientPcmRate = Number.isFinite(declared) && declared >= 8e3 && declared <= 192e3 ? declared : 16e3;
               }
             } catch {
             }
@@ -697480,6 +697564,9 @@ Rules:
       // VAD segment capture
       captureBuffer = "";
       captureStartTime = 0;
+      consensusMetadataSeen = false;
+      consensusText = null;
+      consensusSignalScore = null;
       silenceTimer = null;
       maxSegmentTimer = null;
       lastSignalScore = null;
@@ -697570,18 +697657,25 @@ Rules:
           let isFinal;
           let snr;
           let confidence2;
+          let consensus;
           if (typeof args[0] === "object" && args[0] !== null) {
             const evt = args[0];
             text2 = evt.text ?? "";
             isFinal = evt.isFinal ?? false;
             snr = evt.snr;
             confidence2 = evt.confidence;
+            consensus = typeof evt.consensus === "boolean" ? evt.consensus : void 0;
           } else {
             text2 = String(args[0] ?? "");
             isFinal = Boolean(args[1]);
+            const meta = args[2];
+            if (meta && typeof meta === "object") {
+              const m2 = meta;
+              if (typeof m2.consensus === "boolean") consensus = m2.consensus;
+            }
           }
           if (!text2.trim()) return;
-          this.handleTranscript(text2.trim(), isFinal, snr, confidence2);
+          this.handleTranscript(text2.trim(), isFinal, snr, confidence2, consensus);
         };
         this._onError = (err) => {
           const msg = err instanceof Error ? err.message : String(err);
@@ -697648,7 +697742,7 @@ Rules:
       // ---------------------------------------------------------------------------
       // Transcript handling — VAD-style segment capture (Voryn pattern)
       // ---------------------------------------------------------------------------
-      handleTranscript(text2, isFinal, snr, confidence2) {
+      handleTranscript(text2, isFinal, snr, confidence2, consensus) {
         if (!this.active) return;
         if (this._state !== "LISTENING" && this._state !== "CAPTURING") {
           return;
@@ -697661,7 +697755,7 @@ Rules:
         }
         if (this._state === "LISTENING") {
           this.setState("CAPTURING");
-          this.captureBuffer = "";
+          this.resetCaptureState();
           this.captureStartTime = Date.now();
           this.maxSegmentTimer = setTimeout(() => {
             if (this._state === "CAPTURING") {
@@ -697669,8 +697763,18 @@ Rules:
             }
           }, MAX_SEGMENT_MS);
         }
-        this.captureBuffer = text2;
-        this.lastSignalScore = typeof snr === "number" && !Number.isNaN(snr) ? clamp0116(snr) : computeSignalFromText(text2, confidence2);
+        const signalScore = typeof snr === "number" && !Number.isNaN(snr) ? clamp0116(snr) : computeSignalFromText(text2, confidence2);
+        this.lastSignalScore = signalScore;
+        if (consensus !== void 0) {
+          this.consensusMetadataSeen = true;
+          if (consensus === true) {
+            this.consensusText = text2;
+            this.consensusSignalScore = signalScore;
+            this.captureBuffer = text2;
+          }
+        } else {
+          this.captureBuffer = text2;
+        }
         this.emit("snr", { score: this.lastSignalScore });
         this.onPartialTranscript(text2);
         if (this.silenceTimer) clearTimeout(this.silenceTimer);
@@ -697681,11 +697785,22 @@ Rules:
           }
         }, waitMs);
       }
+      resetCaptureState() {
+        this.captureBuffer = "";
+        this.captureStartTime = 0;
+        this.lastSignalScore = null;
+        this.consensusMetadataSeen = false;
+        this.consensusText = null;
+        this.consensusSignalScore = null;
+      }
       // ---------------------------------------------------------------------------
       // Segment finalization → Transcribing → Thinking → Speaking
       // ---------------------------------------------------------------------------
       finalizeSegment() {
-        const text2 = this.captureBuffer.trim();
+        const rawText = this.captureBuffer.trim();
+        const requiresConsensus = this.consensusMetadataSeen;
+        const text2 = requiresConsensus ? (this.consensusText ?? "").trim() : rawText;
+        const signalScore = requiresConsensus ? this.consensusSignalScore ?? this.lastSignalScore : this.lastSignalScore;
         if (this.silenceTimer) {
           clearTimeout(this.silenceTimer);
           this.silenceTimer = null;
@@ -697694,18 +697809,20 @@ Rules:
           clearTimeout(this.maxSegmentTimer);
           this.maxSegmentTimer = null;
         }
-        this.captureBuffer = "";
+        this.resetCaptureState();
         if (!text2) {
+          if (requiresConsensus && rawText) {
+            if (this.debugSnr) this.onStatus(`Ignoring utterance without dual-ASR consensus: ${truncateForLog(rawText, 48)}`);
+            this.emit("consensusFiltered", { text: rawText });
+          }
           this.setState("LISTENING");
           return;
         }
-        const score = Math.min(this.lastSignalScore ?? 1, computeSignalFromText(text2));
+        const score = Math.min(signalScore ?? 1, computeSignalFromText(text2));
         if (score < MIN_SIGNAL_SCORE || NOISE_ONLY_RE.test(text2)) {
           if (this.debugSnr) this.onStatus(`Ignoring low-signal utterance (SNR:${score.toFixed(2)}): ${truncateForLog(text2, 48)}`);
           this.emit("snrFiltered", { score, text: text2 });
           this.setState("LISTENING");
-          this.captureBuffer = "";
-          this.lastSignalScore = null;
           return;
         }
         this.setState("TRANSCRIBING");

package/dist/scripts/live-whisper.py

@@ -76,10 +76,12 @@ def emit_error(msg: str):
     emit({"type": "error", "message": msg})
 
 
-def emit_transcript(text: str, is_final: bool = False, doa=None):
+def emit_transcript(text: str, is_final: bool = False, doa=None, consensus=None):
     event = {"type": "transcript", "text": text, "isFinal": is_final}
     if doa is not None:
         event["doa"] = doa
+    if consensus is not None:
+        event["consensus"] = bool(consensus)
     emit(event)
 
 # ---------------------------------------------------------------------------
@@ -441,7 +443,7 @@ def main():
             result = run_transcribe(model, amplify(samples))
             text = _segment_filtered_text(result)
             if text:
-                emit_transcript(text, is_final=False)
+                emit_transcript(text, is_final=False, consensus=False)
         except Exception as e:
             emit_error(f"Transcription error: {e}")
 
@@ -481,7 +483,14 @@ def main():
             if text == last_final_text:
                 return
             last_final_text = text
-            emit_transcript(text, is_final=True, doa=tuning.doa() if tuning else None)
+            # consensus=True only when BOTH models ran and converged — the
+            # downstream voicechat gate requires convergence for pass-through.
+            emit_transcript(
+                text,
+                is_final=True,
+                doa=tuning.doa() if tuning else None,
+                consensus=consensus_model is not None,
+            )
         except Exception as e:
             emit_error(f"Transcription error: {e}")
 

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.425",
+  "version": "1.0.427",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.425",
+      "version": "1.0.427",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.425",
+  "version": "1.0.427",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",