omnius 1.0.438 → 1.0.440

package/dist/index.js

@@ -603492,8 +603492,8 @@ function purgeVenvTorch(pip) {
   );
   return `${purge.stdout || ""}${purge.stderr || ""}`;
 }
-function jetsonTorchInstallSource() {
-  const wheel = discoverJetsonTorchWheel();
+function jetsonTorchInstallSource(py = getVenvPython()) {
+  const wheel = discoverJetsonTorchWheel(py);
   if (wheel) {
     return {
       source: wheel,
@@ -603526,11 +603526,16 @@ function torchCudaProbe(py = getVenvPython()) {
     { encoding: "utf8", timeout: 3e4 }
   );
   const log22 = `${probe.stdout || ""}${probe.stderr || ""}`.trim();
-  return { ok: probe.status === 0, log: log22 };
+  try {
+    return { ok: probe.status === 0, log: log22, json: JSON.parse(probe.stdout || "{}") };
+  } catch {
+    return { ok: probe.status === 0, log: log22 };
+  }
 }
 function ensureEmbedDeps() {
   const venv = getVenvDir();
   let log22 = "";
+  const diagnostics = [];
   try {
     mkdirSync67(venv, { recursive: true });
   } catch {
@@ -603542,6 +603547,7 @@ function ensureEmbedDeps() {
   }
   if (isJetsonHost2() && cudaHardwarePresent() && !asrCpuAllowed()) {
     enableVenvSystemSitePackages(venv);
+    diagnostics.push(`Jetson ASR CUDA repair: python=${getVenvPython()} abi=${pythonAbiTag()} cuda=${detectCudaToolkitVersion() || "unknown"} jp_dirs=${jetsonPytorchDirs().join(",")}`);
   }
   if (venvExisted && !(process.env["OMNIUS_INSTALL_FULL_EMBED_DEPS"] === "1")) {
     const check = spawnSync8(
@@ -603557,6 +603563,7 @@ function ensureEmbedDeps() {
       if (cudaProbe.ok) {
         return { ok: true, log: `embed deps already satisfied; torch CUDA OK ${cudaProbe.log}` };
       }
+      diagnostics.push(`Initial venv torch CUDA failed: ${cudaProbe.log}`);
       log22 += `embed deps import OK but torch CUDA probe failed; repairing ASR torch. ${cudaProbe.log}
 `;
     }
@@ -603568,8 +603575,10 @@ function ensureEmbedDeps() {
   if (cudaHardwarePresent() && !asrCpuAllowed()) {
     const before = torchCudaProbe();
     torchOk = before.ok;
+    if (!torchOk) diagnostics.push(`Pre-repair torch probe: ${before.log}`);
     if (!torchOk && isJetsonHost2()) {
       const systemTorch = torchCudaProbe("python3");
+      diagnostics.push(`System python torch probe: ${systemTorch.log || "unavailable"}`);
       if (systemTorch.ok) {
         log22 += `Venv torch CUDA probe failed, but system Jetson torch CUDA works; enabling system site packages and removing venv-local torch.
 venv=${before.log}
@@ -603597,19 +603606,25 @@ system=${systemTorch.log}
       );
       log22 += torchDeps.stdout + torchDeps.stderr;
       const torchSource = jetsonTorchInstallSource();
+      diagnostics.push(`Jetson torch install source: ${torchSource.source || "not found"}`);
       log22 += `Torch CUDA probe failed on Jetson; installing JetPack-compatible PyTorch from ${torchSource.source}
 ${before.log}
 `;
-      const torchInstall = spawnSync8(
-        pip,
-        torchSource.args,
-        { encoding: "utf8", timeout: 9e5 }
-      );
-      log22 += torchInstall.stdout + torchInstall.stderr;
-      const after = torchCudaProbe();
-      torchOk = after.ok;
-      log22 += `Torch CUDA probe after Jetson install: ${after.log}
+      if (!torchSource.source) {
+        torchOk = false;
+      } else {
+        const torchInstall = spawnSync8(
+          pip,
+          torchSource.args,
+          { encoding: "utf8", timeout: 9e5 }
+        );
+        log22 += torchInstall.stdout + torchInstall.stderr;
+        const after = torchCudaProbe();
+        torchOk = after.ok;
+        diagnostics.push(`Post Jetson torch install probe: ${after.log}`);
+        log22 += `Torch CUDA probe after Jetson install: ${after.log}
 `;
+      }
     } else if (!torchOk) {
       log22 += `Torch CUDA probe failed on CUDA host; not allowing ASR CPU fallback. ${before.log}
 `;
@@ -603649,20 +603664,27 @@ ${before.log}
     if (!torchOk) {
       log22 += `Torch CUDA probe still failed after ASR dependency install. ${finalTorch.log}
 `;
+      diagnostics.push(`Post ASR dependency torch probe: ${finalTorch.log}`);
       if (isJetsonHost2()) {
         log22 += "Retrying Jetson torch repair after ASR dependency install to remove any incompatible torch resolver side effects.\n";
         log22 += purgeVenvTorch(pip);
         const torchSource = jetsonTorchInstallSource();
-        const retry = spawnSync8(
-          pip,
-          torchSource.args,
-          { encoding: "utf8", timeout: 9e5 }
-        );
-        log22 += retry.stdout + retry.stderr;
-        const repaired = torchCudaProbe();
-        torchOk = repaired.ok;
-        log22 += `Torch CUDA probe after final Jetson repair from ${torchSource.source}: ${repaired.log}
+        diagnostics.push(`Final Jetson torch repair source: ${torchSource.source || "not found"}`);
+        if (!torchSource.source) {
+          torchOk = false;
+        } else {
+          const retry = spawnSync8(
+            pip,
+            torchSource.args,
+            { encoding: "utf8", timeout: 9e5 }
+          );
+          log22 += retry.stdout + retry.stderr;
+          const repaired = torchCudaProbe();
+          torchOk = repaired.ok;
+          diagnostics.push(`Final repaired torch probe: ${repaired.log}`);
+          log22 += `Torch CUDA probe after final Jetson repair from ${torchSource.source}: ${repaired.log}
 `;
+        }
       }
     }
   }
@@ -603682,6 +603704,21 @@ ${before.log}
     fullOk = full.status === 0;
   }
   const ok3 = asrDepsOk && up.status === 0 && fullOk && torchOk;
+  if (!ok3) {
+    const compact4 = [
+      "ASR dependency setup failed.",
+      ...diagnostics,
+      `pip_base_ok=${up.status === 0}`,
+      `asr_deps_ok=${asrDepsOk}`,
+      `full_embed_deps_ok=${fullOk}`,
+      `torch_cuda_ok=${torchOk}`,
+      "Set OMNIUS_JETSON_TORCH_WHEEL=/path/or/url/to/torch-*-linux_aarch64.whl to override auto-discovery."
+    ].join("\n");
+    return { ok: ok3, log: `${compact4}
+
+--- raw pip tail ---
+${log22.slice(-2e3)}` };
+  }
   return { ok: ok3, log: log22 };
 }
 function runEmbedImage(input) {
@@ -603790,6 +603827,56 @@ function updateListenLiveState(patch) {
 function getListenLiveState() {
   return listenLiveState;
 }
+function clamp0110(value2) {
+  if (!Number.isFinite(value2)) return 0;
+  return Math.max(0, Math.min(1, value2));
+}
+function micColor(text2, level) {
+  const idx = Math.round(clamp0110(level) * (MIC_GRADIENT_256.length - 1));
+  const code8 = MIC_GRADIENT_256[idx] ?? 21;
+  return `\x1B[38;5;${code8}m${text2}\x1B[0m`;
+}
+function micLevelChar(level) {
+  const idx = Math.round(clamp0110(level) * (MIC_LEVEL_CHARS.length - 1));
+  return MIC_LEVEL_CHARS[idx] ?? "▁";
+}
+function micSpectrumChar(level) {
+  const idx = Math.round(clamp0110(level) * (MIC_SPECTRUM_CHARS.length - 1));
+  return MIC_SPECTRUM_CHARS[idx] ?? " ";
+}
+function renderMicWaveform(bucketSquares, bucketCounts, peak) {
+  return bucketSquares.map((sum, idx) => {
+    const count = Math.max(1, bucketCounts[idx] ?? 1);
+    const level = Math.sqrt(sum / count);
+    const normalized = clamp0110(level / Math.max(0.02, peak || 0.02));
+    return micColor(micLevelChar(normalized), normalized);
+  }).join("");
+}
+function renderMicSpectrum(samples) {
+  if (samples.length < 16) return "";
+  const n2 = Math.min(256, samples.length);
+  const src2 = samples.slice(0, n2);
+  const bands = [];
+  const maxK = Math.max(2, Math.floor(n2 / 2) - 1);
+  for (let band = 0; band < MIC_SPECTRUM_BANDS; band++) {
+    const t2 = MIC_SPECTRUM_BANDS <= 1 ? 0 : band / (MIC_SPECTRUM_BANDS - 1);
+    const k = Math.max(1, Math.min(maxK, Math.round(1 + (Math.exp(t2 * Math.log(maxK)) - 1))));
+    let re = 0;
+    let im = 0;
+    for (let i2 = 0; i2 < n2; i2++) {
+      const w = 0.5 - 0.5 * Math.cos(2 * Math.PI * i2 / Math.max(1, n2 - 1));
+      const angle = 2 * Math.PI * k * i2 / n2;
+      const sample = (src2[i2] ?? 0) * w;
+      re += sample * Math.cos(angle);
+      im -= sample * Math.sin(angle);
+    }
+    const mag = Math.sqrt(re * re + im * im) / n2;
+    const db = 20 * Math.log10(Math.max(1e-7, mag));
+    const normalized = clamp0110((db + 75) / 75);
+    bands.push(micColor(micSpectrumChar(normalized), normalized));
+  }
+  return bands.join("");
+}
 async function resolveMicSource() {
   const configured = String(process.env["OMNIUS_MIC_DEVICE"] ?? "").trim();
   if (configured.startsWith("hw:") || configured.startsWith("plughw:")) {
@@ -604228,7 +604315,7 @@ function getListenEngine(config) {
   }
   return _engine;
 }
-var MANAGED_TRANSCRIBE_CLI_DIR2, AUDIO_EXTENSIONS, VIDEO_EXTENSIONS2, listenLiveState, lastHardwareVadAt, WhisperFallbackTranscriber, ListenEngine, _bgInstallPromise, _managedTranscribeCliInstallPromise, _engine;
+var MANAGED_TRANSCRIBE_CLI_DIR2, AUDIO_EXTENSIONS, VIDEO_EXTENSIONS2, listenLiveState, lastHardwareVadAt, MIC_WAVEFORM_BINS, MIC_SPECTRUM_BANDS, MIC_WATERFALL_ROWS, MIC_LEVEL_CHARS, MIC_SPECTRUM_CHARS, MIC_GRADIENT_256, WhisperFallbackTranscriber, ListenEngine, _bgInstallPromise, _managedTranscribeCliInstallPromise, _engine;
 var init_listen = __esm({
   "packages/cli/src/tui/listen.ts"() {
     "use strict";
@@ -604265,6 +604352,9 @@ var init_listen = __esm({
       micLevelDb: null,
       noiseFloorDb: null,
       speechActive: false,
+      waveform: "",
+      spectrum: "",
+      waterfall: [],
       doa: null,
       lastTranscript: "",
       lastTranscriptAt: 0,
@@ -604274,6 +604364,33 @@ var init_listen = __esm({
       updatedAt: 0
     };
     lastHardwareVadAt = 0;
+    MIC_WAVEFORM_BINS = 56;
+    MIC_SPECTRUM_BANDS = 28;
+    MIC_WATERFALL_ROWS = 8;
+    MIC_LEVEL_CHARS = "▁▂▃▄▅▆▇█";
+    MIC_SPECTRUM_CHARS = " .:-=+*#%@";
+    MIC_GRADIENT_256 = [
+      21,
+      27,
+      33,
+      39,
+      45,
+      51,
+      50,
+      49,
+      48,
+      47,
+      82,
+      118,
+      154,
+      190,
+      226,
+      220,
+      214,
+      208,
+      202,
+      196
+    ];
     WhisperFallbackTranscriber = class extends EventEmitter6 {
       constructor(model, scriptPath2) {
         super();
@@ -604285,6 +604402,8 @@ var init_listen = __esm({
       process = null;
       _ready = false;
       registeredBrokerName = null;
+      lastVadSpeech = null;
+      lastVisibleVadAt = 0;
       get ready() {
         return this._ready;
       }
@@ -604298,7 +604417,7 @@ var init_listen = __esm({
             const res = ensure();
             if (res?.log) this.emit("status", res.log.slice(-500));
             if (res && res.ok === false) {
-              throw new Error(`ASR Python dependency setup failed: ${String(res.log ?? "").slice(-1200)}`);
+              throw new Error(String(res.log ?? "ASR Python dependency setup failed").slice(0, 2400));
             }
           }
           if (typeof getPy === "function") {
@@ -604318,7 +604437,7 @@ var init_listen = __esm({
         const consensusEnv = String(process.env["OMNIUS_ASR_CONSENSUS"] ?? "").trim().toLowerCase();
         const consensusModel = consensusEnv === "0" || consensusEnv === "off" || consensusEnv === "false" ? "off" : consensusEnv || defaultConsensusModel(this.model);
         const consensusThreshold = String(process.env["OMNIUS_ASR_CONSENSUS_THRESHOLD"] ?? "0.45");
-        const silenceMs = String(process.env["OMNIUS_ASR_SILENCE_MS"] ?? "800");
+        const silenceMs = String(process.env["OMNIUS_ASR_SILENCE_MS"] ?? "3000");
         this.process = spawn29(
           pyPath,
           [
@@ -604386,9 +604505,15 @@ var init_listen = __esm({
                 case "vad": {
                   lastHardwareVadAt = Date.now();
                   const doa = Number(evt.doa);
-                  this.emit("status", `Whisper VAD: ${evt.speech === true ? "speech" : "silence"}`);
+                  const speech = evt.speech === true;
+                  const now2 = Date.now();
+                  if (process.env["OMNIUS_ASR_VERBOSE_VAD"] === "1" && (speech !== this.lastVadSpeech || now2 - this.lastVisibleVadAt > 1e4)) {
+                    this.emit("status", `Whisper VAD: ${speech ? "speech" : "silence"}`);
+                    this.lastVisibleVadAt = now2;
+                  }
+                  this.lastVadSpeech = speech;
                   updateListenLiveState({
-                    speechActive: evt.speech === true,
+                    speechActive: speech,
                     doa: Number.isFinite(doa) ? doa : null
                   });
                   break;
@@ -604557,6 +604682,7 @@ var init_listen = __esm({
       micNoiseFloorDb = null;
       micSpeechActive = false;
       lastLevelEmit = 0;
+      micWaterfall = [];
       /**
        * Spawn the mic capture process and wire PCM piping + level metering.
        * Shared by start() and resume() so both paths meter identically.
@@ -604596,9 +604722,20 @@ var init_listen = __esm({
         let sumSquares = 0;
         const stride = samples > 4096 ? 4 : 1;
         let counted = 0;
+        const bucketSquares = new Array(MIC_WAVEFORM_BINS).fill(0);
+        const bucketCounts = new Array(MIC_WAVEFORM_BINS).fill(0);
+        const spectrumSamples = [];
+        const spectrumDecimate = Math.max(1, Math.floor(samples / 256));
         for (let i2 = 0; i2 + 1 < chunk.length; i2 += 2 * stride) {
           const v = chunk.readInt16LE(i2) / 32768;
           sumSquares += v * v;
+          const sampleIndex = Math.floor(i2 / 2);
+          const bucket = Math.min(MIC_WAVEFORM_BINS - 1, Math.floor(sampleIndex / Math.max(1, samples) * MIC_WAVEFORM_BINS));
+          bucketSquares[bucket] += v * v;
+          bucketCounts[bucket] += 1;
+          if (sampleIndex % spectrumDecimate === 0 && spectrumSamples.length < 256) {
+            spectrumSamples.push(v);
+          }
           counted++;
         }
         if (counted === 0) return;
@@ -604613,18 +604750,33 @@ var init_listen = __esm({
         const speech = hardwareFresh ? listenLiveState.speechActive : this.micLevelDb > Math.max(-58, floor + 6);
         const changed = speech !== this.micSpeechActive;
         this.micSpeechActive = speech;
+        const peak = Math.sqrt(Math.max(...bucketSquares.map((sum, idx) => sum / Math.max(1, bucketCounts[idx] ?? 1)), 0));
+        const waveform = renderMicWaveform(bucketSquares, bucketCounts, peak);
+        const spectrum = renderMicSpectrum(spectrumSamples);
+        if (spectrum) {
+          this.micWaterfall.push(spectrum);
+          if (this.micWaterfall.length > MIC_WATERFALL_ROWS) {
+            this.micWaterfall = this.micWaterfall.slice(-MIC_WATERFALL_ROWS);
+          }
+        }
         const now2 = Date.now();
         if (changed || now2 - this.lastLevelEmit >= 300) {
           this.lastLevelEmit = now2;
           updateListenLiveState({
             micLevelDb: Number(this.micLevelDb.toFixed(1)),
             noiseFloorDb: Number(floor.toFixed(1)),
-            speechActive: speech
+            speechActive: speech,
+            waveform,
+            spectrum,
+            waterfall: [...this.micWaterfall]
           });
           this.emit("level", {
             levelDb: this.micLevelDb,
             noiseFloorDb: floor,
-            speechActive: speech
+            speechActive: speech,
+            waveform,
+            spectrum,
+            waterfall: [...this.micWaterfall]
           });
         }
       }
@@ -604684,12 +604836,16 @@ var init_listen = __esm({
        */
       async start() {
         if (this.active) return "Already listening.";
+        this.micWaterfall = [];
         updateListenLiveState({
           phase: "bootstrapping",
           model: this.config.model,
           backend: "",
           lastStatus: "starting ASR backend...",
-          speechActive: false
+          speechActive: false,
+          waveform: "",
+          spectrum: "",
+          waterfall: []
         });
         const micCmd = await findMicCaptureCommand();
         if (!micCmd) {
@@ -604896,7 +605052,8 @@ transcribe-cli error: ${transcribeCliError}` : "";
         this.active = false;
         this.owners.clear();
         this.blinkState = false;
-        updateListenLiveState({ phase: "idle", speechActive: false, micLevelDb: null, lastStatus: "stopped" });
+        this.micWaterfall = [];
+        updateListenLiveState({ phase: "idle", speechActive: false, micLevelDb: null, waveform: "", spectrum: "", waterfall: [], lastStatus: "stopped" });
         if (this.blinkTimer) {
           clearInterval(this.blinkTimer);
           this.blinkTimer = null;
@@ -604957,7 +605114,7 @@ transcribe-cli error: ${transcribeCliError}` : "";
           }
           this.micProcess = null;
         }
-        updateListenLiveState({ phase: "paused", speechActive: false, lastStatus: "mic paused (TTS speaking)" });
+        updateListenLiveState({ phase: "paused", speechActive: false, waveform: "", spectrum: "", lastStatus: "mic paused (TTS speaking)" });
         this.emit("paused");
         this.emit("recording", false);
       }
@@ -606981,6 +607138,7 @@ function formatLiveDashboardFromSnapshot(snapshot, opts = {}) {
     };
     const asrMicBits = asr.micLevelDb !== null ? `  mic=${asr.micLevelDb.toFixed(1)}dB${asr.speechActive ? "  ●speech" : ""}${asr.doa !== null ? `  dir=${asr.doa}°` : ""}` : "";
     const asrLine = asr.phase !== "idle" ? `│  ├─ whisper: ${[asr.backend, asr.model].filter(Boolean).join(" ") || "starting"} — ${asrPhaseLabel[asr.phase] ?? asr.phase}${asrMicBits}` : "";
+    const asrStateLine = asr.phase !== "idle" ? `│  ├─ state: phase=${asr.phase} backend=${asr.backend || "unknown"} model=${asr.model || "unknown"} vad=${asr.speechActive ? "speech" : "quiet"} floor=${asr.noiseFloorDb !== null ? `${asr.noiseFloorDb.toFixed(1)}dB` : "n/a"}` : "";
     const asrFinalAge = asr.lastTranscriptAt > 0 ? Math.max(0, Math.round((now2 - asr.lastTranscriptAt) / 1e3)) : null;
     const asrPartialAge = asr.lastPartialTranscriptAt > 0 ? Math.max(0, Math.round((now2 - asr.lastPartialTranscriptAt) / 1e3)) : null;
     const showPartial = asr.lastPartialTranscript && (!asr.lastTranscriptAt || asr.lastPartialTranscriptAt >= asr.lastTranscriptAt);
@@ -606990,14 +607148,23 @@ function formatLiveDashboardFromSnapshot(snapshot, opts = {}) {
     const audioBits = [
       `├─ input: ${audioSnapshot?.input || snapshot.config.selectedAudioInput || "none"}`,
       asrLine,
+      asrStateLine,
       asrStatusLine,
       asrHeardLine,
+      asr.waveform ? `│  ├─ waveform: ${asr.waveform}` : "",
+      asr.spectrum ? `│  ├─ spectrum: ${asr.spectrum} ${dashboardDim("low→high freq")}` : "",
       audioSnapshot?.activity ? `│  ├─ activity: ${audioSnapshot.activity.waveform}  rms=${audioSnapshot.activity.rmsDb.toFixed(1)}dB  active=${Math.round(audioSnapshot.activity.activeRatio * 100)}%` : "",
       audioError ? `│  ├─ error: ${trimOneLine(audioError, 140)}` : "",
       transcript ? `│  ├─ asr${audioSnapshot?.asrBackend ? ` (${audioSnapshot.asrBackend})` : ""}: ${trimOneLine(transcript, 140)}` : "",
       audioSnapshot?.intake ? `│  ├─ intake: ${audioSnapshot.intake.intendedForAgent ? "for-agent" : "ambient"}  conf=${audioSnapshot.intake.confidence.toFixed(2)}  action=${audioSnapshot.intake.action}` : ""
     ].filter(Boolean);
-    for (const item of audioBits.slice(0, 9)) pushDashboardWrapped(lines, item, width);
+    for (const item of audioBits.slice(0, 11)) pushDashboardWrapped(lines, item, width);
+    if (asr.waterfall.length > 0) {
+      lines.push(`│${truncateAnsiCell(`  ├─ speech waterfall ${dashboardDim("blue quiet → red loud")}`, width - 2)}│`);
+      for (const row of asr.waterfall.slice(-8)) {
+        lines.push(`│${truncateAnsiCell(`  │  ${row}`, width - 2)}│`);
+      }
+    }
     if (coloredSoundSummary) {
       lines.push(`│${truncateAnsiCell(`  └─ ${coloredSoundSummary}`, width - 2)}│`);
     }
@@ -668510,7 +668677,7 @@ function normalizePersonName(name10) {
 function personKey(name10) {
   return `person:${normalizePersonName(name10)}`;
 }
-function clamp0110(value2, fallback = 0) {
+function clamp0111(value2, fallback = 0) {
   if (typeof value2 !== "number" || !Number.isFinite(value2)) return fallback;
   return Math.max(0, Math.min(1, value2));
 }
@@ -668527,7 +668694,7 @@ function parseStructuredIdentifyResult(result) {
   const matches = faces.filter((face) => face["identified"] === true && typeof face["name"] === "string" && String(face["name"]).trim()).map((face) => ({
     name: String(face["name"]).trim(),
     personId: typeof face["person_id"] === "string" ? face["person_id"] : void 0,
-    confidence: clamp0110(face["confidence"], 0),
+    confidence: clamp0111(face["confidence"], 0),
     margin: typeof face["margin"] === "number" ? face["margin"] : void 0,
     bbox: Array.isArray(face["bbox"]) ? face["bbox"].map((n2) => Number(n2)).filter(Number.isFinite) : void 0
   }));
@@ -668610,7 +668777,7 @@ function activePendingVisualIdentities(store2, scope, sessionId, limit = 3) {
       pendingId,
       name: name10,
       relation: String(meta["relation"] || "depicts"),
-      confidence: clamp0110(meta["confidence"], 0.92),
+      confidence: clamp0111(meta["confidence"], 0.92),
       note: typeof meta["note"] === "string" ? meta["note"] : void 0,
       episodeId: ep.id,
       createdAt: ep.timestamp,
@@ -668719,7 +668886,7 @@ function stageVisualIdentityAssertion(options2) {
           target: "next_visual_media",
           name: name10,
           relation: options2.relation || "depicts",
-          confidence: clamp0110(options2.confidence, 0.92),
+          confidence: clamp0111(options2.confidence, 0.92),
           note: options2.note,
           createdAt: Date.now(),
           expiresAt
@@ -668727,7 +668894,7 @@ function stageVisualIdentityAssertion(options2) {
         identityAssertions: [{
           name: name10,
           relation: options2.relation || "named_as",
-          confidence: clamp0110(options2.confidence, 0.92),
+          confidence: clamp0111(options2.confidence, 0.92),
           assertedBy: options2.sender,
           note: options2.note || "Explicit user-provided identity staged for the next visual media."
         }],
@@ -668758,7 +668925,7 @@ function stageVisualIdentityAssertion(options2) {
           target: "next_visual_media",
           name: name10,
           relation: options2.relation || "depicts",
-          confidence: clamp0110(options2.confidence, 0.92),
+          confidence: clamp0111(options2.confidence, 0.92),
           note: options2.note,
           createdAt: Date.now(),
           expiresAt
@@ -668767,7 +668934,7 @@ function stageVisualIdentityAssertion(options2) {
       identityAssertions: [{
         name: name10,
         relation: options2.relation || "named_as",
-        confidence: clamp0110(options2.confidence, 0.92),
+        confidence: clamp0111(options2.confidence, 0.92),
         assertedBy: options2.sender,
         note: options2.note || "Explicit user-provided identity staged for the next visual media."
       }]
@@ -677563,7 +677730,7 @@ ${result.output}`,
 });
 
 // packages/cli/src/tui/stimulation.ts
-function clamp0111(value2) {
+function clamp0112(value2) {
   return Math.max(0, Math.min(1, value2));
 }
 function cloneState(state) {
@@ -677619,7 +677786,7 @@ var init_stimulation = __esm({
           ...DEFAULT_STATE,
           ...state,
           phase: normalizePhase(state.phase) ?? DEFAULT_STATE.phase,
-          attention: clamp0111(Number.isFinite(state.attention) ? Number(state.attention) : DEFAULT_STATE.attention),
+          attention: clamp0112(Number.isFinite(state.attention) ? Number(state.attention) : DEFAULT_STATE.attention),
           updatedAtMs: Number.isFinite(state.updatedAtMs) ? Number(state.updatedAtMs) : now2,
           lastStimulusAtMs: Number.isFinite(state.lastStimulusAtMs) ? Number(state.lastStimulusAtMs) : now2,
           messagesSinceAnalysis: Math.max(0, Math.floor(Number(state.messagesSinceAnalysis ?? 0))),
@@ -677660,11 +677827,11 @@ var init_stimulation = __esm({
       applyAgentDecision(channelId, decision2, nowMs = Date.now()) {
         const state = this.stateFor(channelId, nowMs);
         if (Number.isFinite(decision2.attentionScore)) {
-          state.attention = clamp0111(Number(decision2.attentionScore));
+          state.attention = clamp0112(Number(decision2.attentionScore));
         } else if (Number.isFinite(decision2.attentionDelta)) {
-          state.attention = clamp0111(state.attention + Number(decision2.attentionDelta));
+          state.attention = clamp0112(state.attention + Number(decision2.attentionDelta));
         } else {
-          state.attention = clamp0111(state.attention + (decision2.shouldReply ? 0.22 : -0.1));
+          state.attention = clamp0112(state.attention + (decision2.shouldReply ? 0.22 : -0.1));
         }
         if (decision2.phase) {
           state.attention = Math.max(state.attention, PHASE_FLOORS[decision2.phase]);
@@ -677720,7 +677887,7 @@ var init_stimulation = __esm({
 });
 
 // packages/cli/src/tui/pid-controller.ts
-function clamp0112(x) {
+function clamp0113(x) {
   if (!Number.isFinite(x)) return 0;
   if (x < 0) return 0;
   if (x > 1) return 1;
@@ -677784,7 +677951,7 @@ var init_pid_controller = __esm({
         const dt = st.lastSampleAt > 0 ? now2 - st.lastSampleAt : 1e3;
         const derivative = dt > 0 ? (error - st.lastError) / dt : 0;
         const u = st.config.kp * error + st.config.ki * st.integral + st.config.kd * derivative;
-        st.output = clamp0112(st.output + u);
+        st.output = clamp0113(st.output + u);
         st.lastError = error;
         st.lastSampleAt = now2;
         st.samples += 1;
@@ -678313,7 +678480,7 @@ function buildReplyOpportunities(input, openQuestions) {
 function daydreamOpportunityId(input, trigger) {
   return createHash39("sha1").update(`${input.sessionKey}:${input.generatedAtMs}:${trigger}`).digest("hex").slice(0, 16);
 }
-function clamp0113(value2) {
+function clamp0114(value2) {
   if (!Number.isFinite(value2)) return 0;
   return Math.max(0, Math.min(1, value2));
 }
@@ -678330,7 +678497,7 @@ function pushStimulationSignal(signals, signal, source, weight) {
   signals.push({
     signal: cleanSignal,
     source: cleanSource,
-    weight: clamp0113(weight)
+    weight: clamp0114(weight)
   });
 }
 function buildMetaAnalysisSignals(input) {
@@ -678460,7 +678627,7 @@ function buildCuriosityThreads(input, openQuestions, stimulationSignals) {
       question: text2.endsWith("?") || text2.endsWith("？") ? text2 : `What should be learned or clarified from: ${text2 || entry.mediaSummary || "recent media"}?`,
       rationale: "Human curiosity, uncertainty, or multimodal content makes this a useful idle exploration target.",
       sourceMessages: messageId,
-      intensity: clamp0113(0.5 + replyBoost + mediaBoost + questionBoost)
+      intensity: clamp0114(0.5 + replyBoost + mediaBoost + questionBoost)
     });
   }
   for (const question of openQuestions.slice(-4)) {
@@ -678483,7 +678650,7 @@ function buildCuriosityThreads(input, openQuestions, stimulationSignals) {
       question: `Is there a useful clarification or memory consolidation around ${strongest.source}?`,
       rationale: "Strongest stimulation signal can seed a low-intrusion reflection target.",
       sourceMessages: [],
-      intensity: clamp0113(strongest.weight * 0.72)
+      intensity: clamp0114(strongest.weight * 0.72)
     });
   }
   return threads.sort((a2, b) => b.intensity - a2.intensity).slice(0, 8);
@@ -678563,7 +678730,7 @@ function buildOutreachPlans(input, curiosityThreads) {
       purpose: "Continue the public thread only when the live model judges that the group would benefit from a concise follow-up.",
       draftIntent: "Ask one concrete clarification, offer one useful synthesis, or stay silent if the room has moved on.",
       gate: "model_decision",
-      confidence: clamp0113(thread.intensity * 0.86)
+      confidence: clamp0114(thread.intensity * 0.86)
     });
     const participant = participantForThread(input, thread);
     if (!participant) continue;
@@ -678575,7 +678742,7 @@ function buildOutreachPlans(input, curiosityThreads) {
       purpose: "Offer a one-to-one follow-up only if private contact is allowed and the issue is personal, unresolved, or better handled outside the group.",
       draftIntent: "Reference the public thread briefly, ask permission to continue privately, and do not reveal hidden meta-analysis.",
       gate: "admin_review",
-      confidence: clamp0113(thread.intensity * 0.58)
+      confidence: clamp0114(thread.intensity * 0.58)
     });
   }
   return plans.slice(0, 8);
@@ -679907,7 +680074,7 @@ function numberOr(value2, fallback) {
 function isNumber(value2) {
   return typeof value2 === "number" && Number.isFinite(value2);
 }
-function clamp0114(value2) {
+function clamp0115(value2) {
   return Math.max(0, Math.min(1, Number.isFinite(value2) ? value2 : 0));
 }
 function iso(ts) {
@@ -680091,8 +680258,8 @@ function normalizeRelationship(raw) {
     kind: value2.kind,
     fromKey: String(value2.fromKey),
     toKey: String(value2.toKey),
-    confidence: clamp0114(numberOr(value2.confidence, 0)),
-    weight: clamp0114(numberOr(value2.weight, 0)),
+    confidence: clamp0115(numberOr(value2.confidence, 0)),
+    weight: clamp0115(numberOr(value2.weight, 0)),
     firstSeenAt: numberOr(value2.firstSeenAt, Date.now()),
     lastSeenAt: numberOr(value2.lastSeenAt, Date.now()),
     evidenceMessageIds: Array.isArray(value2.evidenceMessageIds) ? value2.evidenceMessageIds.filter(isNumber).slice(-40) : [],
@@ -680111,7 +680278,7 @@ function normalizePreferences(raw) {
       if (!evidence || typeof evidence !== "object") continue;
       out[actorKey][key] = {
         value: Math.max(-1, Math.min(1, numberOr(evidence.value, 0))),
-        confidence: clamp0114(numberOr(evidence.confidence, 0)),
+        confidence: clamp0115(numberOr(evidence.confidence, 0)),
         updatedAt: numberOr(evidence.updatedAt, Date.now()),
         evidenceMessageIds: Array.isArray(evidence.evidenceMessageIds) ? evidence.evidenceMessageIds.filter(isNumber).slice(-12) : [],
         note: compactOptional(evidence.note, 220)
@@ -680125,7 +680292,7 @@ function normalizePreferences(raw) {
         out[actorKey].replyMode = {
           mode,
           scope: normalizeReplyPreferenceScope(record["scope"]),
-          confidence: clamp0114(numberOr(record["confidence"], 0.8)),
+          confidence: clamp0115(numberOr(record["confidence"], 0.8)),
           updatedAt: numberOr(record["updatedAt"], Date.now()),
           evidenceMessageIds: Array.isArray(record["evidenceMessageIds"]) ? record["evidenceMessageIds"].filter(isNumber).slice(-12) : [],
           note: compactOptional(record["note"], 220),
@@ -680215,7 +680382,7 @@ function normalizeOutcome2(raw) {
     replyToMessageId: typeof value2.replyToMessageId === "number" ? value2.replyToMessageId : void 0,
     route: value2.route === "action" ? "action" : "chat",
     shouldReply: value2.shouldReply === true,
-    confidence: clamp0114(numberOr(value2.confidence, 0)),
+    confidence: clamp0115(numberOr(value2.confidence, 0)),
     reason: compact3(value2.reason || "", 280),
     source: compact3(value2.source || "unknown", 80),
     silentDisposition: compactOptional(value2.silentDisposition, 280),
@@ -680227,7 +680394,7 @@ function normalizeOutcome2(raw) {
     scenarioNote: compactOptional(value2.scenarioNote, 360),
     scenarioId: compactOptional(value2.scenarioId, 160),
     scenarioLabel: compactOptional(value2.scenarioLabel, 160),
-    scenarioConfidence: typeof value2.scenarioConfidence === "number" && Number.isFinite(value2.scenarioConfidence) ? clamp0114(value2.scenarioConfidence) : void 0,
+    scenarioConfidence: typeof value2.scenarioConfidence === "number" && Number.isFinite(value2.scenarioConfidence) ? clamp0115(value2.scenarioConfidence) : void 0,
     scenarioObjective: compactOptional(value2.scenarioObjective, 360),
     scenarioStateLoop: compactOptional(value2.scenarioStateLoop, 360),
     salienceSignals: Array.isArray(value2.salienceSignals) ? value2.salienceSignals.map(String).slice(0, 16) : [],
@@ -680245,7 +680412,7 @@ function normalizeDaydreamOpportunity(raw) {
     artifactId: String(value2.artifactId || "unknown"),
     generatedAt: String(value2.generatedAt || (/* @__PURE__ */ new Date()).toISOString()),
     trigger: compact3(value2.trigger || "", 240),
-    confidence: clamp0114(numberOr(value2.confidence, 0)),
+    confidence: clamp0115(numberOr(value2.confidence, 0)),
     lifecycle,
     firstSeenAt: numberOr(value2.firstSeenAt, Date.now()),
     updatedAt: numberOr(value2.updatedAt, Date.now()),
@@ -680302,7 +680469,7 @@ function commitTelegramSocialDecision(state, input) {
     replyToMessageId: input.replyToMessageId,
     route: input.route,
     shouldReply: input.shouldReply,
-    confidence: clamp0114(input.confidence),
+    confidence: clamp0115(input.confidence),
     reason: compact3(input.reason, 280),
     source: compact3(input.source, 80),
     silentDisposition: compactOptional(input.silentDisposition, 280),
@@ -680314,7 +680481,7 @@ function commitTelegramSocialDecision(state, input) {
     scenarioNote: compactOptional(input.scenarioNote, 360),
     scenarioId: compactOptional(input.scenarioId, 160),
     scenarioLabel: compactOptional(input.scenarioLabel, 160),
-    scenarioConfidence: input.scenarioConfidence === void 0 ? void 0 : clamp0114(input.scenarioConfidence),
+    scenarioConfidence: input.scenarioConfidence === void 0 ? void 0 : clamp0115(input.scenarioConfidence),
     scenarioObjective: compactOptional(input.scenarioObjective, 360),
     scenarioStateLoop: compactOptional(input.scenarioStateLoop, 360),
     salienceSignals: [...new Set((input.salienceSignals ?? []).map(String))].slice(0, 16),
@@ -680338,7 +680505,7 @@ function registerDaydreamOpportunities(state, opportunities, now2 = Date.now())
       artifactId: opportunity.artifactId || "unknown",
       generatedAt: opportunity.generatedAt || new Date(now2).toISOString(),
       trigger: compact3(opportunity.trigger, 240),
-      confidence: clamp0114(opportunity.confidence),
+      confidence: clamp0115(opportunity.confidence),
       lifecycle: "proposed",
       firstSeenAt: now2,
       updatedAt: now2,
@@ -680348,7 +680515,7 @@ function registerDaydreamOpportunities(state, opportunities, now2 = Date.now())
     };
     if (existing) {
       item.trigger = compact3(opportunity.trigger, 240) || item.trigger;
-      item.confidence = clamp0114(opportunity.confidence);
+      item.confidence = clamp0115(opportunity.confidence);
       item.updatedAt = now2;
     }
     state.daydreamOpportunities[id2] = item;
@@ -680375,7 +680542,7 @@ function setTelegramReplyModePreference(state, input) {
   const next = {
     mode: input.mode,
     scope: input.scope,
-    confidence: Math.max(existing?.confidence ?? 0, clamp0114(input.confidence ?? 0.84)),
+    confidence: Math.max(existing?.confidence ?? 0, clamp0115(input.confidence ?? 0.84)),
     updatedAt: now2,
     evidenceMessageIds: appendUnique(existing?.evidenceMessageIds ?? [], input.messageId, 12),
     note: compactOptional(input.note, 220),
@@ -680555,8 +680722,8 @@ function upsertRelationship(state, kind, fromKey, toKey, messageId, confidence2,
     evidenceMessageIds: [],
     source
   };
-  edge.confidence = Math.max(edge.confidence, clamp0114(confidence2));
-  edge.weight = Math.min(1, edge.weight + 0.12 + clamp0114(confidence2) * 0.2);
+  edge.confidence = Math.max(edge.confidence, clamp0115(confidence2));
+  edge.weight = Math.min(1, edge.weight + 0.12 + clamp0115(confidence2) * 0.2);
   edge.lastSeenAt = now2;
   edge.evidenceMessageIds = appendUnique(edge.evidenceMessageIds, messageId, 40);
   edge.note = compactOptional(note, 260) || edge.note;
@@ -680598,7 +680765,7 @@ function setPreference(vector, key, value2, confidence2, messageId, now2, note)
   const existing = vector[key];
   vector[key] = {
     value: existing ? existing.value * 0.7 + value2 * 0.3 : value2,
-    confidence: Math.max(existing?.confidence ?? 0, clamp0114(confidence2)),
+    confidence: Math.max(existing?.confidence ?? 0, clamp0115(confidence2)),
     updatedAt: now2,
     evidenceMessageIds: appendUnique(existing?.evidenceMessageIds ?? [], messageId, 12),
     note
@@ -680780,7 +680947,7 @@ async function queryVisionModel(modelName, imagePath, prompt = "Describe what yo
     return "";
   }
 }
-function clamp0115(value2, fallback = 0) {
+function clamp0116(value2, fallback = 0) {
   if (typeof value2 !== "number" || !Number.isFinite(value2)) return fallback;
   return Math.max(0, Math.min(1, value2));
 }
@@ -680797,9 +680964,9 @@ function parseVisualMemoryRecognize(raw) {
     label: String(match.label).trim(),
     objectId: typeof match.object_id === "string" ? match.object_id : void 0,
     matchedAlias: typeof match.matched_alias === "string" ? match.matched_alias : void 0,
-    blendedScore: clamp0115(match.blended_score, 0),
-    imageSimilarity: typeof match.image_similarity === "number" ? clamp0115(match.image_similarity, 0) : void 0,
-    textSimilarity: typeof match.text_similarity === "number" ? clamp0115(match.text_similarity, 0) : void 0
+    blendedScore: clamp0116(match.blended_score, 0),
+    imageSimilarity: typeof match.image_similarity === "number" ? clamp0116(match.image_similarity, 0) : void 0,
+    textSimilarity: typeof match.text_similarity === "number" ? clamp0116(match.text_similarity, 0) : void 0
   })).sort((a2, b) => b.blendedScore - a2.blendedScore).slice(0, 5);
 }
 function formatVisualFamiliarityContext(result) {
@@ -699750,7 +699917,7 @@ __export(voicechat_exports, {
   isLikelyAgentSpeechEcho: () => isLikelyAgentSpeechEcho
 });
 import { EventEmitter as EventEmitter12 } from "node:events";
-function clamp0116(x) {
+function clamp0117(x) {
   return x < 0 ? 0 : x > 1 ? 1 : x;
 }
 function alnumRatio(s2) {
@@ -699803,9 +699970,9 @@ function computeSignalFromText(text2, confidence2) {
   score -= repeatingCharPenalty(t2) * 0.4;
   score -= gibberishTokenPenalty(t2) * 0.8;
   if (typeof confidence2 === "number" && !Number.isNaN(confidence2)) {
-    score = 0.7 * score + 0.3 * clamp0116(confidence2);
+    score = 0.7 * score + 0.3 * clamp0117(confidence2);
   }
-  return clamp0116(score);
+  return clamp0117(score);
 }
 function truncateForLog(s2, n2) {
   return s2.length <= n2 ? s2 : s2.slice(0, n2 - 1) + "…";
@@ -700297,7 +700464,7 @@ Rules:
             }
           }, MAX_SEGMENT_MS);
         }
-        const signalScore = typeof snr === "number" && !Number.isNaN(snr) ? clamp0116(snr) : computeSignalFromText(text2, confidence2);
+        const signalScore = typeof snr === "number" && !Number.isNaN(snr) ? clamp0117(snr) : computeSignalFromText(text2, confidence2);
         this.lastSignalScore = signalScore;
         if (consensus !== void 0) {
           this.consensusMetadataSeen = true;
@@ -739376,7 +739543,16 @@ Respond concisely and safely. Remember: you are talking to the general public.`;
       if (!available) {
         return "ASR runtime unavailable. Omnius attempted managed transcribe-cli and Whisper fallback setup; check ~/.omnius/runtimes/asr and ~/.omnius/venv.";
       }
+      let lastRenderedTranscript = "";
+      let lastRenderedTranscriptAt = 0;
       engine.on("transcript", (text2, isFinal) => {
+        const normalizedTranscript = text2.trim();
+        const now2 = Date.now();
+        if (normalizedTranscript && normalizedTranscript === lastRenderedTranscript && now2 - lastRenderedTranscriptAt < 2500) {
+          return;
+        }
+        lastRenderedTranscript = normalizedTranscript;
+        lastRenderedTranscriptAt = now2;
         if (engine.currentMode === "confirm") {
           writeContent(
             () => renderInfo(`Heard: "${text2}" ${c3.dim("(press Enter to submit)")}`)
@@ -739801,7 +739977,7 @@ Respond concisely and safely. Remember: you are talking to the general public.`;
           statusBar.setMicActivity(evt.speechActive, evt.levelDb);
         });
         listenEng.on("info", (msg) => {
-          if (/consensus rejected|loading whisper|whisper model loaded|Whisper VAD|no accepted speech text|very short utterance/i.test(msg)) {
+          if (/consensus rejected|loading whisper|whisper model loaded/i.test(msg)) {
             writeContent(() => renderInfo(`[voicechat/asr] ${msg}`));
           }
         });

package/dist/scripts/live-whisper.py

@@ -454,7 +454,7 @@ def main():
                         help="Second whisper model run in parallel on finalized utterances; transcripts are only emitted when both models agree. 'off' disables.")
     parser.add_argument("--consensus-threshold", type=float, default=0.55,
                         help="Token similarity (0-1) required between the two models' transcripts.")
-    parser.add_argument("--silence-ms", type=float, default=900,
+    parser.add_argument("--silence-ms", type=float, default=3000,
                         help="Sustained silence that finalizes an utterance.")
     parser.add_argument("--preroll-ms", type=float, default=600,
                         help="Audio kept from before speech onset.")
@@ -577,7 +577,6 @@ def main():
     def transcribe_final(samples):
         nonlocal last_final_text
         if len(samples) < int(0.4 * SAMPLE_RATE):
-            emit_status("Whisper ignored very short utterance.")
             return
         samples = amplify(samples)
         try:
@@ -585,7 +584,6 @@ def main():
 
             text = _segment_filtered_text(result)
             if not text:
-                emit_status("Whisper final produced no accepted speech text.")
                 return
             if text == last_final_text:
                 return

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.438",
+  "version": "1.0.440",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.438",
+      "version": "1.0.440",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.438",
+  "version": "1.0.440",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",