omnius 1.0.428 → 1.0.429

package/dist/index.js

@@ -603651,6 +603651,7 @@ var init_listen = __esm({
                 case "vad": {
                   lastHardwareVadAt = Date.now();
                   const doa = Number(evt.doa);
+                  this.emit("status", `Whisper VAD: ${evt.speech === true ? "speech" : "silence"}`);
                   updateListenLiveState({
                     speechActive: evt.speech === true,
                     doa: Number.isFinite(doa) ? doa : null
@@ -606474,6 +606475,7 @@ var init_live_sensors = __esm({
       previewTickInFlight = false;
       lastStreamFrameMtime = /* @__PURE__ */ new Map();
       lastStreamPersistAt = 0;
+      videoGeneration = 0;
       videoInFlight = false;
       audioInFlight = false;
       audioActivityInFlight = false;
@@ -606676,7 +606678,12 @@ var init_live_sensors = __esm({
         const next = (current + delta) % 360;
         return this.setCameraRotation(device, next, "manual", `manual ${direction === "cw" ? "clockwise" : "counter-clockwise"} dashboard override`);
       }
+      isVideoWorkCurrent(generation) {
+        return this.config.videoEnabled && generation === this.videoGeneration;
+      }
       async autoOrientCamera(device = this.config.selectedCamera, options2 = {}) {
+        const generation = this.videoGeneration;
+        if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
         const target = device || this.config.selectedCamera || firstDeviceId(this.devices.video);
         if (!target) return "No camera selected. Use /live camera <device> or select a camera in /live.";
         if (options2.onlyIfUnset && this.getCameraOrientation(target)) {
@@ -606693,17 +606700,20 @@ var init_live_sensors = __esm({
             framePath = streamFrame.path;
           }
           if (!framePath) {
+            if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
             const captured = await new CameraCaptureTool().execute({
               action: "capture",
               device: target,
               rotate_degrees: 0
             });
+            if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
             if (!captured.success) {
               return `Camera orientation detection could not capture ${target}: ${captured.error || captured.output || "unknown error"}`;
             }
             framePath = extractSavedImagePath(captured.output, this.repoRoot);
           }
           if (!framePath) return `Camera orientation detection captured ${target}, but no frame path was returned.`;
+          if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
           const cvScores = await scoreCameraOrientationWithOpenCv(framePath);
           const cvDecision = cvScores ? chooseCameraOrientationFromScores(cvScores) : null;
           const prompt = [
@@ -606721,6 +606731,7 @@ var init_live_sensors = __esm({
             prompt,
             model: "moondream3-preview"
           });
+          if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
           if (!vision.success) {
             if (cvDecision) {
               const orientation2 = this.setCameraRotation(target, cvDecision.rotation, "auto", cvDecision.reason, cvDecision.confidence);
@@ -606774,12 +606785,18 @@ var init_live_sensors = __esm({
         return this.devices;
       }
       configure(patch) {
+        const wasVideoEnabled = this.config.videoEnabled;
         this.config = {
           ...this.config,
           ...patch,
           videoIntervalMs: Math.max(MIN_VIDEO_INTERVAL_MS, Number(patch.videoIntervalMs ?? this.config.videoIntervalMs)),
           audioIntervalMs: Math.max(MIN_AUDIO_INTERVAL_MS, Number(patch.audioIntervalMs ?? this.config.audioIntervalMs))
         };
+        if (wasVideoEnabled && !this.config.videoEnabled) {
+          this.videoGeneration++;
+          this.lastStreamFrameMtime.clear();
+          this.cameraStreamers?.stopAll();
+        }
         this.persist();
         this.clearLoopTimers();
         this.ensureLoops();
@@ -606858,6 +606875,8 @@ var init_live_sensors = __esm({
         return options2.agent ? "Live PFC run enabled: visual/audio feedback is active and significant events can launch background agent review." : "Live run enabled: visual/audio feedback is active and live context will update each turn.";
       }
       async previewCamera(device = this.config.selectedCamera, options2 = {}) {
+        const generation = options2.generation ?? this.videoGeneration;
+        if (!this.isVideoWorkCurrent(generation)) return { ok: false, message: "Video context is disabled." };
         const source = device || firstDeviceId(this.devices.video);
         if (!source) return { ok: false, message: "No camera selected. Use /live to select a camera after device discovery." };
         const orientation = this.getCameraOrientation(source);
@@ -606867,6 +606886,7 @@ var init_live_sensors = __esm({
           const displayPath2 = relative11(this.repoRoot, streamFrame.path).startsWith("..") ? streamFrame.path : relative11(this.repoRoot, streamFrame.path);
           const previewWidth2 = Math.max(42, Math.min(120, Math.floor(Number(options2.previewWidth ?? (process.stdout.columns || 100) - 14))));
           const preview2 = await buildImageAsciiPreview(streamFrame.path, { width: previewWidth2 });
+          if (!this.isVideoWorkCurrent(generation)) return { ok: false, message: "Video context is disabled." };
           if (preview2?.ascii || preview2?.plainAscii) {
             const observedAt2 = Date.now();
             const cameraView2 = {
@@ -606909,6 +606929,7 @@ var init_live_sensors = __esm({
           device: source,
           rotate_degrees: rotation
         });
+        if (!this.isVideoWorkCurrent(generation)) return { ok: false, message: "Video context is disabled." };
         if (!result.success) {
           const error = result.error || result.output || "camera capture failed";
           const observedAt2 = Date.now();
@@ -606953,6 +606974,7 @@ var init_live_sensors = __esm({
         const displayPath = relative11(this.repoRoot, framePath).startsWith("..") ? framePath : relative11(this.repoRoot, framePath);
         const previewWidth = Math.max(42, Math.min(120, Math.floor(Number(options2.previewWidth ?? (process.stdout.columns || 100) - 14))));
         const preview = await buildImageAsciiPreview(framePath, { width: previewWidth });
+        if (!this.isVideoWorkCurrent(generation)) return { ok: false, message: "Video context is disabled." };
         const asciiContext = preview ? formatImageAsciiContext(preview, displayPath) : void 0;
         const observedAt = Date.now();
         const cameraView = {
@@ -607061,7 +607083,10 @@ var init_live_sensors = __esm({
         }
       }
       async sampleSingleVideoNow(source, forceInfer, previewWidth, options2 = {}) {
-        const preview = await this.previewCamera(source, { emit: false, previewWidth });
+        const generation = options2.generation ?? this.videoGeneration;
+        if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
+        const preview = await this.previewCamera(source, { emit: false, previewWidth, generation });
+        if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
         void this.runLiveVlm(source, preview.ok ? preview.framePath : void 0).catch(() => {
         });
         let inference = "";
@@ -607096,6 +607121,7 @@ var init_live_sensors = __esm({
             crop_faces: true,
             recognize_faces: true
           });
+          if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
           inference = result.success ? String(result.output || result.llmContent || "") : `Video inference failed: ${result.error || result.output || "unknown error"}`;
           const livePayload = result.success ? parseLiveMediaPayload(result.llmContent || result.output) : null;
           if (livePayload) livePayload.source = source;
@@ -607168,6 +607194,7 @@ var init_live_sensors = __esm({
         if (this.config.clipEnabled && preview.ok && preview.framePath && source && sampledAt - lastClip >= 5e3) {
           this.lastClipAt.set(source, sampledAt);
           clipSummary = await this.recognizeFrameObjects(preview.framePath, source, sampledAt, { emit: false });
+          if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
           const current = this.snapshot.cameras?.[source];
           if (current) {
             const clipLine = formatClipDashboardSummary(clipSummary);
@@ -607196,9 +607223,12 @@ ${clipSummary}` : ""].filter(Boolean).join("\n");
         return result;
       }
       async sampleVideoNow(forceInfer = this.config.inferEnabled, options2 = {}) {
+        const generation = options2.generation ?? this.videoGeneration;
+        if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
         if (this.videoInFlight) return "Video sampler is already running.";
         this.videoInFlight = true;
         try {
+          if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
           const allSources = this.activeVideoSources();
           if (allSources.length === 0) return "No camera selected. Use /live refresh after connecting hardware.";
           const sources = options2.mode === "round-robin" ? [allSources[this.nextVideoSourceIndex++ % allSources.length]] : allSources;
@@ -607206,8 +607236,10 @@ ${clipSummary}` : ""].filter(Boolean).join("\n");
           const outputs = [];
           const previewWidth = liveDashboardPreviewWidthForSources(allSources.length);
           for (const source of sources) {
-            outputs.push(await this.sampleSingleVideoNow(source, forceInfer, previewWidth, options2));
+            if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
+            outputs.push(await this.sampleSingleVideoNow(source, forceInfer, previewWidth, { ...options2, generation }));
           }
+          if (!this.isVideoWorkCurrent(generation)) return "Video context is disabled.";
           this.emitDashboard();
           return outputs.map((output, index) => `## Camera ${sources[index]}
 ${output}`).join("\n\n");
@@ -737161,7 +737193,7 @@ 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/i.test(msg)) {
+        if (/consensus rejected|loading whisper|whisper model loaded|Whisper VAD|no accepted speech text|very short utterance/i.test(msg)) {
           writeContent(() => renderInfo(`[voicechat/asr] ${msg}`));
         }
       });

package/dist/scripts/live-whisper.py

@@ -443,13 +443,18 @@ def main():
             result = run_transcribe(model, amplify(samples))
             text = _segment_filtered_text(result)
             if text:
-                emit_transcript(text, is_final=False, consensus=False)
+                emit_transcript(
+                    text,
+                    is_final=False,
+                    consensus=False if consensus_model is not None else None,
+                )
         except Exception as e:
             emit_error(f"Transcription error: {e}")
 
     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:
@@ -466,6 +471,7 @@ def main():
 
             text = _segment_filtered_text(result)
             if not text:
+                emit_status("Whisper final produced no accepted speech text.")
                 return
             if secondary is not None:
                 secondary_text = _segment_filtered_text(secondary)

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.428",
+  "version": "1.0.429",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.428",
+      "version": "1.0.429",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.428",
+  "version": "1.0.429",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",