omnius 1.0.359 → 1.0.360

package/dist/index.js

@@ -592185,6 +592185,78 @@ var init_dedup_gate = __esm({
   }
 });
 
+// packages/orchestrator/dist/conversational-scrutiny.js
+function auditPerceptionClaims(text2, evidence) {
+  if (!text2 || !text2.trim())
+    return [];
+  const tools = (evidence.toolsUsed ?? []).map((t2) => t2.toLowerCase());
+  const ran = (names) => names.some((n2) => tools.some((t2) => t2.includes(n2)));
+  const out = [];
+  for (const rule of RULES2) {
+    const m2 = text2.match(rule.pattern);
+    if (!m2)
+      continue;
+    if (ran(rule.satisfiedBy))
+      continue;
+    out.push({
+      claim: m2[0].trim().slice(0, 140),
+      kind: rule.kind,
+      requiredTool: rule.requiredTool
+    });
+  }
+  return out;
+}
+function confabulationDirective(violations) {
+  const items = violations.map((v) => `  • You claimed: "${v.claim}" — but NO ${v.kind} tool ran this turn. ACTUALLY do it: call ${v.requiredTool}.`).join("\n");
+  return [
+    `[UNSUPPORTED CLAIM — you asserted an action you did not perform]`,
+    items,
+    ``,
+    `Do NOT claim you ran vision / OCR / tests / verification unless you actually called the tool this turn.`,
+    `Either CALL the tool now and base your answer on its real output, or remove the claim and state plainly that you have not done it yet.`
+  ].join("\n");
+}
+var RULES2;
+var init_conversational_scrutiny = __esm({
+  "packages/orchestrator/dist/conversational-scrutiny.js"() {
+    "use strict";
+    RULES2 = [
+      {
+        kind: "vision",
+        // "I ran/performed full vision", "vision captioning is complete",
+        // "I analyzed/captioned/looked at each image", "ran OCR on", "I examined the screenshots".
+        pattern: /\b(ran|run|performed|did|completed|executed|finished)\b[^.!?\n]{0,40}\b(full\s+)?(vision|captioning|caption|ocr|image\s+analy[sz])|(vision|captioning|caption|ocr)\b[^.!?\n]{0,30}\b(is\s+)?(complete|done|finished|run)\b|\b(i|i've|i have)\b[^.!?\n]{0,30}\b(analy[sz]ed|captioned|ocr'?d|examined|inspected|looked\s+at|read)\b[^.!?\n]{0,30}\b(image|images|photo|photos|screenshot|screenshots|picture|pictures)\b/i,
+        satisfiedBy: [
+          "vision",
+          "ocr",
+          "image_analyze",
+          "image_read",
+          "telegram_image_analyze",
+          "video_understand",
+          "pdf_to_text",
+          "ocr_pdf",
+          "ocr_image_advanced"
+        ],
+        requiredTool: "telegram_image_analyze / vision / ocr_image_advanced"
+      },
+      {
+        kind: "test",
+        // "tests pass", "all tests passing", "I ran the tests", "the suite is green".
+        pattern: /\b(tests?\s+(pass|passed|passing|are\s+green|succeed)|all\s+tests?\b[^.!?\n]{0,20}\bpass|i\s+ran\s+the\s+tests?|(test\s+)?suite\s+(is\s+)?green)\b/i,
+        satisfiedBy: ["shell", "run_tests", "test"],
+        requiredTool: "shell (run the test command)"
+      },
+      {
+        kind: "verification",
+        // "I verified by running", "I confirmed it works by", "I checked and it runs".
+        pattern: /\bi\s+(verified|confirmed|checked|validated)\b[^.!?\n]{0,40}\b(by\s+running|works|runs|passes|it\s+is\s+running|live|deployed)\b/i,
+        satisfiedBy: ["shell", "vision", "ocr", "image_read", "file_read"],
+        requiredTool: "shell / the relevant verification tool"
+      }
+    ];
+  }
+});
+
 // packages/orchestrator/dist/index.js
 var dist_exports3 = {};
 __export(dist_exports3, {
@@ -592241,6 +592313,7 @@ __export(dist_exports3, {
   appendSteeringOutcome: () => appendSteeringOutcome,
   arcSummary: () => arcSummary,
   auditCompletionClaims: () => auditCompletionClaims,
+  auditPerceptionClaims: () => auditPerceptionClaims,
   buildAgentNotification: () => buildAgentNotification,
   buildAgentTypeSummary: () => buildAgentTypeSummary,
   buildCompletionScenarioDecomposition: () => buildCompletionScenarioDecomposition,
@@ -592276,6 +592349,7 @@ __export(dist_exports3, {
   computeArc: () => computeArc,
   computeStabilityHash: () => computeStabilityHash,
   computeTodoReminder: () => computeTodoReminder,
+  confabulationDirective: () => confabulationDirective,
   createAppState: () => createAppState,
   createChildAbortController: () => createChildAbortController,
   createCompletionLedger: () => createCompletionLedger,
@@ -592507,6 +592581,7 @@ var init_dist8 = __esm({
     init_exploration_fanout();
     init_consolidation_runtime();
     init_completion_evidence_gate();
+    init_conversational_scrutiny();
   }
 });
 
@@ -674879,6 +674954,8 @@ Join: ${newUrl}`
           streamText: "",
           visibleReplyText: "",
           currentStreamToolNames: /* @__PURE__ */ new Set(),
+          toolNamesThisRun: /* @__PURE__ */ new Set(),
+          successfulToolNamesThisRun: /* @__PURE__ */ new Set(),
           completionBoundarySeen: false,
           intermediateLines: [],
           startedAtMs: Date.now(),
@@ -675159,6 +675236,8 @@ Join: ${newUrl}`
           streamText: "",
           visibleReplyText: "",
           currentStreamToolNames: /* @__PURE__ */ new Set(),
+          toolNamesThisRun: /* @__PURE__ */ new Set(),
+          successfulToolNamesThisRun: /* @__PURE__ */ new Set(),
           completionBoundarySeen: false,
           intermediateLines: [],
           startedAtMs: Date.now(),
@@ -675864,6 +675943,7 @@ ${conversationStream}`
             suppressExternalEvent = true;
           }
           if (event.type === "tool_call" && event.toolName) {
+            subAgent.toolNamesThisRun.add(event.toolName);
             if (event.toolName === "task_complete") {
               subAgent.completionBoundarySeen = true;
               const toolNamesAtCompletion = new Set(
@@ -675895,6 +675975,7 @@ ${conversationStream}`
               `tool: ${event.toolName}(${argsPreview})`
             );
           } else if (event.type === "tool_result" && event.toolName) {
+            if (event.success) subAgent.successfulToolNamesThisRun.add(event.toolName);
             const preview = (event.content || "").slice(0, 140);
             this.subAgentViewCallbacks?.onWrite(
               subAgent.viewId,
@@ -679877,9 +679958,14 @@ ${knownList}` : "Private-user telegram_send_file target must be this DM or a kno
         const safeCaption = caption ? ` — caption: ${telegramContextJsonString(caption, 220)}` : "";
         const cacheKey = `${String(msg.chatId)}:${String(sourceMessageId ?? 0)}:${fileUniqueId}`;
         const existingEntry = this.mediaCache.get(cacheKey);
-        if (existingEntry && existsSync146(existingEntry.localPath)) {
-          existingEntry.cachedAt = Date.now();
-          return existingEntry.extractedContent || `[${sourceLabel}${type} received: path_alias=${mediaAlias}${safeCaption}]`;
+        let cacheEntry = existingEntry && existsSync146(existingEntry.localPath) ? existingEntry : void 0;
+        let localPath = cacheEntry?.localPath;
+        if (cacheEntry) {
+          cacheEntry.cachedAt = Date.now();
+          const needsDeferredImageAnalysis = isImageMedia && !eager && cacheEntry.analysisComplete !== true;
+          if (!needsDeferredImageAnalysis) {
+            return cacheEntry.extractedContent || `[${sourceLabel}${type} received: path_alias=${mediaAlias}${safeCaption}]`;
+          }
         }
         let ext = ".bin";
         if (isImageMedia) ext = telegramImageExtension(media);
@@ -679890,32 +679976,44 @@ ${knownList}` : "Private-user telegram_send_file target must be this DM or a kno
           const dotIdx = media.fileName.lastIndexOf(".");
           if (dotIdx >= 0) ext = media.fileName.slice(dotIdx);
         }
-        const localPath = await this.downloadTelegramFile(fileId, ext);
-        if (!localPath) return `[Media: ${type} — failed to download]`;
-        const cacheEntry = {
-          localPath,
-          fileId,
-          fileUniqueId,
-          chatId: msg.chatId,
-          messageId: sourceMessageId ?? 0,
-          username: msg.username,
-          mediaType: type,
-          mimeType,
-          caption,
-          cachedAt: Date.now()
-        };
-        this.mediaCache.set(cacheKey, cacheEntry);
-        const metadataKey = String(msg.chatId);
-        if (!this.mediaMetadata.has(metadataKey)) {
-          this.mediaMetadata.set(metadataKey, []);
+        if (!localPath) {
+          localPath = await this.downloadTelegramFile(fileId, ext) ?? void 0;
+          if (!localPath) return `[Media: ${type} — failed to download]`;
+        }
+        if (!cacheEntry) {
+          cacheEntry = {
+            localPath,
+            fileId,
+            fileUniqueId,
+            chatId: msg.chatId,
+            messageId: sourceMessageId ?? 0,
+            username: msg.username,
+            mediaType: type,
+            mimeType,
+            caption,
+            cachedAt: Date.now(),
+            analysisComplete: false
+          };
+          this.mediaCache.set(cacheKey, cacheEntry);
+          const metadataKey = String(msg.chatId);
+          if (!this.mediaMetadata.has(metadataKey)) {
+            this.mediaMetadata.set(metadataKey, []);
+          }
+          this.mediaMetadata.get(metadataKey).push({
+            timestamp: Date.now(),
+            type,
+            mimeType,
+            caption,
+            username: msg.username
+          });
+        } else {
+          cacheEntry.fileId = fileId;
+          cacheEntry.fileUniqueId = fileUniqueId;
+          cacheEntry.mediaType = type;
+          cacheEntry.mimeType = mimeType;
+          cacheEntry.caption = caption;
+          cacheEntry.username = msg.username;
         }
-        this.mediaMetadata.get(metadataKey).push({
-          timestamp: Date.now(),
-          type,
-          mimeType,
-          caption,
-          username: msg.username
-        });
         let description = `[${type}${caption ? `: ${caption}` : ""}]`;
         if (isImageMedia && !eager) {
           let visionContext = "";
@@ -679931,6 +680029,7 @@ ${knownList}` : "Private-user telegram_send_file target must be this DM or a kno
             );
             visionContext = formatImageContextPrefix2(ingressResult);
             cacheEntry.extractedContent = ingressResult.contextBlock;
+            cacheEntry.analysisComplete = true;
           } catch {
           }
           if (visionContext) {
@@ -679987,6 +680086,9 @@ ${visionContext}]`;
             }
           }
           description = appendMediaContextBlock(description, visualIdentityContext);
+        } else if (isImageMedia) {
+          description = `[${sourceLabel}image received: path_alias=${mediaAlias}${safeCaption}. Full visual comprehension pending; use image='${source === "reply" ? "reply" : "latest"}' or image='${mediaAlias}' with telegram_image_analyze detail='full'.]`;
+          cacheEntry.analysisComplete = false;
         } else if (type === "audio" || type === "voice") {
           let transcription = null;
           try {
@@ -679996,6 +680098,7 @@ ${visionContext}]`;
             if (result?.text?.trim()) {
               transcription = result.text.trim();
               cacheEntry.extractedContent = transcription;
+              cacheEntry.analysisComplete = true;
             }
           } catch {
           }

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.359",
+  "version": "1.0.360",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.359",
+      "version": "1.0.360",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.359",
+  "version": "1.0.360",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",