omnius 1.0.193 → 1.0.195

package/README.md

@@ -158,8 +158,8 @@ pnpm -r build
 node scripts/build-publish.mjs
 cd publish
 mkdir -p .npm-cache
-NPM_CONFIG_CACHE=$(pwd)/.npm-cache npm pack
-NPM_CONFIG_CACHE=$(pwd)/.npm-cache npm publish --access public
+NPM_CONFIG_CACHE=$(pwd)/.npm-cache npm pack --prefer-online --cache-min=0 --registry https://registry.npmjs.org/
+NPM_CONFIG_CACHE=$(pwd)/.npm-cache npm publish --access public --prefer-online --cache-min=0 --registry https://registry.npmjs.org/
 ```
 
 Before publishing, verify `README.md`, `package.json`, `dist/index.js`, and `dist/launcher.cjs` are in the tarball, and that `package.json` includes `readmeFilename: "README.md"` plus a string `readme`.

package/dist/index.js

@@ -23930,7 +23930,25 @@ function listMediaModelCatalog(modality) {
 function resolveMediaModel(id, modality) {
   const wanted = normalizeModality(modality);
   const normalized = normalizeRepoId(id);
-  return listMediaModelCatalog(wanted).find((entry) => entry.spec.id === normalized || entry.spec.repoId === normalized || mediaModelSlug(entry.spec.id) === mediaModelSlug(normalized));
+  const normalizedLower = normalized.toLowerCase();
+  const normalizedSlug = mediaModelSlug(normalized).toLowerCase();
+  return listMediaModelCatalog(wanted).find((entry) => {
+    const spec = entry.spec;
+    const keys = [
+      spec.id,
+      spec.repoId,
+      basename4(spec.repoId),
+      spec.label,
+      mediaModelSlug(spec.id),
+      mediaModelSlug(spec.repoId),
+      mediaModelSlug(basename4(spec.repoId)),
+      mediaModelSlug(spec.label)
+    ];
+    return keys.some((key) => {
+      const candidate = String(key ?? "").trim();
+      return candidate.toLowerCase() === normalizedLower || mediaModelSlug(candidate).toLowerCase() === normalizedSlug;
+    });
+  });
 }
 function listRuntimeMediaModelSpecs(modality) {
   return listMediaModelCatalog(modality).map((entry) => entry.spec).filter((spec) => (spec.status === "active" || spec.status === "metadata-only") && mediaBackendCompatibleWithModality(spec.backend, modality));
@@ -267033,7 +267051,9 @@ ${llmAnnotation}` : result.llmContent;
 import { mkdir as mkdir15, stat as stat5, writeFile as writeFile20 } from "node:fs/promises";
 import { dirname as dirname9, extname as extname5, isAbsolute as isAbsolute2, join as join44, resolve as resolvePath } from "node:path";
 function builtInRuntimeFor(spec, kind) {
-  if (kind !== "cad" || spec.modality !== "cad")
+  if (spec.modality !== "cad")
+    return null;
+  if (kind !== "cad" && !spec.modalities.includes("3d"))
     return null;
   const id = spec.id.toLowerCase();
   if (id === "campedersen/cad0" || id === "campedersen/cad0-mini")
@@ -267061,9 +267081,9 @@ function resolveModelOutputPath(cwd4, outputPath3, ext) {
   }
   return resolved;
 }
-function formatCadSuccessOutput(args) {
+function formatCadSuccessOutput(args, warnings = []) {
   const prompt = args.prompt.length > 140 ? args.prompt.slice(0, 137) + "..." : args.prompt;
-  return [
+  const lines = [
     `CAD generated: ${args.filepath}`,
     `  Model: ${args.model}`,
     `  Backend: ${args.backend}`,
@@ -267072,7 +267092,13 @@ function formatCadSuccessOutput(args) {
     `  Prompt: "${prompt}"`,
     "  Compact IR:",
     ...args.compactIr.split("\n").map((line) => `    ${line}`)
-  ].join("\n");
+  ];
+  if (warnings.length > 0) {
+    lines.push("  Notes:");
+    for (const warning of warnings)
+      lines.push(`    - ${warning}`);
+  }
+  return lines.join("\n");
 }
 function buildCadScadArtifact(prompt, modelId) {
   const profile = inferCadProfile(prompt);
@@ -267104,7 +267130,7 @@ function inferCadProfile(prompt) {
   const lower = prompt.toLowerCase();
   const dims = inferDimensions(prompt);
   const holeCount = inferHoleCount(lower);
-  const family = /gear|tooth|teeth|cog/.test(lower) ? "gear" : /flange|bolt circle|bcd/.test(lower) ? "flange" : /standoff|spacer|bushing|sleeve/.test(lower) ? "standoff" : /enclosure|box|case|shell/.test(lower) ? "enclosure" : /l-bracket|bracket|angle/.test(lower) ? "bracket" : "plate";
+  const family = /sphere|ball|orb|globe|planet/.test(lower) ? "sphere" : /cone|conical/.test(lower) ? "cone" : /cylinder|tube|pipe|rod/.test(lower) ? "cylinder" : /cube|boxy block|block/.test(lower) && !/enclosure|case|shell/.test(lower) ? "cube" : /gear|tooth|teeth|cog/.test(lower) ? "gear" : /flange|bolt circle|bcd/.test(lower) ? "flange" : /standoff|spacer|bushing|sleeve/.test(lower) ? "standoff" : /enclosure|box|case|shell/.test(lower) ? "enclosure" : /l-bracket|bracket|angle/.test(lower) ? "bracket" : "plate";
   const holeRadius = inferHoleRadius(lower);
   return {
     family,
@@ -267201,7 +267227,11 @@ function holePositions(count, width, depth) {
 }
 function compactIrForProfile(profile) {
   const lines = [];
-  if (profile.family === "standoff" || profile.family === "flange" || profile.family === "gear") {
+  if (profile.family === "sphere") {
+    lines.push(`S ${num(profile.width / 2)}`);
+  } else if (profile.family === "cone") {
+    lines.push(`CN ${num(profile.width / 2)} 0 ${num(profile.height)}`);
+  } else if (profile.family === "standoff" || profile.family === "flange" || profile.family === "gear" || profile.family === "cylinder") {
     lines.push(`Y ${num(profile.width / 2)} ${num(profile.height)}`);
   } else {
     lines.push(`C ${num(profile.width)} ${num(profile.depth)} ${num(profile.height)}`);
@@ -267223,6 +267253,38 @@ function compactIrForProfile(profile) {
   return lines.join("\n");
 }
 function scadBodyForProfile(profile) {
+  if (profile.family === "sphere") {
+    return [
+      "difference() {",
+      "  sphere(r = width / 2);",
+      "  for (p = hole_positions) translate([p[0], p[1], -width / 2 - 0.2]) cylinder(h = width + 0.4, r = hole_r);",
+      "}"
+    ].join("\n");
+  }
+  if (profile.family === "cube") {
+    return [
+      "difference() {",
+      "  translate([-width / 2, -width / 2, -width / 2]) cube([width, width, width]);",
+      "  for (p = hole_positions) translate([p[0], p[1], -width / 2 - 0.2]) cylinder(h = width + 0.4, r = hole_r);",
+      "}"
+    ].join("\n");
+  }
+  if (profile.family === "cylinder") {
+    return [
+      "difference() {",
+      "  cylinder(h = height, r = width / 2);",
+      "  for (p = hole_positions) translate([p[0], p[1], -0.2]) cylinder(h = height + 0.4, r = hole_r);",
+      "}"
+    ].join("\n");
+  }
+  if (profile.family === "cone") {
+    return [
+      "difference() {",
+      "  cylinder(h = height, r1 = width / 2, r2 = 0);",
+      "  for (p = hole_positions) translate([p[0], p[1], -0.2]) cylinder(h = height + 0.4, r = hole_r);",
+      "}"
+    ].join("\n");
+  }
   if (profile.family === "standoff") {
     return [
       "difference() {",
@@ -267333,10 +267395,62 @@ function statusRank(status) {
     return 2;
   return 3;
 }
+function isGenericModelSelector(value2, kind) {
+  const raw = value2.trim().toLowerCase();
+  if (!raw)
+    return true;
+  if (raw === "auto" || raw === "default" || raw === "best" || raw === "smallest" || raw === "recommended")
+    return true;
+  if (kind === "3d") {
+    return /^(3d|3d model|3d-model|model|mesh|asset|object|world|glb|obj|stl|ply)$/.test(raw);
+  }
+  return /^(cad|cad model|text-to-cad|parametric cad|scad|step|openscad)$/.test(raw);
+}
+function looksLikeExplicitModelAdapter(value2) {
+  const raw = value2.trim();
+  if (!raw)
+    return false;
+  if (/^https?:\/\/huggingface\.co\//i.test(raw))
+    return true;
+  if (/^[A-Za-z0-9_.-]+\/[A-Za-z0-9_.-]+(?:\/.*)?$/.test(raw))
+    return true;
+  if (/^(hf|huggingface):/i.test(raw))
+    return true;
+  return false;
+}
+function shouldUseTextCadFallbackFor3d(kind, prompt, input, args) {
+  if (kind !== "3d")
+    return false;
+  if (!prompt || input)
+    return false;
+  const outputFormat = String(args["output_format"] ?? "").trim().toLowerCase().replace(/^\./, "");
+  if (outputFormat && outputFormat !== "scad" && outputFormat !== "step")
+    return false;
+  const outputPath3 = String(args["output_path"] ?? "").trim().toLowerCase();
+  if (/\.(glb|obj|stl|ply|png|json)$/.test(outputPath3))
+    return false;
+  return true;
+}
+function renderSelectionError(kind, selection) {
+  const label = adapterLabel(kind);
+  const available = selection.available.slice(0, 8).map((entry) => entry.spec.id);
+  if (selection.missingRequested) {
+    return [
+      `Requested ${label} "${selection.missingRequested}" was not found in the /models catalog.`,
+      available.length > 0 ? `Available ${label}s include: ${available.join(", ")}.` : `No ${label}s are currently available.`,
+      `Call generate_model with action='list_models' and kind='${kind}' to inspect selectable adapters.`
+    ].join("\n");
+  }
+  return [
+    `No ${label} is available in the /models catalog.`,
+    "Built-in CAD/3D adapters may be disabled by OMNIUS_MEDIA_MODEL_DISABLE_BUILTINS=1.",
+    "Use hf_model_discover or hf_model_intake to add a Hugging Face media model adapter."
+  ].join("\n");
+}
 function renderCatalogForKind(kind, entries) {
   if (entries.length === 0)
-    return `No ${kindLabel(kind)} model adapters are available.`;
-  const lines = [`Available ${kindLabel(kind)} model adapters:`];
+    return `No ${adapterLabel(kind)}s are available.`;
+  const lines = [`Available ${adapterLabel(kind)}s:`];
   for (const entry of entries) {
     const spec = entry.spec;
     const resources = resourceSummary(spec);
@@ -267474,6 +267588,9 @@ function resourceSummary(spec) {
 function kindLabel(kind) {
   return kind === "cad" ? "CAD" : "3D model";
 }
+function adapterLabel(kind) {
+  return kind === "cad" ? "CAD model adapter" : "3D model adapter";
+}
 var RUNTIME_BLOCKER, ModelGenerateTool;
 var init_model_generate = __esm({
   "packages/execution/dist/tools/model-generate.js"() {
@@ -267557,18 +267674,28 @@ var init_model_generate = __esm({
               durationMs: performance.now() - start2
             };
           }
-          const selected = this.selectModel(kind, args);
-          if (!selected) {
+          const selection = this.selectModel(kind, args);
+          if (!selection.entry) {
+            const error = renderSelectionError(kind, selection);
             return {
               success: false,
-              output: "",
-              error: `No ${kindLabel(kind)} model adapter is available. Use hf_model_discover or hf_model_intake to add one.`,
+              output: error,
+              error,
+              llmContent: JSON.stringify({
+                kind,
+                created: false,
+                reason: selection.missingRequested ? "model_adapter_not_found" : "no_model_adapters",
+                requestedModel: selection.missingRequested,
+                available: selection.available.map((entry) => modelSummary(entry))
+              }, null, 2),
               durationMs: performance.now() - start2
             };
           }
+          const selected = selection.entry;
           const spec = selected.spec;
           this.emit(start2, "setup", `Selected ${spec.id} (${spec.modality}/${spec.backend})`, 10);
           const readiness = await this.checkReadiness(spec, kind);
+          readiness.warnings.unshift(...selection.warnings);
           const body = renderReadiness(kind, spec, readiness);
           if (action === "check") {
             return {
@@ -267590,7 +267717,7 @@ No artifact created: ${readiness.blockers[0] ?? `${kindLabel(kind)} generation i
               durationMs: performance.now() - start2
             };
           }
-          const prompt = String(args["prompt"] ?? "").trim();
+          const prompt = String(args["prompt"] ?? "").trim() || selection.promptFromModelValue || "";
           const input = String(args["input_image"] ?? args["input"] ?? "").trim();
           if (!prompt && !input) {
             return {
@@ -267667,21 +267794,46 @@ No artifact created: provide prompt and/or input_image for ${kindLabel(kind)} ge
         return "3d";
       }
       selectModel(kind, args) {
+        const warnings = [];
+        const entries = rankCatalogEntries(kind);
         const requested = String(args["model"] ?? "").trim();
         const configured = kind === "cad" ? this.defaults.cadModel : this.defaults.model3dModel;
-        const model = requested || configured || "";
-        if (model)
-          return resolveMediaModel(model, kind);
+        let promptFromModelValue;
+        if (requested && !isGenericModelSelector(requested, kind)) {
+          const resolved = resolveMediaModel(requested, kind);
+          if (resolved)
+            return { entry: resolved, available: entries, warnings };
+          if (looksLikeExplicitModelAdapter(requested)) {
+            return { available: entries, warnings, missingRequested: requested };
+          }
+          promptFromModelValue = requested;
+          warnings.push(`Treating model="${requested}" as the requested object description, not a /models adapter id; using automatic selection.`);
+        }
+        if (configured && !isGenericModelSelector(configured, kind)) {
+          const resolved = resolveMediaModel(configured, kind);
+          if (resolved)
+            return { entry: resolved, available: entries, warnings };
+          warnings.push(`Configured /models ${kind} adapter "${configured}" is not available; falling back to automatic selection.`);
+        }
         const requestedBackend = String(args["backend"] ?? "").trim();
         const backend = requestedBackend || (kind === "cad" ? this.defaults.cadBackend : this.defaults.model3dBackend);
-        const entries = rankCatalogEntries(kind);
         if (backend && backend !== "auto") {
           const wantedBackend = backend.toLowerCase();
           const matching = entries.find((entry) => entry.spec.backend === wantedBackend);
           if (matching)
-            return matching;
+            return { entry: matching, available: entries, warnings, promptFromModelValue };
+          warnings.push(`Requested ${kindLabel(kind)} backend "${backend}" is not available; falling back to automatic selection.`);
+        }
+        const prompt = String(args["prompt"] ?? "").trim() || promptFromModelValue || "";
+        const input = String(args["input_image"] ?? args["input"] ?? "").trim();
+        if (shouldUseTextCadFallbackFor3d(kind, prompt, input, args)) {
+          const cadFallback = rankCatalogEntries("cad").find((entry) => builtInRuntimeFor(entry.spec, "cad") === "cad0-scad");
+          if (cadFallback) {
+            warnings.push("Text-only 3D generation has no wired mesh runtime yet; using the built-in CAD/SCAD adapter to create a real 3D artifact.");
+            return { entry: cadFallback, available: entries, warnings, promptFromModelValue };
+          }
         }
-        return entries[0];
+        return { entry: entries[0], available: entries, warnings, promptFromModelValue };
       }
       async checkReadiness(spec, kind) {
         const blockers = [];
@@ -267789,7 +267941,7 @@ No artifact created: ${msg}`,
           prompt: args.prompt,
           sizeKB: Math.max(1, Math.round(info.size / 1024)),
           compactIr: artifact.compactIr
-        });
+        }, args.readiness.warnings);
         return {
           success: true,
           output,
@@ -551987,6 +552139,14 @@ function normalizeProviderToolMessage(msg, model) {
     toolCalls: toolCalls.length > 0 ? toolCalls : void 0
   };
 }
+function isGenerationToolName(toolName) {
+  return /^(generate_image|generate_audio|generate_video|generate_model|generate_tts|create_audio_file)$/.test(toolName);
+}
+function isGenerationArtifactSuccess(toolName, output) {
+  if (!isGenerationToolName(toolName))
+    return false;
+  return /(?:Image generated|Music generated|Sound generated|Video generated|3D model generated|CAD generated|Model generated|TTS generated|Created [A-Z]+ file|Created|Saved to|Output saved to):?\s+/i.test(output);
+}
 function inferEpisodeModality(toolName) {
   if (VISUAL_TOOLS.has(toolName))
     return "visual";
@@ -553486,6 +553646,8 @@ ${result.output ?? ""}`;
        * no todo list, returns an empty array to allow task completion.
        */
       getOpenTodoItems() {
+        if (this.options.disableTodoCompletionGuard)
+          return [];
         const todos = this.readSessionTodos();
         if (!todos || todos.length === 0)
           return [];
@@ -557651,7 +557813,7 @@ ${_staleSamples.join("\n")}` : ``,
             const userMsg = this.pendingUserMessages.shift();
             await this.appendInjectedUserMessage(userMsg, messages2, turn);
           }
-          {
+          if (!this.options.disableTodoPlanningNudges) {
             const maybeReminder = this.getTodoReminderContent(turn);
             if (maybeReminder) {
               messages2.push({ role: "user", content: maybeReminder });
@@ -557663,7 +557825,7 @@ ${_staleSamples.join("\n")}` : ``,
             }
           }
           const turnTier = this.options.modelTier ?? "large";
-          if (turn === 0 && (turnTier === "small" || turnTier === "medium")) {
+          if (turn === 0 && !this.options.disableTodoPlanningNudges && (turnTier === "small" || turnTier === "medium")) {
             const goal = this._taskState.goal || "";
             const wordCount2 = goal.split(/\s+/).length;
             const hasMultipleActions = /\band\b.*\band\b|then.*then|also.*also/i.test(goal);
@@ -558775,6 +558937,9 @@ ${criticDecision.cachedResult.slice(0, 500)}` : `[BLOCKED — the observer confi
                   mode: "step_repetition",
                   rationale: `force_progress_block on ${tc.name} after ${criticDecision.hitNumber} identical calls`
                 });
+                const generationCompletionHint = isGenerationArtifactSuccess(tc.name, criticDecision.cachedResult) ? `
+
+[GENERATION ALREADY COMPLETE] This exact ${tc.name} call already succeeded. Do not call it again. Use the cached artifact/path above; if delivery is needed, send it, otherwise call task_complete.` : "";
                 const header = criticDecision.compacted ? `[RE-SERVED FROM CACHE — the original result was compacted from context. Here is the data again. Do not retry this exact call.]
 
 ` : `[SKIPPED DUPLICATE — exact ${tc.name} call not re-run. The cached result below is from the prior successful call. Do not retry this exact call.]
@@ -558789,7 +558954,7 @@ ${truncatedCache}`);
                   tc,
                   output: `${criticDecision.blockMessage}
 
-${header}${truncatedCache}`,
+${header}${truncatedCache}${generationCompletionHint}`,
                   success: true
                 };
               }
@@ -558807,6 +558972,9 @@ ${header}${truncatedCache}`,
                   turn,
                   timestamp: (/* @__PURE__ */ new Date()).toISOString()
                 });
+                const generationCompletionHint = isGenerationArtifactSuccess(tc.name, criticDecision.cachedResult) ? `
+
+[GENERATION ALREADY COMPLETE] This exact ${tc.name} call already succeeded. Do not call it again. Use the cached artifact/path above; if delivery is needed, send it, otherwise call task_complete.` : "";
                 const header = criticDecision.compacted ? `[RE-SERVED FROM CACHE — the original result was compacted from context. Here is the data again. No need to call this tool again.]
 
 ` : `[DUPLICATE CALL #${criticDecision.hitNumber} — you already called ${tc.name} with these exact arguments. The result is identical. Do NOT call this again. Use the data you already have to make progress. One more identical call will trigger a hard progress block.]
@@ -558814,7 +558982,7 @@ ${header}${truncatedCache}`,
 `;
                 const truncatedCache = criticDecision.cachedResult.length > 500 ? criticDecision.cachedResult.slice(0, 500) + `
 ... [${criticDecision.cachedResult.length - 500} chars omitted — same as before]` : criticDecision.cachedResult;
-                const dedupOutput = header + truncatedCache;
+                const dedupOutput = header + truncatedCache + generationCompletionHint;
                 markSyntheticToolLog(dedupOutput);
                 this.emit({
                   type: "tool_result",
@@ -560008,6 +560176,9 @@ Do NOT retry ${tc.name} with similar arguments.`);
               } else if (result.success && tc.name !== "task_complete") {
                 sameToolFailStreak = 0;
                 sameToolFailName = null;
+                if (isGenerationArtifactSuccess(tc.name, result.output ?? "")) {
+                  this.pendingUserMessages.push(`[GENERATION COMPLETE] ${tc.name} succeeded. Do not call the same generation tool again for the same request. Use the artifact/path from the tool result; if delivery is needed, send it, otherwise call task_complete.`);
+                }
               }
               if (filePath && (tc.name === "file_read" || tc.name === "file_write" || tc.name === "file_edit" || tc.name === "batch_edit" || tc.name === "file_patch")) {
                 const isModify = tc.name !== "file_read";
@@ -564537,6 +564708,8 @@ Example: ${tool.name}(${JSON.stringify(meta.examples[0].args ?? {})})` : "";
           "batch_edit"
         ]);
         const taskText = (this._taskState.originalGoal || this._taskState.goal || "").toLowerCase();
+        const wants3dModelGeneration = /\b(?:make|create|generate|build|produce|render|give)\b/.test(taskText) && /\b(?:3d|three[-\s]?d|mesh|glb|obj|stl|ply|cad|scad|step|printable|model)\b/.test(taskText) && !/\b(?:image|picture|photo|rendering|screenshot)\b/.test(taskText);
+        const wantsModelCatalogManagement = /\b(?:discover|search|find|add|install|intake|inspect|validate|save|catalog|adapter|hugging\s*face|hf)\b/.test(taskText) && /\b(?:model|adapter|hugging\s*face|hf)\b/.test(taskText);
         const taskWords = new Set(taskText.split(/\s+/).filter((w) => w.length > 2));
         const scored = [];
         for (const tool of allTools) {
@@ -564557,6 +564730,15 @@ Example: ${tool.name}(${JSON.stringify(meta.examples[0].args ?? {})})` : "";
           if (taskText.includes(tool.name.replace(/_/g, " ")) || taskText.includes(tool.name)) {
             score += customMeta ? 16 : 10;
           }
+          if (wants3dModelGeneration) {
+            if (tool.name === "generate_model")
+              score += 40;
+            if (tool.name === "generate_image")
+              score -= 12;
+            if ((tool.name === "hf_model_discover" || tool.name === "hf_model_intake") && !wantsModelCatalogManagement) {
+              score -= 18;
+            }
+          }
           if (customMeta) {
             const lastStatus = customMeta.qualityGate?.lastTest?.status;
             if (lastStatus === "passed")
@@ -633935,6 +634117,9 @@ file access, and code analysis. Respond thoroughly and helpfully.
 When asked to send a generated or existing file to Telegram, call telegram_send_file with
 path and target. Do not search for Telegram bot tokens, environment secrets, or Bot API
 credentials; upload authorization is encapsulated by telegram_send_file.
+For admin-DM artifact generation requests, send the generated file with
+telegram_send_file after the generation tool succeeds unless the admin
+explicitly asked only for a local path.
 
 When asked to generate speech, narration, or TTS, use generate_tts or audio_playback.
 Those tools handle first-use backend setup where supported. Do not fall back to shell
@@ -633943,6 +634128,9 @@ commands or generic audio generation for speech synthesis while those tools are
 When asked to generate a 3D asset, mesh, printable model, or CAD part, use
 generate_model. Use hf_model_discover or hf_model_intake only when the admin is
 asking to add, inspect, validate, or save model adapters.
+For text-only 3D requests, pass the requested object in prompt and do not put
+the object name in model; model is only for exact /models adapter ids. Do not
+use generate_image as a substitute for a 3D/CAD artifact.
 
 Keep responses concise for Telegram but don't withhold information from the admin.
 `.trim();
@@ -634057,6 +634245,7 @@ Telegram response contract:
 - Do not summarize the fact that you answered; the visible assistant text must be the answer itself.
 - If you delegated long-running work, include the sub-agent id/status and what the admin should expect next.
 - Do not narrate retry strategy ("let me try one more approach", "actually", "wait, let me try") in assistant text. If you are stuck, send a single concise blocker sentence as the reply instead of streaming deliberation.
+- Treat todo_write as optional scratch state in Telegram runs. Do not create or maintain todos for simple Q&A, capability checks, or single artifact-generation requests.
 `.trim();
     TELEGRAM_EXTERNAL_ACQUISITION_CONTRACT = `
 External acquisition contract:
@@ -634243,10 +634432,6 @@ Telegram link integrity contract:
     TELEGRAM_ALLOWED_UPDATES = ["message", "guest_message", "callback_query", "poll", "message_reaction", "message_reaction_count"];
     TELEGRAM_DEFAULT_LONG_POLL_TIMEOUT_SECONDS = 50;
     TELEGRAM_DEFAULT_ROUTER_MODEL_CANDIDATES = [
-      "qwen3:0.6b",
-      "qwen3:1.7b",
-      "qwen3:4b",
-      "qwen3:8b",
       "qwen2.5:3b",
       "qwen2.5:7b",
       "llama3.2:1b",
@@ -634254,7 +634439,11 @@ Telegram link integrity contract:
       "gemma3:1b",
       "gemma3:4b",
       "phi3:mini",
-      "phi4-mini:latest"
+      "phi4-mini:latest",
+      "qwen3:0.6b",
+      "qwen3:1.7b",
+      "qwen3:4b",
+      "qwen3:8b"
     ];
     TELEGRAM_PUBLIC_TOOL_QUOTAS = {
       web: { limit: 20, windowMs: 60 * 6e4 },
@@ -639205,6 +639394,19 @@ ${retryText}`,
         const candidates = raw ? raw.split(/[,\s]+/).map((part) => part.trim()).filter(Boolean) : TELEGRAM_DEFAULT_ROUTER_MODEL_CANDIDATES;
         return Array.from(new Set(candidates));
       }
+      telegramRouterAllowThinkHeavyAutoModels() {
+        const raw = (process.env["OMNIUS_TG_ROUTER_ALLOW_THINK_MODELS"] ?? "").trim().toLowerCase();
+        return raw === "1" || raw === "true" || raw === "on";
+      }
+      telegramRouterModelLooksThinkHeavy(name10) {
+        return /\b(?:qwen3|qwq|deepseek-r1|r1-|reasoning)\b/i.test(name10);
+      }
+      orderTelegramRouterCandidates(candidates) {
+        if (this.telegramRouterAllowThinkHeavyAutoModels()) return candidates;
+        const stable = candidates.filter((candidate) => !this.telegramRouterModelLooksThinkHeavy(candidate));
+        const thinkHeavy = candidates.filter((candidate) => this.telegramRouterModelLooksThinkHeavy(candidate));
+        return [...stable, ...thinkHeavy];
+      }
       normalizeOllamaModelNameForMatch(name10) {
         return name10.trim().toLowerCase().replace(/:latest$/, "");
       }
@@ -639235,7 +639437,7 @@ ${retryText}`,
             source: "main"
           };
         }
-        const candidates = this.telegramRouterCandidateModels();
+        const candidates = this.orderTelegramRouterCandidates(this.telegramRouterCandidateModels());
         const cacheKey = `${config.backendUrl}
 ${config.model}
 ${candidates.join(",")}`;
@@ -639263,7 +639465,7 @@ ${candidates.join(",")}`;
               atMs: now,
               model: selected,
               source: "auto-small",
-              detail: "selected first installed OMNIUS_TG_ROUTER_MODEL_CANDIDATES entry from Ollama /api/tags"
+              detail: "selected first installed Telegram router candidate from Ollama /api/tags; think-heavy models are tried last unless OMNIUS_TG_ROUTER_ALLOW_THINK_MODELS=1"
             };
             this.telegramRouterModelCache = resolved;
             return {
@@ -640425,7 +640627,7 @@ ${TELEGRAM_PUBLIC_ORCHESTRATOR_CONTRACT}`);
             return result.result?.message_id ?? null;
           } catch (err) {
             const errStr = err instanceof Error ? err.message : String(err);
-            if (this.shouldLogTelegramSendFailure(errStr)) {
+            if (process.env["OMNIUS_TELEGRAM_DEBUG_SEND_FAILURES"] === "1" && this.shouldLogTelegramSendFailure(errStr)) {
               this.tuiWrite(() => renderWarning(`Failed to send Telegram live message: ${errStr}`));
             }
             this.updateTelegramTextDeliveryCapability(chatId, {
@@ -641851,6 +642053,8 @@ ${conversationStream}`
           compactionThreshold: this.telegramFallbackCompactionThreshold(modelTier),
           contextWindowSize,
           modelTier,
+          disableTodoCompletionGuard: true,
+          disableTodoPlanningNudges: true,
           streamEnabled: true,
           dynamicContext: sessionContext.context,
           captureContextFrame: true,
@@ -642044,7 +642248,7 @@ ${currentTelegramPrompt}`;
             "If the user asks you to create an image, audio file, video, 3D/CAD model, or document artifact, create it with the scoped creative tools. Freshly generated artifacts are recorded and automatically attached to this Telegram chat when the turn completes, so do not call telegram_send_file for those same artifacts unless the user asked for a specific caption, existing/unrecorded file, or non-default target.",
             "For image generation requests, decide from the conversation whether generate_image is appropriate; do not ask the user to use a hardcoded shortcut when the request is clear.",
             "For video generation requests, decide whether generate_video is appropriate. Use mode='i2v' (image-to-video) when the user references an attached or already-generated image; otherwise mode='t2v'. Warn the user up front that video generation typically takes 2-10 minutes on consumer GPUs, and that the public Telegram video-generation quota is tight (2/hour/user).",
-            "For 3D asset, mesh, printable model, or CAD generation requests, use generate_model. Use kind='cad' for parametric/mechanical parts and kind='3d' for mesh/asset/world requests. If the tool reports a catalog/runtime blocker, explain that blocker plainly.",
+            "For 3D asset, mesh, printable model, or CAD generation requests, use generate_model. Use kind='cad' for parametric/mechanical parts and kind='3d' for mesh/asset/world requests. For text-only 3D requests, put the object description in prompt and omit model unless selecting an exact /models adapter id. Do not use generate_image as a substitute for a 3D/CAD artifact. If the tool reports a catalog/runtime blocker, explain that blocker plainly.",
             creativeWorkspace
           ].filter(Boolean).join("\n\n");
           userPrompt = `${systemPrompt}${discretionPrompt}

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.193",
+  "version": "1.0.195",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.193",
+      "version": "1.0.195",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.193",
+  "version": "1.0.195",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",
@@ -142,5 +142,5 @@
     "transcribe-cli": "^2.0.1",
     "viem": "2.47.4"
   },
-  "readme": "# Omnius\n\nOmnius is a local-first agentic coding runtime: terminal UI, autonomous coding loop, REST daemon, model router, memory layer, media tools, Telegram bridge, and peer-to-peer inference mesh in one CLI.\n\nIt is designed for open-weight and user-controlled models first, while still routing cleanly through Ollama, vLLM, OpenAI-compatible endpoints, OpenRouter, Groq, Chutes, sponsor peers, COHERE peers, and other configured providers.\n\n[![npm](https://img.shields.io/npm/v/omnius.svg)](https://www.npmjs.com/package/omnius)\n[![Node](https://img.shields.io/badge/node-%3E%3D22-brightgreen.svg)](https://nodejs.org/)\n[![License](https://img.shields.io/badge/license-CC--BY--NC--4.0-blue.svg)](LICENSE)\n\n## Install\n\n```bash\nnpm install -g omnius\nomnius\n```\n\nRequirements:\n\n- Node.js 22 or newer\n- npm 10 or newer for published CLI use\n- pnpm 9 or newer for workspace development\n- A local model or configured remote endpoint\n\nStart the REST daemon:\n\n```bash\nomnius serve\n```\n\nThe daemon defaults to `http://127.0.0.1:11435`. Open the interactive API docs at `http://127.0.0.1:11435/docs`.\n\n## What Omnius Does\n\n- Runs autonomous coding tasks, edits files, executes tools, tests changes, and iterates on failures.\n- Provides a dense terminal UI for model selection, endpoint routing, task control, shell output, voice, sponsors, Telegram, and system telemetry.\n- Exposes a REST daemon with OpenAI/Ollama-compatible inference, agentic task execution, memory, skills, tools, MCP, events, voice, projects, and governance endpoints.\n- Routes models through local, cloud, sponsor, and peer-to-peer endpoints without assuming local Ollama is the only source.\n- Supports realtime spoken conversation for ASR/TTS clients through `/realtime` and REST `realtime: true`.\n- Supports image, video, sound, music, TTS, ASR, voice clone references, Telegram media workflows, and sponsor-provided media generation.\n- Keeps project runtime state in `.omnius/`, which is intentionally ignored by git.\n\n## Common Workflows\n\n```bash\nomnius \"inspect this repo and summarize the main entrypoints\"\nomnius serve\n```\n\n```text\n/help                 command help\n/model                select or inspect the active model\n/endpoint             select or configure local, cloud, sponsor, or peer endpoints\n/realtime             toggle short ASR/TTS-oriented conversation mode\n/broker               inspect model broker, RAM/VRAM thresholds, and loaded models\n/sponsor              expose local or upstream capacity to peers\n/cohere               participate in distributed COHERE inference\n/telegram             configure or toggle the Telegram bridge\n/skills               list explorable skills and docs memories\n/pause                pause after the current turn boundary\n/stop                 interrupt the active run\n/resume               resume saved state\n```\n\n## Current Feature Areas\n\n| Area | What to read |\n| --- | --- |\n| Install and setup | [Install](docs/getting-started/install.md), [First run](docs/getting-started/first-run.md), [Model providers](docs/getting-started/model-providers.md) |\n| Terminal workflows | [TUI workflows](docs/guides/tui-workflows.md), [Slash commands](docs/reference/slash-commands.md) |\n| REST daemon | [REST reference](docs/reference/rest-api.md), [REST quickref](docs/rest/QUICKREF.md), [OpenAPI source](docs/rest/openapi-source.md) |\n| Realtime voice chat | [Realtime guide](docs/guides/realtime.md) |\n| Sponsor and COHERE mesh | [Sponsor and COHERE guide](docs/guides/sponsor-and-cohere.md) |\n| Telegram bridge | [Telegram guide](docs/guides/telegram.md) |\n| Media generation | [Media guide](docs/guides/media-generation.md) |\n| Operations | [Runtime hygiene](docs/operations/runtime-hygiene.md), [Security and remote access](docs/operations/security-and-remote-access.md) |\n| Architecture | [Architecture overview](docs/architecture/overview.md) |\n| Agent-explorable docs | [Agent memory docs index](docs/agent-memory/INDEX.md) |\n\n## Recent Highlights\n\n- `/realtime` and REST `realtime: true` provide short, natural, SOUL.md-aware conversation for ASR/TTS clients.\n- Endpoint setup and sponsor setup aggregate models from all enabled endpoints, including external OpenAI-compatible routers.\n- `/sponsor` can expose text inference and media generation for image, video, sound, and music with per-modality limits.\n- Sponsor and COHERE status surfaces now use shared telemetry concepts: concurrency, request rate, daily tokens, peer usage, model usage, and remote system metrics.\n- The TUI reports token production rate as `t/s`, supports Shift+Enter multiline input, and renders dynamic shell output inside bounded Unicode cards.\n- Telegram state is scoped by user and group, supports durable reply preferences, and feeds raw platform/tool failures back into the agent loop.\n- Ollama pool cleanup now accounts for process groups and orphan runner processes that can keep VRAM pinned.\n- REST documentation is available both as human docs and as Omnius-discoverable docs skills.\n\n## REST API\n\nStart:\n\n```bash\nomnius serve\n```\n\nUseful entrypoints:\n\n```text\nGET  /docs\nGET  /openapi.json\nPOST /v1/chat\nPOST /v1/chat/completions\nPOST /v1/run\nGET  /v1/events\nGET  /v1/skills\nPOST /v1/tools/{name}/call\nWS   /v1/voicechat/ws\n```\n\nFor shared deployments, use bearer keys:\n\n```bash\nOMNIUS_API_KEYS=\"read-key:read:grafana,run-key:run:ci:60:100000:3,admin-key:admin:ops\" omnius serve\n```\n\nSee [docs/reference/rest-api.md](docs/reference/rest-api.md) for the maintained endpoint inventory. The canonical machine contract is generated from [packages/cli/src/api/openapi.ts](packages/cli/src/api/openapi.ts).\n\n## Agent-Explorable Documentation\n\nOmnius discovers project-local docs skills from `.aiwg/addons/*/skills`. The docs bundles in this repo expose high-signal entrypoints for agents:\n\n```text\n/skills omnius docs\nskill_execute name=\"omnius-docs\"\nskill_execute name=\"omnius-rest-docs\"\nskill_extract name=\"omnius-realtime-docs\" query=\"How does realtime REST mode work?\"\n```\n\nThe intended pattern is index first, targeted document second, not loading the whole manual into the active context.\n\n## Development\n\n```bash\npnpm install\npnpm -r build\npnpm docs:check\n```\n\nFocused checks used for the docs skill surface:\n\n```bash\npnpm --filter @omnius/execution exec vitest run tests/skill-discovery.test.ts\npnpm --filter omnius exec vitest run tests/realtime-mode.test.ts tests/command-registry.test.ts\n```\n\n## Publishing\n\nPublish only from `publish/`.\n\n```bash\ncd omnius\npnpm -r clean || true\nfind . -name 'tsconfig.tsbuildinfo' -not -path '*/node_modules/*' -delete\npnpm -r build\nnode scripts/build-publish.mjs\ncd publish\nmkdir -p .npm-cache\nNPM_CONFIG_CACHE=$(pwd)/.npm-cache npm pack\nNPM_CONFIG_CACHE=$(pwd)/.npm-cache npm publish --access public\n```\n\nBefore publishing, verify `README.md`, `package.json`, `dist/index.js`, and `dist/launcher.cjs` are in the tarball, and that `package.json` includes `readmeFilename: \"README.md\"` plus a string `readme`.\n\n## License\n\nOmnius is released under [CC-BY-NC-4.0](LICENSE) for non-commercial use. Commercial use, redistribution, hosted services, and enterprise deployment require a commercial license.\n"
+  "readme": "# Omnius\n\nOmnius is a local-first agentic coding runtime: terminal UI, autonomous coding loop, REST daemon, model router, memory layer, media tools, Telegram bridge, and peer-to-peer inference mesh in one CLI.\n\nIt is designed for open-weight and user-controlled models first, while still routing cleanly through Ollama, vLLM, OpenAI-compatible endpoints, OpenRouter, Groq, Chutes, sponsor peers, COHERE peers, and other configured providers.\n\n[![npm](https://img.shields.io/npm/v/omnius.svg)](https://www.npmjs.com/package/omnius)\n[![Node](https://img.shields.io/badge/node-%3E%3D22-brightgreen.svg)](https://nodejs.org/)\n[![License](https://img.shields.io/badge/license-CC--BY--NC--4.0-blue.svg)](LICENSE)\n\n## Install\n\n```bash\nnpm install -g omnius\nomnius\n```\n\nRequirements:\n\n- Node.js 22 or newer\n- npm 10 or newer for published CLI use\n- pnpm 9 or newer for workspace development\n- A local model or configured remote endpoint\n\nStart the REST daemon:\n\n```bash\nomnius serve\n```\n\nThe daemon defaults to `http://127.0.0.1:11435`. Open the interactive API docs at `http://127.0.0.1:11435/docs`.\n\n## What Omnius Does\n\n- Runs autonomous coding tasks, edits files, executes tools, tests changes, and iterates on failures.\n- Provides a dense terminal UI for model selection, endpoint routing, task control, shell output, voice, sponsors, Telegram, and system telemetry.\n- Exposes a REST daemon with OpenAI/Ollama-compatible inference, agentic task execution, memory, skills, tools, MCP, events, voice, projects, and governance endpoints.\n- Routes models through local, cloud, sponsor, and peer-to-peer endpoints without assuming local Ollama is the only source.\n- Supports realtime spoken conversation for ASR/TTS clients through `/realtime` and REST `realtime: true`.\n- Supports image, video, sound, music, TTS, ASR, voice clone references, Telegram media workflows, and sponsor-provided media generation.\n- Keeps project runtime state in `.omnius/`, which is intentionally ignored by git.\n\n## Common Workflows\n\n```bash\nomnius \"inspect this repo and summarize the main entrypoints\"\nomnius serve\n```\n\n```text\n/help                 command help\n/model                select or inspect the active model\n/endpoint             select or configure local, cloud, sponsor, or peer endpoints\n/realtime             toggle short ASR/TTS-oriented conversation mode\n/broker               inspect model broker, RAM/VRAM thresholds, and loaded models\n/sponsor              expose local or upstream capacity to peers\n/cohere               participate in distributed COHERE inference\n/telegram             configure or toggle the Telegram bridge\n/skills               list explorable skills and docs memories\n/pause                pause after the current turn boundary\n/stop                 interrupt the active run\n/resume               resume saved state\n```\n\n## Current Feature Areas\n\n| Area | What to read |\n| --- | --- |\n| Install and setup | [Install](docs/getting-started/install.md), [First run](docs/getting-started/first-run.md), [Model providers](docs/getting-started/model-providers.md) |\n| Terminal workflows | [TUI workflows](docs/guides/tui-workflows.md), [Slash commands](docs/reference/slash-commands.md) |\n| REST daemon | [REST reference](docs/reference/rest-api.md), [REST quickref](docs/rest/QUICKREF.md), [OpenAPI source](docs/rest/openapi-source.md) |\n| Realtime voice chat | [Realtime guide](docs/guides/realtime.md) |\n| Sponsor and COHERE mesh | [Sponsor and COHERE guide](docs/guides/sponsor-and-cohere.md) |\n| Telegram bridge | [Telegram guide](docs/guides/telegram.md) |\n| Media generation | [Media guide](docs/guides/media-generation.md) |\n| Operations | [Runtime hygiene](docs/operations/runtime-hygiene.md), [Security and remote access](docs/operations/security-and-remote-access.md) |\n| Architecture | [Architecture overview](docs/architecture/overview.md) |\n| Agent-explorable docs | [Agent memory docs index](docs/agent-memory/INDEX.md) |\n\n## Recent Highlights\n\n- `/realtime` and REST `realtime: true` provide short, natural, SOUL.md-aware conversation for ASR/TTS clients.\n- Endpoint setup and sponsor setup aggregate models from all enabled endpoints, including external OpenAI-compatible routers.\n- `/sponsor` can expose text inference and media generation for image, video, sound, and music with per-modality limits.\n- Sponsor and COHERE status surfaces now use shared telemetry concepts: concurrency, request rate, daily tokens, peer usage, model usage, and remote system metrics.\n- The TUI reports token production rate as `t/s`, supports Shift+Enter multiline input, and renders dynamic shell output inside bounded Unicode cards.\n- Telegram state is scoped by user and group, supports durable reply preferences, and feeds raw platform/tool failures back into the agent loop.\n- Ollama pool cleanup now accounts for process groups and orphan runner processes that can keep VRAM pinned.\n- REST documentation is available both as human docs and as Omnius-discoverable docs skills.\n\n## REST API\n\nStart:\n\n```bash\nomnius serve\n```\n\nUseful entrypoints:\n\n```text\nGET  /docs\nGET  /openapi.json\nPOST /v1/chat\nPOST /v1/chat/completions\nPOST /v1/run\nGET  /v1/events\nGET  /v1/skills\nPOST /v1/tools/{name}/call\nWS   /v1/voicechat/ws\n```\n\nFor shared deployments, use bearer keys:\n\n```bash\nOMNIUS_API_KEYS=\"read-key:read:grafana,run-key:run:ci:60:100000:3,admin-key:admin:ops\" omnius serve\n```\n\nSee [docs/reference/rest-api.md](docs/reference/rest-api.md) for the maintained endpoint inventory. The canonical machine contract is generated from [packages/cli/src/api/openapi.ts](packages/cli/src/api/openapi.ts).\n\n## Agent-Explorable Documentation\n\nOmnius discovers project-local docs skills from `.aiwg/addons/*/skills`. The docs bundles in this repo expose high-signal entrypoints for agents:\n\n```text\n/skills omnius docs\nskill_execute name=\"omnius-docs\"\nskill_execute name=\"omnius-rest-docs\"\nskill_extract name=\"omnius-realtime-docs\" query=\"How does realtime REST mode work?\"\n```\n\nThe intended pattern is index first, targeted document second, not loading the whole manual into the active context.\n\n## Development\n\n```bash\npnpm install\npnpm -r build\npnpm docs:check\n```\n\nFocused checks used for the docs skill surface:\n\n```bash\npnpm --filter @omnius/execution exec vitest run tests/skill-discovery.test.ts\npnpm --filter omnius exec vitest run tests/realtime-mode.test.ts tests/command-registry.test.ts\n```\n\n## Publishing\n\nPublish only from `publish/`.\n\n```bash\ncd omnius\npnpm -r clean || true\nfind . -name 'tsconfig.tsbuildinfo' -not -path '*/node_modules/*' -delete\npnpm -r build\nnode scripts/build-publish.mjs\ncd publish\nmkdir -p .npm-cache\nNPM_CONFIG_CACHE=$(pwd)/.npm-cache npm pack --prefer-online --cache-min=0 --registry https://registry.npmjs.org/\nNPM_CONFIG_CACHE=$(pwd)/.npm-cache npm publish --access public --prefer-online --cache-min=0 --registry https://registry.npmjs.org/\n```\n\nBefore publishing, verify `README.md`, `package.json`, `dist/index.js`, and `dist/launcher.cjs` are in the tarball, and that `package.json` includes `readmeFilename: \"README.md\"` plus a string `readme`.\n\n## License\n\nOmnius is released under [CC-BY-NC-4.0](LICENSE) for non-commercial use. Commercial use, redistribution, hosted services, and enterprise deployment require a commercial license.\n"
 }