@lazyingart/agintiflow 0.20.13 → 0.20.15

package/README.md

@@ -21,6 +21,10 @@ AgInTiFlow is AgInTi's web-first agent platform for controlled website automatio
 
 It is designed for workflows where an AI agent should act, but every tool, log, and session state should remain inspectable.
 
+<p align="center">
+  <img src="./demos/agintiflow-cli-launch.jpg" alt="AgInTiFlow interactive CLI launch screen with colorful terminal banner, Docker workspace status, and chat input panel" width="960" />
+</p>
+
 ## Product Snapshot
 
 | Area | Direction |

package/demos/README.md

@@ -0,0 +1,6 @@
+# AgInTiFlow Demos
+
+This folder stores shareable screenshots and visual demos used by the repository README.
+
+- `agintiflow-cli-launch.png`: interactive CLI launch screen with the AgInTi Flow terminal banner, Docker workspace status, and input panel.
+- `agintiflow-cli-launch.jpg`: optimized README/web preview derived from the same screenshot.

package/docs/model-selection.md

@@ -46,9 +46,27 @@ Interactive commands:
 /auxiliary model grsai/nano-banana-2
 ```
 
-In the interactive CLI, `/provider`, `/route`, `/model`, `/spare`, and `/auxiliary model` without arguments open selectors. Use Up/Down/Left/Right to move through choices, Enter to confirm, and Esc to cancel. Slash-command hints use the same arrow selection behavior: type a prefix such as `/mo`, use arrows to choose `/model` or `/models`, then press Enter or Tab.
+In the interactive CLI, `/provider`, `/route`, `/model`, `/spare`, and `/auxiliary model` without arguments open selectors. Use Up/Down/Left/Right to move through choices and Enter to confirm. Esc cancels at the top level; inside multi-level selectors it goes back one level. Slash-command hints use the same arrow selection behavior: type a prefix such as `/mo`, use arrows to choose `/model` or `/models`, then press Enter or Tab.
 
-`/route`, `/model`, and `/spare` intentionally share the same text-model selector so users do not need to learn three different catalogs. The shared list is grouped as DeepSeek, Venice Uncensored, Venice GPT/Claude/Gemma/Qwen, OpenAI, Qwen, and Mock. OpenAI entries show the recommended default reasoning effort in the description; `/spare` stores that reasoning value when selected.
+`/route`, `/model`, and `/spare` intentionally share the same text-model selector so users do not need to learn three different catalogs. The selector is hierarchical:
+
+1. Choose a provider family.
+2. Choose the model in that family.
+3. For OpenAI models, choose reasoning effort: `low`, `medium`, `high`, or `xhigh`.
+
+Provider families:
+
+| Family | Models |
+| --- | --- |
+| DeepSeek | `deepseek-v4-flash`, `deepseek-v4-pro` |
+| Venice | `venice-uncensored-1-2`, `venice-uncensored`, `gemma-4-uncensored` |
+| OpenAI | `gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.3-codex`, `gpt-5.3-codex-spark`, `gpt-5.2`; each has low/medium/high/xhigh reasoning |
+| Venice GPT | GPT-family Venice routes such as `openai-gpt-55`, `openai-gpt-54`, `openai-gpt-54-mini`, `openai-gpt-53-codex`, `openai-gpt-52` |
+| Venice Gemma | Gemma instruct routes such as `google-gemma-4-31b-it`, `google-gemma-4-26b-a4b-it`, `google-gemma-3-27b-it` |
+| Venice Claude | Claude Sonnet/Opus routes such as `claude-sonnet-4-6`, `claude-opus-4-7`, `claude-opus-4-6` |
+| Venice Qwen | Qwen routes such as `qwen3-6-27b`, `qwen-3-6-plus`, `qwen3-coder-480b-a35b-instruct-turbo` |
+| Qwen | `qwen-plus`, `qwen-turbo`, `qwen-max` |
+| Mock | `mock-agent` |
 
 `/venice` opens a two-step selector for the Venice route and main models. The current text choices are:
 
@@ -77,14 +95,21 @@ It keeps smart routing enabled. If the Venice key is missing, run `/auth venice`
 | `deepseek` | DeepSeek | Default route/main because V4 Flash and V4 Pro are cheap and strong. |
 | `openai` | OpenAI | Spare/frontier checks, Codex-family work, and explicit manual routes. |
 | `qwen` | Qwen | Chinese and general-purpose OpenAI-compatible tasks. |
-| `venice-uncensored` | Venice | Venice-native uncensored text models. |
+| `venice` | Venice | Venice-native text shortcuts: Venice 1.2, Venice 1.1, and Gemma 4 Uncensored. |
 | `venice-gpt` | Venice | GPT-family models through Venice. |
 | `venice-claude` | Venice | Claude-family models through Venice. |
-| `venice-gemma` | Venice | Gemma-family models through Venice. |
+| `venice-gemma` | Venice | Gemma instruct models through Venice, excluding the Gemma 4 Uncensored shortcut. |
 | `venice-qwen` | Venice | Qwen-family models through Venice. |
 | `venice-image` | Venice | Image generation/editing such as Nano Banana, GPT Image, Wan, Qwen Image. |
 | `grsai` | GRS AI | Auxiliary image generation only. |
 
+`/auxiliary model` uses the same two-level pattern for image tools:
+
+| Family | Models |
+| --- | --- |
+| GRS AI | `nano-banana-2`, `nano-banana-2-edit`, `gpt-image-2`, `gpt-image-2-edit` when the configured GRS AI-compatible endpoint supports them |
+| Venice Image | `wan-2-7-pro-edit`, `nano-banana-2`, `gpt-image-2`, `grok-imagine-image`, `qwen-image-2-pro`, `bria-bg-remover`, `recraft-v4`, `flux-2-pro`, and related image/edit routes |
+
 ## Keys
 
 Keys are set per project and stored only in ignored `.aginti/.env`:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lazyingart/agintiflow",
-  "version": "0.20.13",
+  "version": "0.20.15",
   "type": "module",
   "description": "AgInTiFlow is a web-first coding agent and CLI with DeepSeek routing, sandboxed tools, model providers, canvas artifacts, and optional wrappers.",
   "license": "Apache-2.0",
@@ -37,6 +37,7 @@
     "bin/",
     "docs/",
     "docker/",
+    "demos/agintiflow-cli-launch.jpg",
     "i18n/",
     "logos/banner-opaque.png",
     "logos/logo.png",

package/public/app.js

@@ -989,7 +989,7 @@ function fieldValue(field) {
 }
 
 function providerModelOptions(provider = providerField.value) {
-  return modelCatalog[provider] || [];
+  return (modelCatalog[provider] || []).filter((item) => !item.hidden);
 }
 
 function mergeModelOptions(...groups) {

package/references/model-routing-provider-design.md

@@ -55,6 +55,8 @@ Interactive equivalents:
 
 `/route`, `/model`, and `/spare` share a single text-model selector. The selector is grouped as DeepSeek, Venice Uncensored, Venice GPT/Claude/Gemma/Qwen, OpenAI, Qwen, and Mock so the user sees the same union of route/main/spare-capable models in every role.
 
+Interactive selectors are hierarchical. Enter moves forward or confirms. Esc cancels only at the first level; at model/reasoning levels it returns to the previous selector instead of dropping the whole flow.
+
 `/venice` opens a route/main selector for Venice text models. The selector includes `venice/venice-uncensored-1-2` (Venice 1.2), `venice/venice-uncensored` (Venice 1.1), `venice/gemma-4-uncensored` (Gemma 4), and a Disable Venice option. In non-interactive shells, `/venice` keeps script compatibility by selecting `venice/venice-uncensored-1-2` for both roles. `/venice 1.2 gemma` sets route to Venice 1.2 and main to Gemma 4; `/venice off` or the Disable Venice selector option restores `deepseek/deepseek-v4-flash` for route and `deepseek/deepseek-v4-pro` for main.
 
 The web UI should expose model names as dropdowns, not free-text fields. The left panel should stay focused on common daily controls, while model-role editing and less-used switches live in an Advanced settings modal. The terminal-like capability panels belong after the runtime log so the left control panel remains short.

package/scripts/smoke-model-roles.js

@@ -109,9 +109,14 @@ const fastRoute = selectModelRoute({
 assert(fastRoute.model === "deepseek-v4-flash", "fast route did not use route model override");
 
 assert(MODEL_PROVIDER_GROUPS["venice-gpt"].provider === "venice", "venice-gpt group missing");
-assert(modelsForProviderGroup("venice-gemma").some((item) => item.id === "gemma-4-uncensored"), "venice-gemma bucket missing Gemma");
-assert(modelsForProviderGroup("venice-uncensored").some((item) => item.id === "e2ee-venice-uncensored-24b-p"), "venice-uncensored bucket missing Venice 1.1");
+assert(modelsForProviderGroup("venice").some((item) => item.id === "venice-uncensored-1-2"), "venice group missing Venice 1.2");
+assert(modelsForProviderGroup("venice-gemma").some((item) => item.id === "google-gemma-4-31b-it"), "venice-gemma bucket missing Gemma 4 instruct");
+assert(!modelsForProviderGroup("venice-gemma").some((item) => item.id === "gemma-4-uncensored"), "venice-gemma bucket should not include Gemma 4 Uncensored shortcut");
+assert(modelsForProviderGroup("venice-gpt").some((item) => item.id === "openai-gpt-54-mini"), "venice-gpt bucket missing GPT-5.4 Mini");
+assert(modelsForProviderGroup("venice-claude").some((item) => item.id === "claude-opus-4-7"), "venice-claude bucket missing Claude Opus 4.7");
+assert(modelsForProviderGroup("venice-qwen").some((item) => item.id === "qwen3-coder-480b-a35b-instruct-turbo"), "venice-qwen bucket missing Qwen Coder");
 assert(AUXILIARY_MODEL_CATALOG["venice-image"].some((item) => item.id === "gpt-image-2"), "Venice image catalog missing GPT Image 2");
+assert(AUXILIARY_MODEL_CATALOG["venice-image"].some((item) => item.id === "wan-2-7-pro-edit"), "Venice image catalog missing Wan edit");
 const routeChoices = modelRoleChoices("route").map((item) => `${item.provider}/${item.model}`);
 const mainChoices = modelRoleChoices("main").map((item) => `${item.provider}/${item.model}`);
 const spareChoices = modelRoleChoices("spare").map((item) => `${item.provider}/${item.model}`);
@@ -123,9 +128,14 @@ for (const expected of [
   "venice/venice-uncensored-1-2",
   "venice/venice-uncensored",
   "venice/gemma-4-uncensored",
+  "venice/google-gemma-4-31b-it",
+  "venice/google-gemma-4-26b-a4b-it",
   "venice/openai-gpt-55",
+  "venice/openai-gpt-54-mini",
   "venice/claude-sonnet-4-6",
+  "venice/claude-opus-4-7",
   "venice/qwen3-6-27b",
+  "venice/qwen3-coder-480b-a35b-instruct-turbo",
   "openai/gpt-5.5",
   "openai/gpt-5.4",
   "openai/gpt-5.4-mini",
@@ -137,7 +147,9 @@ for (const expected of [
   assert(routeChoices.includes(expected), `shared model selector missing ${expected}`);
 }
 assert(!routeChoices.includes("venice/e2ee-venice-uncensored-24b-p"), "shared model selector should hide unstable E2EE Venice 1.1");
+assert(modelRoleChoices("route").some((item) => item.provider === "openai" && item.model === "gpt-5.5" && item.reasoningOptions.includes("xhigh")), "OpenAI selector missing reasoning levels");
 assert(modelRoleChoices("auxiliary").some((item) => item.provider === "grsai"), "auxiliary selector missing GRS AI");
+assert(modelRoleChoices("auxiliary").some((item) => item.provider === "venice" && item.model === "wan-2-7-pro-edit"), "auxiliary selector missing Venice Wan edit");
 const parsedTextToolCalls = parseTextToolCalls('[TOOL_CALLS]list_files[ARGS]call_123[ARGS]{"path":".","maxDepth":1}');
 assert(parsedTextToolCalls.length === 1, "Venice text tool-call parser did not detect encoded tool call");
 assert(parsedTextToolCalls[0].function.name === "list_files", "Venice text tool-call parser returned wrong tool name");

package/src/interactive-cli.js

@@ -1651,8 +1651,10 @@ function createState(args = {}) {
     preferredWrapper: args.preferredWrapper || "codex",
     routeProvider: args.routeProvider || "",
     routeModel: args.routeModel || "",
+    routeReasoning: args.routeReasoning || "",
     mainProvider: args.mainProvider || "",
     mainModel: args.mainModel || "",
+    mainReasoning: args.mainReasoning || "",
     spareProvider: args.spareProvider || "openai",
     spareModel: args.spareModel || "gpt-5.4",
     spareReasoning: args.spareReasoning || "medium",
@@ -1694,8 +1696,10 @@ function useDeepSeekDefaults(state) {
   state.model = "";
   state.routeProvider = "deepseek";
   state.routeModel = "deepseek-v4-flash";
+  state.routeReasoning = "";
   state.mainProvider = "deepseek";
   state.mainModel = "deepseek-v4-pro";
+  state.mainReasoning = "";
 }
 
 function veniceTextModelChoices() {
@@ -1766,8 +1770,10 @@ function printModelRoles(state) {
   const roles = getModelRoleDefaults({
     routeProvider: state.routeProvider,
     routeModel: state.routeModel,
+    routeReasoning: state.routeReasoning,
     mainProvider: state.mainProvider,
     mainModel: state.mainModel,
+    mainReasoning: state.mainReasoning,
     spareProvider: state.spareProvider,
     spareModel: state.spareModel,
     spareReasoning: state.spareReasoning,
@@ -1827,13 +1833,17 @@ function compactReasoning(model = {}) {
   return `${preferred} reasoning`;
 }
 
+function isReasoningLevel(value = "") {
+  return ["low", "medium", "high", "xhigh"].includes(String(value || "").toLowerCase());
+}
+
 function modelSelectorGroup(provider, model = {}) {
   if (provider === "deepseek") return "DeepSeek";
   if (provider === "openai") return "OpenAI";
   if (provider === "qwen") return "Qwen";
   if (provider === "mock") return "Mock";
   if (provider === "venice") {
-    if (model.bucket === "venice-uncensored") return "Venice Uncensored";
+    if (model.bucket === "venice") return "Venice";
     if (model.bucket === "venice-gpt") return "Venice GPT";
     if (model.bucket === "venice-claude") return "Venice Claude";
     if (model.bucket === "venice-gemma") return "Venice Gemma";
@@ -1843,25 +1853,96 @@ function modelSelectorGroup(provider, model = {}) {
   return provider;
 }
 
-function textModelRoleChoices() {
-  const providerOrder = ["deepseek", "venice", "openai", "qwen", "mock"];
-  const veniceOrder = [
-    "venice-uncensored-1-2",
-    "venice-uncensored",
-    "gemma-4-uncensored",
-    "openai-gpt-55",
-    "claude-sonnet-4-6",
-    "qwen3-6-27b",
-  ];
+const TEXT_MODEL_GROUPS = [
+  {
+    id: "deepseek",
+    label: "DeepSeek",
+    description: "DeepSeek V4 Flash and Pro defaults.",
+  },
+  {
+    id: "venice",
+    label: "Venice",
+    description: "Venice 1.2, Venice 1.1, and Gemma 4 Uncensored.",
+  },
+  {
+    id: "openai",
+    label: "OpenAI",
+    description: "GPT and Codex models with selectable reasoning.",
+  },
+  {
+    id: "venice-gpt",
+    label: "Venice GPT",
+    description: "OpenAI/GPT-family text models through Venice.",
+  },
+  {
+    id: "venice-gemma",
+    label: "Venice Gemma",
+    description: "Gemma instruct models through Venice, excluding Gemma 4 Uncensored.",
+  },
+  {
+    id: "venice-claude",
+    label: "Venice Claude",
+    description: "Claude Sonnet and Opus models through Venice.",
+  },
+  {
+    id: "venice-qwen",
+    label: "Venice Qwen",
+    description: "Qwen text and coder models through Venice.",
+  },
+  {
+    id: "qwen",
+    label: "Qwen",
+    description: "DashScope Qwen Plus/Turbo/Max.",
+  },
+  {
+    id: "mock",
+    label: "Mock",
+    description: "Deterministic local test model.",
+  },
+];
+
+function modelGroupForSelection(provider, model = {}) {
+  if (provider === "venice") return model.bucket || "venice";
+  return provider;
+}
+
+function modelGroupChoices() {
+  return TEXT_MODEL_GROUPS.map((group) => ({
+    action: "group",
+    groupId: group.id,
+    label: group.label,
+    description: group.description,
+  }));
+}
+
+function modelsForSelectorGroup(groupId = "") {
+  const group = String(groupId || "");
+  if (group === "deepseek") return (PROVIDER_MODEL_CATALOG.deepseek || []).filter((model) => !model.hidden);
+  if (group === "openai") return (PROVIDER_MODEL_CATALOG.openai || []).filter((model) => !model.hidden);
+  if (group === "qwen") return (PROVIDER_MODEL_CATALOG.qwen || []).filter((model) => !model.hidden);
+  if (group === "mock") return (PROVIDER_MODEL_CATALOG.mock || []).filter((model) => !model.hidden);
+  if (group === "venice") {
+    const order = ["venice-uncensored-1-2", "venice-uncensored", "gemma-4-uncensored"];
+    const byId = new Map((PROVIDER_MODEL_CATALOG.venice || []).filter((model) => !model.hidden).map((model) => [model.id, model]));
+    return order.map((id) => byId.get(id)).filter(Boolean);
+  }
+  if (group.startsWith("venice-")) {
+    return (PROVIDER_MODEL_CATALOG.venice || []).filter((model) => !model.hidden && model.bucket === group);
+  }
+  return [];
+}
+
+function providerForSelectorGroup(groupId = "") {
+  if (String(groupId).startsWith("venice")) return "venice";
+  return groupId;
+}
+
+function textModelRoleChoices(groupId = "") {
+  const groups = groupId ? [groupId] : TEXT_MODEL_GROUPS.map((group) => group.id);
   const choices = [];
-  for (const provider of providerOrder) {
-    let models = PROVIDER_MODEL_CATALOG[provider] || [];
-    if (provider === "venice") {
-      const byId = new Map(models.filter((model) => !model.hidden).map((model) => [model.id, model]));
-      models = veniceOrder.map((id) => byId.get(id)).filter(Boolean);
-    } else {
-      models = models.filter((model) => !model.hidden);
-    }
+  for (const groupIdItem of groups) {
+    const provider = providerForSelectorGroup(groupIdItem);
+    const models = modelsForSelectorGroup(groupIdItem);
     for (const model of models) {
       const group = modelSelectorGroup(provider, model);
       const reasoning = compactReasoning(model);
@@ -1874,6 +1955,7 @@ function textModelRoleChoices() {
         group,
         description: details,
         reasoningDefault: reasoning ? reasoning.split(" ")[0] : "",
+        reasoningOptions: Array.isArray(model.reasoning) ? model.reasoning : [],
       });
     }
   }
@@ -1882,30 +1964,47 @@ function textModelRoleChoices() {
 
 export function modelRoleChoices(role = "main") {
   if (role !== "auxiliary") return textModelRoleChoices();
-  const common = [
-    {
-      provider: "grsai",
-      model: "nano-banana-2",
-      label: "GRS AI Nano Banana",
-      description: "default auxiliary image model",
-      auxiliary: true,
-    },
+  return auxiliaryModelRoleChoices();
+}
+
+function auxiliaryGroupChoices() {
+  return [
     {
-      provider: "venice",
-      model: "gpt-image-2",
-      label: "Venice GPT Image 2",
-      description: "high-quality Venice image generation",
-      auxiliary: true,
+      action: "group",
+      groupId: "grsai",
+      label: "GRS AI",
+      description: "Nano Banana and GPT image models through GRS AI-compatible keys.",
     },
     {
-      provider: "venice",
-      model: "nano-banana-2",
-      label: "Venice Nano Banana 2",
-      description: "Venice image generation route",
-      auxiliary: true,
+      action: "group",
+      groupId: "venice-image",
+      label: "Venice Image",
+      description: "Venice image, edit, background-removal, Qwen, Wan, Grok, Recraft, Flux routes.",
     },
   ];
-  return common.filter((item) => item.auxiliary).sort((a, b) => Number(b.provider === "grsai") - Number(a.provider === "grsai"));
+}
+
+function auxiliaryProviderForGroup(groupId = "") {
+  return groupId === "venice-image" ? "venice" : "grsai";
+}
+
+function auxiliaryModelRoleChoices(groupId = "") {
+  const groups = groupId ? [groupId] : ["grsai", "venice-image"];
+  const choices = [];
+  for (const group of groups) {
+    const provider = auxiliaryProviderForGroup(group);
+    for (const model of AUXILIARY_MODEL_CATALOG[group] || []) {
+      choices.push({
+        provider,
+        model: model.id,
+        label: `${group === "venice-image" ? "Venice Image" : "GRS AI"} · ${model.label || model.id}`,
+        group,
+        description: [model.type, model.price, model.description].filter(Boolean).join("; "),
+        auxiliary: true,
+      });
+    }
+  }
+  return choices;
 }
 
 function providerChoices() {
@@ -1945,6 +2044,7 @@ function providerChoices() {
 
 function modelChoiceLine(option) {
   if (option.action === "disable") return `${option.label}  ${option.description}`;
+  if (option.action === "group" || option.action === "reasoning") return `${option.label}  ${option.description}`;
   return `${option.label}  ${option.provider}/${option.model}  ${option.description}`;
 }
 
@@ -1979,7 +2079,13 @@ function renderSelector({ title, subtitle, options, selectedIndex, lineCount = 0
   return rows.length;
 }
 
-function selectModelChoice({ title, subtitle, options, initialIndex = 0 }) {
+const SELECTOR_BACK = Symbol("selector-back");
+
+function isSelectorBack(value) {
+  return value === SELECTOR_BACK;
+}
+
+function selectModelChoice({ title, subtitle, options, initialIndex = 0, escapeValue = null }) {
   if (!input.isTTY || !output.isTTY || typeof input.setRawMode !== "function") return Promise.resolve(null);
   return new Promise((resolve) => {
     emitKeypressEvents(input);
@@ -2007,7 +2113,7 @@ function selectModelChoice({ title, subtitle, options, initialIndex = 0 }) {
         return;
       }
       if (key.name === "escape") {
-        finish(null);
+        finish(escapeValue);
         return;
       }
       if (key.name === "up" || key.name === "left") {
@@ -2034,7 +2140,6 @@ function selectModelChoice({ title, subtitle, options, initialIndex = 0 }) {
 }
 
 async function pickModelRole(role, state) {
-  const options = modelRoleChoices(role);
   const currentProvider =
     role === "route"
       ? state.routeProvider || "deepseek"
@@ -2051,24 +2156,103 @@ async function pickModelRole(role, state) {
         : role === "auxiliary"
           ? state.auxiliaryModel || "nano-banana-2"
           : state.mainModel || "deepseek-v4-pro";
-  const initialIndex = Math.max(
-    options.findIndex((item) => item.provider === currentProvider && item.model === currentModel),
+
+  const groupOptions = role === "auxiliary" ? auxiliaryGroupChoices() : modelGroupChoices();
+  const currentModelEntry =
+    role === "auxiliary"
+      ? modelRoleChoices("auxiliary").find((item) => item.provider === currentProvider && item.model === currentModel)
+      : modelRoleChoices(role).find((item) => item.provider === currentProvider && item.model === currentModel);
+  const currentGroup = currentModelEntry?.group || currentProvider;
+  let selected = null;
+  let selectedReasoning = "";
+  let initialGroupIndex = Math.max(
+    groupOptions.findIndex((item) => item.groupId === currentGroup),
     0
   );
-  const selected = await selectModelChoice({
-    title:
-      role === "route"
-        ? "Select route model"
-        : role === "spare"
-          ? "Select spare model"
-          : role === "auxiliary"
-            ? "Select auxiliary model"
-            : "Select main model",
-    subtitle: "Up/Down/Left/Right selects, Enter confirms, Esc cancels.",
-    options,
-    initialIndex,
-  });
-  if (!selected) return false;
+
+  while (!selected) {
+    const group = await selectModelChoice({
+      title:
+        role === "route"
+          ? "Select route model family"
+          : role === "spare"
+            ? "Select spare model family"
+            : role === "auxiliary"
+              ? "Select auxiliary provider"
+              : "Select main model family",
+      subtitle: "First choose a provider family. Up/Down/Left/Right selects, Enter confirms, Esc cancels.",
+      options: groupOptions,
+      initialIndex: initialGroupIndex,
+    });
+    if (!group) return false;
+    initialGroupIndex = Math.max(
+      groupOptions.findIndex((item) => item.groupId === group.groupId),
+      0
+    );
+
+    const options = role === "auxiliary" ? auxiliaryModelRoleChoices(group.groupId) : textModelRoleChoices(group.groupId);
+    let initialModelIndex = Math.max(
+      options.findIndex((item) => item.provider === currentProvider && item.model === currentModel),
+      0
+    );
+
+    while (!selected) {
+      const modelSelection = await selectModelChoice({
+        title:
+          role === "route"
+            ? `Select route model: ${group.label}`
+            : role === "spare"
+              ? `Select spare model: ${group.label}`
+              : role === "auxiliary"
+                ? `Select auxiliary model: ${group.label}`
+                : `Select main model: ${group.label}`,
+        subtitle:
+          role === "auxiliary"
+            ? "Choose the image/edit auxiliary model. Enter confirms, Esc goes back."
+            : "Choose the text model. OpenAI choices ask for reasoning next. Esc goes back.",
+        options,
+        initialIndex: initialModelIndex,
+        escapeValue: SELECTOR_BACK,
+      });
+      if (isSelectorBack(modelSelection)) break;
+      if (!modelSelection) return false;
+      initialModelIndex = Math.max(
+        options.findIndex((item) => item.provider === modelSelection.provider && item.model === modelSelection.model),
+        0
+      );
+
+      selectedReasoning = modelSelection.reasoningDefault || "";
+      if (Array.isArray(modelSelection.reasoningOptions) && modelSelection.reasoningOptions.length > 0) {
+        const currentReasoning =
+          role === "route"
+            ? state.routeReasoning || selectedReasoning || "medium"
+            : role === "spare"
+              ? state.spareReasoning || selectedReasoning || "medium"
+              : state.mainReasoning || selectedReasoning || "medium";
+        const reasoningOptions = modelSelection.reasoningOptions.map((level) => ({
+          action: "reasoning",
+          label: level,
+          value: level,
+          description: `${modelSelection.label} with ${level} reasoning`,
+        }));
+        const reasoning = await selectModelChoice({
+          title: `Select reasoning: ${modelSelection.label}`,
+          subtitle: "Up/Down selects reasoning effort. Enter confirms, Esc goes back.",
+          options: reasoningOptions,
+          initialIndex: Math.max(
+            reasoningOptions.findIndex((item) => item.value === currentReasoning),
+            reasoningOptions.findIndex((item) => item.value === selectedReasoning),
+            0
+          ),
+          escapeValue: SELECTOR_BACK,
+        });
+        if (isSelectorBack(reasoning)) continue;
+        if (!reasoning) return false;
+        selectedReasoning = reasoning.value;
+      }
+      selected = modelSelection;
+    }
+  }
 
   state.routingMode = "smart";
   state.provider = "deepseek";
@@ -2076,11 +2260,12 @@ async function pickModelRole(role, state) {
   if (role === "route") {
     state.routeProvider = selected.provider;
     state.routeModel = selected.model;
-    printSystemLine(`route=${state.routeProvider}/${state.routeModel}`);
+    state.routeReasoning = selectedReasoning || state.routeReasoning || "";
+    printSystemLine(`route=${state.routeProvider}/${state.routeModel}${state.routeReasoning ? ` reasoning=${state.routeReasoning}` : ""}`);
   } else if (role === "spare") {
     state.spareProvider = selected.provider;
     state.spareModel = selected.model;
-    state.spareReasoning = selected.reasoningDefault || state.spareReasoning || "medium";
+    state.spareReasoning = selectedReasoning || state.spareReasoning || "medium";
     printSystemLine(`spare=${state.spareProvider}/${state.spareModel} reasoning=${state.spareReasoning}`);
   } else if (role === "auxiliary") {
     state.auxiliaryProvider = selected.provider;
@@ -2090,51 +2275,60 @@ async function pickModelRole(role, state) {
   } else {
     state.mainProvider = selected.provider;
     state.mainModel = selected.model;
-    printSystemLine(`main=${state.mainProvider}/${state.mainModel}`);
+    state.mainReasoning = selectedReasoning || state.mainReasoning || "";
+    printSystemLine(`main=${state.mainProvider}/${state.mainModel}${state.mainReasoning ? ` reasoning=${state.mainReasoning}` : ""}`);
   }
   return true;
 }
 
 async function pickVeniceRouteAndMain(state) {
   const options = veniceTextModelChoices();
-  const routeIndex = Math.max(
+  let routeIndex = Math.max(
     options.findIndex((item) => item.provider === state.routeProvider && item.model === state.routeModel),
     0
   );
-  const route = await selectModelChoice({
-    title: "Select Venice route model",
-    subtitle: "Route handles planning and short turns. Up/Down selects, Enter confirms, Esc cancels.",
-    options,
-    initialIndex: routeIndex,
-  });
-  if (!route) return false;
-  if (route.action === "disable") {
-    useDeepSeekDefaults(state);
-    printSystemLine("venice=off routing=smart route=deepseek/deepseek-v4-flash main=deepseek/deepseek-v4-pro");
-    return true;
-  }
+  while (true) {
+    const route = await selectModelChoice({
+      title: "Select Venice route model",
+      subtitle: "Route handles planning and short turns. Up/Down selects, Enter confirms, Esc cancels.",
+      options,
+      initialIndex: routeIndex,
+    });
+    if (!route) return false;
+    routeIndex = Math.max(
+      options.findIndex((item) => item.provider === route.provider && item.model === route.model),
+      0
+    );
+    if (route.action === "disable") {
+      useDeepSeekDefaults(state);
+      printSystemLine("venice=off routing=smart route=deepseek/deepseek-v4-flash main=deepseek/deepseek-v4-pro");
+      return true;
+    }
 
-  const mainIndex = Math.max(
-    options.findIndex((item) => item.action !== "disable" && item.provider === state.mainProvider && item.model === state.mainModel),
-    options.findIndex((item) => item.action !== "disable" && item.model === route.model),
-    0
-  );
-  const main = await selectModelChoice({
-    title: "Select Venice main model",
-    subtitle: "Main handles complex work. Up/Down selects, Enter confirms, Esc cancels.",
-    options,
-    initialIndex: mainIndex,
-  });
-  if (!main) return false;
-  if (main.action === "disable") {
-    useDeepSeekDefaults(state);
-    printSystemLine("venice=off routing=smart route=deepseek/deepseek-v4-flash main=deepseek/deepseek-v4-pro");
+    const mainIndex = Math.max(
+      options.findIndex((item) => item.action !== "disable" && item.provider === state.mainProvider && item.model === state.mainModel),
+      options.findIndex((item) => item.action !== "disable" && item.model === route.model),
+      0
+    );
+    const main = await selectModelChoice({
+      title: "Select Venice main model",
+      subtitle: "Main handles complex work. Up/Down selects, Enter confirms, Esc goes back.",
+      options,
+      initialIndex: mainIndex,
+      escapeValue: SELECTOR_BACK,
+    });
+    if (isSelectorBack(main)) continue;
+    if (!main) return false;
+    if (main.action === "disable") {
+      useDeepSeekDefaults(state);
+      printSystemLine("venice=off routing=smart route=deepseek/deepseek-v4-flash main=deepseek/deepseek-v4-pro");
+      return true;
+    }
+
+    useVeniceModels(state, route.model, main.model);
+    printSystemLine(`venice=on routing=smart route=venice/${state.routeModel} main=venice/${state.mainModel}`);
     return true;
   }
-
-  useVeniceModels(state, route.model, main.model);
-  printSystemLine(`venice=on routing=smart route=venice/${state.routeModel} main=venice/${state.mainModel}`);
-  return true;
 }
 
 async function pickProvider(state) {
@@ -2274,11 +2468,16 @@ async function handleCommand(line, state, packageDir) {
           `Image generation: ${keys.grsai ? "GRSAI key available" : "missing GRSAI key"}`,
           `Venice image: ${keys.venice ? "Venice key available" : "missing Venice key"}`,
           `Selected auxiliary: ${state.auxiliaryProvider || "grsai"}/${state.auxiliaryModel || "nano-banana-2"}`,
-          "Use `/auxiliary grsai` or `/auxiliary venice` to paste a key, `/auxiliary model grsai/nano-banana-2`, `/auxiliary image`, or `/auxiliary off`.",
+          "Use `/auxiliary model` for the two-level selector, `/auxiliary grsai` or `/auxiliary venice` to paste a key, `/auxiliary image`, or `/auxiliary off`.",
         ].join("\n")
       );
       return true;
     }
+    if (action === "select") {
+      const changed = await pickModelRole("auxiliary", state);
+      if (!changed) printSystemLine(`auxiliary=${state.auxiliaryProvider || "grsai"}/${state.auxiliaryModel || "nano-banana-2"}`);
+      return true;
+    }
     if (action === "on") {
       state.allowAuxiliaryTools = true;
       printSystemLine("auxiliary=on");
@@ -2303,7 +2502,7 @@ async function handleCommand(line, state, packageDir) {
         return true;
       }
       const selected = parseProviderModel(modelValue, state.auxiliaryProvider || "grsai");
-      state.auxiliaryProvider = selected.provider || "grsai";
+      state.auxiliaryProvider = selected.provider === "venice-image" ? "venice" : selected.provider || "grsai";
       state.auxiliaryModel = selected.model || state.auxiliaryModel || "nano-banana-2";
       printSystemLine(`auxiliary=${state.auxiliaryProvider}/${state.auxiliaryModel}`);
       return true;
@@ -2312,7 +2511,7 @@ async function handleCommand(line, state, packageDir) {
       await promptAndSaveProviderKey(action === "venice" ? "venice" : "grsai", state);
       return true;
     }
-    printAgentMessage("Usage: /auxiliary [status|grsai|venice|model grsai/nano-banana-2|on|off|image]");
+    printAgentMessage("Usage: /auxiliary [status|select|grsai|venice|model [grsai/nano-banana-2]|on|off|image]");
     return true;
   }
   if (command === "new") {
@@ -2464,10 +2663,12 @@ async function handleCommand(line, state, packageDir) {
       printSystemLine(`routing=${state.routingMode}`);
       return true;
     }
-    const selected = parseProviderModel(value, state.routeProvider || "deepseek");
+    const parts = value.split(/\s+/).filter(Boolean);
+    const selected = parseProviderModel(parts[0] || value, state.routeProvider || "deepseek");
     state.routeProvider = selected.provider || "deepseek";
     state.routeModel = selected.model || state.routeModel || "deepseek-v4-flash";
-    printSystemLine(`route=${state.routeProvider}/${state.routeModel}`);
+    if (isReasoningLevel(parts[1])) state.routeReasoning = parts[1].toLowerCase();
+    printSystemLine(`route=${state.routeProvider}/${state.routeModel}${state.routeReasoning ? ` reasoning=${state.routeReasoning}` : ""}`);
     return true;
   }
   if (command === "main" || command === "model") {
@@ -2486,6 +2687,7 @@ async function handleCommand(line, state, packageDir) {
       if (command === "main") {
         state.mainProvider = "";
         state.mainModel = "";
+        state.mainReasoning = "";
         printSystemLine("main=auto");
       } else {
         state.model = "";
@@ -2493,15 +2695,17 @@ async function handleCommand(line, state, packageDir) {
       }
       return true;
     }
-    const selected = parseProviderModel(value, state.mainProvider || state.provider || "deepseek");
+    const parts = value.split(/\s+/).filter(Boolean);
+    const selected = parseProviderModel(parts[0] || value, state.mainProvider || state.provider || "deepseek");
     if (command === "main" || selected.provider) {
       state.mainProvider = selected.provider || "deepseek";
       state.mainModel = selected.model || "deepseek-v4-pro";
+      if (isReasoningLevel(parts[1])) state.mainReasoning = parts[1].toLowerCase();
       if (command === "model") {
         state.provider = state.mainProvider;
         state.model = state.mainModel;
       }
-      printSystemLine(`main=${state.mainProvider}/${state.mainModel}`);
+      printSystemLine(`main=${state.mainProvider}/${state.mainModel}${state.mainReasoning ? ` reasoning=${state.mainReasoning}` : ""}`);
       return true;
     }
     state.model = value;

package/src/model-routing.js

@@ -88,11 +88,11 @@ export const MODEL_PROVIDER_GROUPS = {
     role: "regional/general",
     description: "DashScope/OpenAI-compatible Qwen route for Chinese and general work.",
   },
-  "venice-uncensored": {
-    label: "Venice Uncensored",
+  venice: {
+    label: "Venice",
     provider: "venice",
     role: "alternate text",
-    description: "Venice uncensored text models for optional manual routes.",
+    description: "Venice 1.2/1.1 and Gemma 4 Uncensored for optional manual routes.",
   },
   "venice-gpt": {
     label: "Venice GPT",
@@ -128,21 +128,46 @@ export const AUXILIARY_MODEL_CATALOG = {
       type: "image",
       description: "Default auxiliary image-generation route through GRS AI-compatible APIs.",
     },
+    {
+      id: "nano-banana-2-edit",
+      label: "Nano Banana 2 Edit",
+      type: "inpaint",
+      description: "Image edit route through GRS AI-compatible APIs when available.",
+    },
     {
       id: "gpt-image-2",
       label: "GPT Image 2",
       type: "image",
       description: "High-quality image generation when available through the configured auxiliary endpoint.",
     },
+    {
+      id: "gpt-image-2-edit",
+      label: "GPT Image 2 Edit",
+      type: "inpaint",
+      description: "High-quality image editing when available through the configured auxiliary endpoint.",
+    },
   ],
   "venice-image": [
+    { id: "wan-2-7-pro-edit", label: "Wan 2.7 Pro Edit", type: "inpaint", price: "$0.09/edit" },
     { id: "nano-banana-2", label: "Nano Banana 2", type: "image", price: "$0.10/image" },
+    { id: "nano-banana-2-edit", label: "Nano Banana 2 Edit", type: "inpaint", price: "$0.10/edit" },
     { id: "gpt-image-2", label: "GPT Image 2", type: "image", price: "$0.27/image" },
     { id: "gpt-image-2-edit", label: "GPT Image 2 Edit", type: "inpaint", price: "$0.36/edit" },
-    { id: "wan-2-7-pro-edit", label: "Wan 2.7 Pro Edit", type: "inpaint", price: "$0.09/edit" },
+    { id: "grok-imagine-image-pro", label: "Grok Imagine Pro", type: "image", price: "$0.09/image" },
     { id: "grok-imagine-image", label: "Grok Imagine", type: "image", price: "$0.03/image" },
+    { id: "wan-2-7-text-to-image", label: "Wan 2.7", type: "image", price: "$0.04/image" },
+    { id: "wan-2-7-pro-text-to-image", label: "Wan 2.7 Pro", type: "image", price: "$0.09/image" },
+    { id: "qwen-image-2", label: "Qwen Image 2", type: "image", price: "$0.05/image" },
     { id: "qwen-image-2-pro", label: "Qwen Image 2 Pro", type: "image", price: "$0.10/image" },
+    { id: "qwen-image-2-edit", label: "Qwen Image 2 Edit", type: "inpaint", price: "$0.05/edit" },
+    { id: "qwen-image-2-pro-edit", label: "Qwen Image 2 Pro Edit", type: "inpaint", price: "$0.10/edit" },
     { id: "bria-bg-remover", label: "Background Remover", type: "image", price: "$0.03/image" },
+    { id: "recraft-v4", label: "Recraft V4", type: "image", price: "$0.05/image" },
+    { id: "recraft-v4-pro", label: "Recraft V4 Pro", type: "image", price: "$0.29/image" },
+    { id: "flux-2-pro", label: "Flux 2 Pro", type: "image", price: "$0.04/image" },
+    { id: "flux-2-max", label: "Flux 2 Max", type: "image", price: "$0.09/image" },
+    { id: "nano-banana-pro", label: "Nano Banana Pro", type: "image", price: "$0.18/image" },
+    { id: "nano-banana-pro-edit", label: "Nano Banana Pro Edit", type: "inpaint", price: "$0.18/edit" },
   ],
 };
 
@@ -211,21 +236,21 @@ export const PROVIDER_MODEL_CATALOG = {
     {
       id: "venice-uncensored-1-2",
       label: "Venice Uncensored 1.2",
-      bucket: "venice-uncensored",
+      bucket: "venice",
       context: "128K",
       description: "Current Venice uncensored text route.",
     },
     {
       id: "venice-uncensored",
       label: "Venice Uncensored 1.1",
-      bucket: "venice-uncensored",
+      bucket: "venice",
       context: "32K",
       description: "Working Venice uncensored 1.1 text route.",
     },
     {
       id: "e2ee-venice-uncensored-24b-p",
       label: "Venice Uncensored 1.1 E2EE",
-      bucket: "venice-uncensored",
+      bucket: "venice",
       context: "32K",
       hidden: true,
       description: "Documented E2EE Venice 1.1 route; hidden from selectors until upstream stabilizes.",
@@ -233,7 +258,7 @@ export const PROVIDER_MODEL_CATALOG = {
     {
       id: "venice-uncensored-role-play",
       label: "Venice Role Play Uncensored",
-      bucket: "venice-uncensored",
+      bucket: "venice",
       context: "128K",
       hidden: true,
       description: "Role-play oriented Venice route.",
@@ -241,10 +266,38 @@ export const PROVIDER_MODEL_CATALOG = {
     {
       id: "gemma-4-uncensored",
       label: "Gemma 4 Uncensored",
-      bucket: "venice-gemma",
+      bucket: "venice",
       context: "256K",
       description: "Gemma-family uncensored Venice route.",
     },
+    {
+      id: "google-gemma-4-31b-it",
+      label: "Google Gemma 4 31B Instruct",
+      bucket: "venice-gemma",
+      context: "256K",
+      description: "Gemma 4 instruct model through Venice.",
+    },
+    {
+      id: "google-gemma-4-26b-a4b-it",
+      label: "Google Gemma 4 26B A4B Instruct",
+      bucket: "venice-gemma",
+      context: "256K",
+      description: "Smaller Gemma 4 instruct model through Venice.",
+    },
+    {
+      id: "google-gemma-3-27b-it",
+      label: "Google Gemma 3 27B Instruct",
+      bucket: "venice-gemma",
+      context: "198K",
+      description: "Gemma 3 instruct model through Venice.",
+    },
+    {
+      id: "e2ee-gemma-3-27b-p",
+      label: "Gemma 3 27B E2EE",
+      bucket: "venice-gemma",
+      context: "40K",
+      description: "Private E2EE Gemma 3 route through Venice.",
+    },
     {
       id: "qwen3-6-27b",
       label: "Qwen 3.6 27B",
@@ -252,6 +305,77 @@ export const PROVIDER_MODEL_CATALOG = {
       context: "256K",
       description: "Qwen-family Venice route.",
     },
+    {
+      id: "qwen-3-6-plus",
+      label: "Qwen 3.6 Plus Uncensored",
+      bucket: "venice-qwen",
+      context: "1.0M",
+      description: "Long-context Qwen 3.6 Plus route through Venice.",
+    },
+    {
+      id: "qwen3-5-9b",
+      label: "Qwen 3.5 9B",
+      bucket: "venice-qwen",
+      context: "256K",
+      description: "Small fast Qwen-family Venice route.",
+    },
+    {
+      id: "qwen3-5-35b-a3b",
+      label: "Qwen 3.5 35B A3B",
+      bucket: "venice-qwen",
+      context: "256K",
+      description: "Mid-size Qwen-family Venice route.",
+    },
+    {
+      id: "qwen3-5-397b-a17b",
+      label: "Qwen 3.5 397B A17B",
+      bucket: "venice-qwen",
+      context: "128K",
+      description: "Large Qwen-family Venice route.",
+    },
+    {
+      id: "qwen3-coder-480b-a35b-instruct-turbo",
+      label: "Qwen 3 Coder 480B Turbo",
+      bucket: "venice-qwen",
+      context: "256K",
+      description: "Coding-oriented Qwen route through Venice.",
+    },
+    {
+      id: "qwen3-vl-235b-a22b",
+      label: "Qwen3 VL 235B",
+      bucket: "venice-qwen",
+      context: "256K",
+      description: "Vision-language Qwen route through Venice.",
+    },
+    {
+      id: "qwen3-235b-a22b-thinking-2507",
+      label: "Qwen 3 235B A22B Thinking",
+      bucket: "venice-qwen",
+      context: "128K",
+      description: "Qwen thinking route through Venice.",
+    },
+    {
+      id: "qwen3-235b-a22b-instruct-2507",
+      label: "Qwen 3 235B A22B Instruct",
+      bucket: "venice-qwen",
+      context: "128K",
+      description: "Qwen instruct route through Venice.",
+    },
+    {
+      id: "qwen3-next-80b",
+      label: "Qwen 3 Next 80B",
+      bucket: "venice-qwen",
+      context: "256K",
+      description: "Qwen Next route through Venice.",
+    },
+    {
+      id: "qwen3-coder-480b-a35b-instruct",
+      label: "Qwen 3 Coder 480B",
+      bucket: "venice-qwen",
+      context: "256K",
+      hidden: true,
+      description: "Deprecated Qwen Coder route; hidden by default.",
+    },
     {
       id: "openai-gpt-55",
       label: "GPT-5.5 via Venice",
@@ -259,6 +383,69 @@ export const PROVIDER_MODEL_CATALOG = {
       context: "1.0M",
       description: "OpenAI-family Venice-routed model.",
     },
+    {
+      id: "openai-gpt-55-pro",
+      label: "GPT-5.5 Pro via Venice",
+      bucket: "venice-gpt",
+      context: "1.0M",
+      description: "High-capacity GPT-5.5 Pro route through Venice.",
+    },
+    {
+      id: "openai-gpt-54",
+      label: "GPT-5.4 via Venice",
+      bucket: "venice-gpt",
+      context: "1.0M",
+      description: "GPT-5.4 route through Venice.",
+    },
+    {
+      id: "openai-gpt-54-pro",
+      label: "GPT-5.4 Pro via Venice",
+      bucket: "venice-gpt",
+      context: "1.0M",
+      description: "High-capacity GPT-5.4 Pro route through Venice.",
+    },
+    {
+      id: "openai-gpt-54-mini",
+      label: "GPT-5.4 Mini via Venice",
+      bucket: "venice-gpt",
+      context: "400K",
+      description: "Small fast GPT route through Venice.",
+    },
+    {
+      id: "openai-gpt-53-codex",
+      label: "GPT-5.3 Codex via Venice",
+      bucket: "venice-gpt",
+      context: "400K",
+      description: "Coding-optimized GPT route through Venice.",
+    },
+    {
+      id: "openai-gpt-52",
+      label: "GPT-5.2 via Venice",
+      bucket: "venice-gpt",
+      context: "256K",
+      description: "Professional-work GPT route through Venice.",
+    },
+    {
+      id: "openai-gpt-52-codex",
+      label: "GPT-5.2 Codex via Venice",
+      bucket: "venice-gpt",
+      context: "256K",
+      description: "Codex-flavored GPT-5.2 route through Venice.",
+    },
+    {
+      id: "openai-gpt-4o-2024-11-20",
+      label: "GPT-4o via Venice",
+      bucket: "venice-gpt",
+      context: "128K",
+      description: "GPT-4o route through Venice.",
+    },
+    {
+      id: "openai-gpt-4o-mini-2024-07-18",
+      label: "GPT-4o Mini via Venice",
+      bucket: "venice-gpt",
+      context: "128K",
+      description: "Small GPT-4o route through Venice.",
+    },
     {
       id: "claude-sonnet-4-6",
       label: "Claude Sonnet 4.6 via Venice",
@@ -266,6 +453,41 @@ export const PROVIDER_MODEL_CATALOG = {
       context: "1.0M",
       description: "Claude-family Venice-routed model.",
     },
+    {
+      id: "claude-opus-4-7",
+      label: "Claude Opus 4.7 via Venice",
+      bucket: "venice-claude",
+      context: "1.0M",
+      description: "High-capacity Claude Opus route through Venice.",
+    },
+    {
+      id: "claude-opus-4-6-fast",
+      label: "Claude Opus 4.6 Fast via Venice",
+      bucket: "venice-claude",
+      context: "1.0M",
+      description: "Fast Claude Opus route through Venice.",
+    },
+    {
+      id: "claude-opus-4-6",
+      label: "Claude Opus 4.6 via Venice",
+      bucket: "venice-claude",
+      context: "1.0M",
+      description: "Claude Opus route through Venice.",
+    },
+    {
+      id: "claude-opus-4-5",
+      label: "Claude Opus 4.5 via Venice",
+      bucket: "venice-claude",
+      context: "198K",
+      description: "Claude Opus 4.5 route through Venice.",
+    },
+    {
+      id: "claude-sonnet-4-5",
+      label: "Claude Sonnet 4.5 via Venice",
+      bucket: "venice-claude",
+      context: "198K",
+      description: "Claude Sonnet 4.5 route through Venice.",
+    },
   ],
   qwen: [
     {
@@ -274,6 +496,12 @@ export const PROVIDER_MODEL_CATALOG = {
       role: "default",
       description: "DashScope OpenAI-compatible route.",
     },
+    {
+      id: "qwen-turbo",
+      label: "Qwen Turbo",
+      role: "fast",
+      description: "Fast DashScope OpenAI-compatible Qwen route.",
+    },
     {
       id: "qwen-max",
       label: "Qwen Max",
@@ -497,6 +725,7 @@ export function getModelRoleDefaults(overrides = {}) {
       command: "/route",
       provider: presets.fast.provider,
       model: presets.fast.model,
+      reasoning: overrides.routeReasoning || process.env.AGINTI_ROUTE_REASONING || "",
       description: "Fast planner and triage model. Default: DeepSeek V4 Flash.",
     },
     main: {
@@ -505,6 +734,7 @@ export function getModelRoleDefaults(overrides = {}) {
       command: "/model",
       provider: presets.complex.provider,
       model: presets.complex.model,
+      reasoning: overrides.mainReasoning || process.env.AGINTI_MAIN_REASONING || "",
       description: "Complex executor for coding, debugging, writing, and long tasks. Default: DeepSeek V4 Pro.",
     },
     spare: {