jinzd-ai-cli 0.4.329 → 0.4.330

package/dist/{doctor-cli-LYZSHYR6.js → doctor-cli-BHS7G2XY.js}

@@ -2,16 +2,16 @@
 import {
   formatDoctorReport,
   runDoctorCli
-} from "./chunk-ABIR2TGW.js";
-import "./chunk-SS3PG7LP.js";
+} from "./chunk-265M6ZLH.js";
+import "./chunk-HMQA2R3O.js";
 import "./chunk-RS7SKYTS.js";
-import "./chunk-I5KLH7TL.js";
-import "./chunk-2CNJ25U7.js";
+import "./chunk-GEC3MEVQ.js";
+import "./chunk-QU74ONFF.js";
 import "./chunk-E2W2MI6H.js";
 import "./chunk-TU3L3PW7.js";
 import "./chunk-HOSJZMQS.js";
 import "./chunk-BMLQPSHP.js";
-import "./chunk-6UDKGZ3D.js";
+import "./chunk-AJITHMRE.js";
 export {
   formatDoctorReport,
   runDoctorCli

package/dist/electron-server.js

@@ -55,7 +55,7 @@ import {
   touchMemoryReferences,
   updateMemoryApproval,
   updateMemoryEntry
-} from "./chunk-LKSMXCFL.js";
+} from "./chunk-2SYDIWRU.js";
 import {
   DEFAULT_PATTERNS,
   redactJson,
@@ -81,7 +81,7 @@ import {
   recoverGbkMojibake,
   runTestsTool,
   shellInvocation
-} from "./chunk-LRRBYGHE.js";
+} from "./chunk-DU44SD6A.js";
 import {
   getCurrentSessionKey,
   getExecutionConfigDir,
@@ -123,7 +123,7 @@ import {
   SUBAGENT_MAX_ROUNDS_LIMIT,
   VERSION,
   buildUserIdentityPrompt
-} from "./chunk-IUC3DRGB.js";
+} from "./chunk-3YVZJHTF.js";
 
 // src/web/access.ts
 import { randomBytes, timingSafeEqual } from "crypto";
@@ -691,8 +691,14 @@ ${err}`
       "deepseek-v4-flash-vision-exp"
     ]);
     const DEEPSEEK_TARGET = "deepseek-flash";
-    const LEGACY_KIMI = /* @__PURE__ */ new Set(["kimi-k2-0711-preview", "kimi-k2-turbo-preview"]);
-    const KIMI_TARGET = "kimi-k2.6";
+    const LEGACY_KIMI = /* @__PURE__ */ new Set([
+      "kimi-k2-0711-preview",
+      "kimi-k2-turbo-preview",
+      "kimi-k2.5",
+      "kimi-k2-thinking",
+      "kimi-k2-thinking-turbo"
+    ]);
+    const KIMI_TARGET = "kimi-k2.7-code";
     const LEGACY_CLAUDE = {
       "claude-sonnet-4-5-20250929": "claude-sonnet-5-5",
       "claude-sonnet-4-5": "claude-sonnet-5-5",
@@ -921,7 +927,6 @@ function deepSeekCnyPricing(cacheRead, input, output) {
     output: output / CNY_PER_USD
   };
 }
-var DEEPSEEK_V4_PRO_ROUTED_TO_FLASH_AT = /* @__PURE__ */ new Date("2026-09-14T04:00:00.000Z");
 var DEEPSEEK_V41_FLASH_ROUTING_CONFIRMED_AT = /* @__PURE__ */ new Date("2026-09-11T00:00:00.000Z");
 function isDeepSeekPeakHour(at) {
   const weekday = at.getUTCDay();
@@ -937,7 +942,6 @@ function getDeepSeekV4Pricing(model, at) {
   const isPro = model.startsWith("deepseek-v4-pro");
   const isFlash = model.startsWith("deepseek-v4-flash");
   if (!isPro && !isFlash) return null;
-  if (isPro && at >= DEEPSEEK_V4_PRO_ROUTED_TO_FLASH_AT) return deepSeekFlash41Pricing(at);
   if (isFlash && at >= DEEPSEEK_V41_FLASH_ROUTING_CONFIRMED_AT) return deepSeekFlash41Pricing(at);
   if (at < DEEPSEEK_TIME_PRICING_EFFECTIVE_AT) {
     return isPro ? deepSeekCnyPricing(0.025, 3, 6) : deepSeekCnyPricing(0.02, 1, 2);
@@ -1016,6 +1020,12 @@ var PRICING_TABLE = {
   // Kimi K3（2026-07 旗舰，1M 上下文）官方定价 platform.kimi.ai/pricing/chat-k3：
   // 输入 cache-miss $3.00 / cache-hit $0.30，输出 $15.00（/MTok）。
   "kimi-k3": { input: 3, output: 15, cacheRead: 0.3 },
+  // K2.6 / K2.7 Code：官方 platform.kimi.ai/docs/pricing/chat（2026-10-18 核对，USD/MTok）。
+  // 国内站人民币价 ¥6.5/¥27（缓存 ¥1.1 / ¥1.3）按 ¥6.9 折算与之相差 <3%，沿用 K3 那行的美元口径。
+  // 此前 K2.6 没有单列，前缀落到 'kimi-k2'（$0.6/$2.5），输出低报约 37%。
+  "kimi-k2.6": { input: 0.95, output: 4, cacheRead: 0.16 },
+  "kimi-k2.7-code": { input: 0.95, output: 4, cacheRead: 0.19 },
+  "kimi-k2.7-code-highspeed": { input: 1.9, output: 8, cacheRead: 0.38 },
   // ── Zhipu GLM ─────────────────────────────────────────────────
   "glm-4-plus": { input: 0.7, output: 0.7 },
   "glm-4": { input: 0.14, output: 0.14 },
@@ -1037,6 +1047,11 @@ var PRICING_TABLE = {
   // 没公布截止日期，静态表表达不了这个时间维度；高估比低估安全——预算告警早触发
   // 好过晚触发。折扣期内 `/cost` 会比实际偏高约一倍，这是刻意的取舍。
   "glm-5.3-flash": { input: 0.116, output: 0.406, cacheRead: 0.033 },
+  // GLM-5.3-FlashX：官方定价页（docs.bigmodel.cn/cn/guide/start/pricing，2026-10-18 浏览器读取）
+  // 输入 2 元、输出 7 元、缓存命中 0.57 元。必须单列——否则前缀匹配会落到上面的 glm-5.3-flash，低报 2.5 倍。
+  "glm-5.3-flashx": { input: 2 / CNY_PER_USD, output: 7 / CNY_PER_USD, cacheRead: 0.57 / CNY_PER_USD },
+  // GLM-4.7-Flash：官方免费（输入/输出/缓存全免）。必须单列——否则前缀落到 'glm-4'（$0.14）。
+  "glm-4.7-flash": { input: 0, output: 0, cacheRead: 0 },
   // GLM-5.2：与 GLM-5.3 同价（同一定价页：8 / 28 / 缓存命中 2 元），口径修正同上。
   "glm-5.2": { input: 8 / CNY_PER_USD, output: 28 / CNY_PER_USD, cacheRead: 2 / CNY_PER_USD },
   // GLM-5V-Turbo（2026-04 视觉编程旗舰）官方直连 API 定价 $1.20 / $4.00。
@@ -1046,8 +1061,26 @@ var PRICING_TABLE = {
   "glm-z1-air": { input: 0.2, output: 0.6 },
   "glm-z1-flash": { input: 0, output: 0 },
   // ── MiniMax (海螺) ────────────────────────────────────────────
-  // GAP (v0.4.184): MiniMax-M3 / M2.x are unpriced — no verified rate anchor yet.
-  // Unpriced → `aicli usage` shows "—" (unknown), not a false $0.00. Add when confirmed.
+  // 官方定价页 platform.minimax.cn/docs/guides/pricing（2026-10-18 核对，元/百万 tokens），按 ¥6.9 折算。
+  // 本 provider 打的是国内站 api.minimaxi.com，按人民币计费。
+  // M3 记的是「永久五折」后的价（官方原文标注永久，不同于智谱的限时折扣），≤512K 输入档；
+  // >512K 那档见 computeCost 的 LONG_PROMPT_TIERS。M3 未公布缓存写入价，缺省按输入价（偏高估）。
+  "minimax-m3": { input: 2.1 / CNY_PER_USD, output: 8.4 / CNY_PER_USD, cacheRead: 0.42 / CNY_PER_USD },
+  "minimax-m2.7": { input: 2.1 / CNY_PER_USD, output: 8.4 / CNY_PER_USD, cacheRead: 0.42 / CNY_PER_USD, cacheWrite: 2.625 / CNY_PER_USD },
+  "minimax-m2.7-highspeed": { input: 4.2 / CNY_PER_USD, output: 16.8 / CNY_PER_USD, cacheRead: 0.42 / CNY_PER_USD, cacheWrite: 2.625 / CNY_PER_USD },
+  // M2.5 / M2.1 / M2（历史模型）同价；highspeed 变体输入输出翻倍、缓存同价。
+  "minimax-m2": { input: 2.1 / CNY_PER_USD, output: 8.4 / CNY_PER_USD, cacheRead: 0.21 / CNY_PER_USD, cacheWrite: 2.625 / CNY_PER_USD },
+  "minimax-m2.5-highspeed": { input: 4.2 / CNY_PER_USD, output: 16.8 / CNY_PER_USD, cacheRead: 0.21 / CNY_PER_USD, cacheWrite: 2.625 / CNY_PER_USD },
+  "minimax-m2.1-highspeed": { input: 4.2 / CNY_PER_USD, output: 16.8 / CNY_PER_USD, cacheRead: 0.21 / CNY_PER_USD, cacheWrite: 2.625 / CNY_PER_USD },
+  // ── Qwen（阿里云百炼，北京地域）───────────────────────────────────
+  // 官方 help.aliyun.com/zh/model-studio/model-pricing（2026-10-18 核对，元/百万 tokens，原价），按 ¥6.9 折算。
+  // 思考与非思考输出同价。qwen3.7-plus 当前限时 8 折，记原价（高估方向，同智谱 glm-5.3-flash 的取舍）。
+  // 官方未给出各模型的缓存单价，不填 cacheRead，命中按输入全价算（高估方向）。
+  // 长提示分档（qwen3.7-plus >256K、qwen-flash >128K / >256K）见 LONG_PROMPT_TIERS；此处为最低档。
+  "qwen3.8-max": { input: 12 / CNY_PER_USD, output: 36 / CNY_PER_USD },
+  "qwen3.7-plus": { input: 2 / CNY_PER_USD, output: 8 / CNY_PER_USD },
+  "qwen3.8-flash": { input: 0.8 / CNY_PER_USD, output: 2.7 / CNY_PER_USD },
+  "qwen-flash": { input: 0.15 / CNY_PER_USD, output: 1.5 / CNY_PER_USD },
   // ── OpenRouter (pass-through — actual cost depends on underlying model) ──
   // 绝大多数 id 仍**刻意留空**：真实价由底层模型 + OpenRouter 加价决定，静态表覆盖不了
   // 396 个模型，未收录的照旧显示 "—"（未知）而不是假的 $0.00。
@@ -1102,11 +1135,21 @@ function getPricing(provider, model, at = /* @__PURE__ */ new Date()) {
   return null;
 }
 var CLAUDE_HAIKU_55_LONG = { input: 0.5, output: 2.5, cacheWrite: 0.625, cacheRead: 0.05 };
-var CLAUDE_HAIKU_55_TIER_TOKENS = 1e5;
+var LONG_PROMPT_TIERS = [
+  { prefix: "claude-haiku-5-5", overTokens: 1e5, pricing: CLAUDE_HAIKU_55_LONG },
+  {
+    prefix: "minimax-m3",
+    overTokens: 512e3,
+    pricing: { input: 4.2 / CNY_PER_USD, output: 16.8 / CNY_PER_USD, cacheRead: 0.84 / CNY_PER_USD }
+  },
+  { prefix: "qwen3.7-plus", overTokens: 256e3, pricing: { input: 6 / CNY_PER_USD, output: 24 / CNY_PER_USD } },
+  { prefix: "qwen-flash", overTokens: 256e3, pricing: { input: 1.2 / CNY_PER_USD, output: 12 / CNY_PER_USD } },
+  { prefix: "qwen-flash", overTokens: 128e3, pricing: { input: 0.6 / CNY_PER_USD, output: 6 / CNY_PER_USD } }
+];
 function longPromptTier(model, usage) {
-  if (!model.toLowerCase().startsWith("claude-haiku-5-5")) return null;
+  const key = model.toLowerCase();
   const prompt = usage.inputTokens + (usage.cacheCreationTokens ?? 0) + (usage.cacheReadTokens ?? 0);
-  return prompt > CLAUDE_HAIKU_55_TIER_TOKENS ? CLAUDE_HAIKU_55_LONG : null;
+  return LONG_PROMPT_TIERS.find((t) => key.startsWith(t.prefix) && prompt > t.overTokens)?.pricing ?? null;
 }
 function computeCost(provider, model, usage, at = /* @__PURE__ */ new Date()) {
   const p = longPromptTier(model, usage) ?? getPricing(provider, model, at);
@@ -4447,7 +4490,9 @@ var DeepSeekProvider = class extends OpenAICompatibleProvider {
       },
       {
         id: "deepseek-v4-pro",
-        displayName: "DeepSeek V4 Pro (text-only; routes to V4.1 Flash from 2026-09-14)",
+        // 2026-09-11 官方撤回了「9-14 起路由到 Flash」，Pro 照常服务、按 Pro 价计费（见 pricing.ts）。
+        // 价格约为 Flash 的 4 倍且不收图片，但 Flash 替代不了它的能力档，故保留。
+        displayName: "DeepSeek V4 Pro (0813, text-only)",
         contextWindow: 1048576,
         supportsStreaming: true,
         supportsThinking: true
@@ -4506,20 +4551,24 @@ var DeepSeekProvider = class extends OpenAICompatibleProvider {
 var ZHIPU_CONFIG = {
   id: "zhipu",
   displayName: "Zhipu (GLM)",
-  defaultModel: "glm-4.6",
+  defaultModel: "glm-5.3-flash",
   apiKeyEnvVar: "AICLI_API_KEY_ZHIPU",
   baseUrl: "https://open.bigmodel.cn/api/paas/v4",
   defaultTimeout: 3e5,
   models: [
     { id: "glm-5.3", displayName: "GLM-5.3 (2026-08 Flagship, 1M, Coding/Security, \u59CB\u7EC8\u601D\u8003)", contextWindow: 1048576, supportsStreaming: true, supportsThinking: true },
+    // GLM-5.3-FlashX（2026-10）：Flash 与 5.3 之间的一档，¥2/¥7、缓存命中 ¥0.57（官方定价页）。
+    // 2026-10-18 实测：同样始终思考（`disabled` → 400/1210），`low` 档 200，base64 图片可识别。
+    { id: "glm-5.3-flashx", displayName: "GLM-5.3 FlashX (\u539F\u751F\u591A\u6A21\u6001, 1M, \u4EF7\u683C\u7EA6\u4E3A 5.3 \u7684 1/4, \u59CB\u7EC8\u601D\u8003)", contextWindow: 1048576, supportsStreaming: true, supportsThinking: true },
     { id: "glm-5.3-flash", displayName: "GLM-5.3 Flash (\u539F\u751F\u591A\u6A21\u6001, 1M, \u8F93\u5165\u7EA6\u4E3A 5.3 \u7684 1/10, \u59CB\u7EC8\u601D\u8003)", contextWindow: 1048576, supportsStreaming: true, supportsThinking: true },
-    { id: "glm-5.2", displayName: "GLM-5.2 (2026-06 Flagship, 1M, Coding)", contextWindow: 1048576, supportsStreaming: true, supportsThinking: true },
-    { id: "glm-5v-turbo", displayName: "GLM-5V-Turbo (Multimodal Vision + Coding)", contextWindow: 202752, supportsStreaming: true },
-    // 2026-10-07 清理：移除 glm-5.1-reasoning / glm-5.1-air（官方模型总览与发布记录里**从未出现过**，
-    // 当初取自第三方标注，同 v0.4.198 的 `glm-5.2[1m]`）与 glm-5 / glm-5.1（已被 glm-5.2 / 5.3 取代，
-    // 要便宜有 glm-5.3-flash；glm-5 这里标的 128K 与官方 200K 也不符）。后两者端点仍可用，`/model` 照收，
-    // pricing.ts 保留其费率供历史成本回算。
-    { id: "glm-4.6", displayName: "GLM-4.6 (200K, \u4E2D\u6587\u5199\u4F5C\u63A8\u8350)", contextWindow: 204800, supportsStreaming: true }
+    // 免费（输入/输出/缓存全免，官方定价页「更多文本模型」）。2026-10-18 实测 200，但慢（一句 ok 约 15s）。
+    { id: "glm-4.7-flash", displayName: "GLM-4.7 Flash (\u514D\u8D39, 200K)", contextWindow: 2e5, supportsStreaming: true }
+    // 2026-10-18 按「老模型没有价格优势就下架」清理（端点仍可用，`/model` 照收，pricing.ts 保留费率供历史回算）：
+    //   - glm-5.2：与 glm-5.3 同价（¥8/¥28/缓存 ¥2），已被取代
+    //   - glm-5v-turbo：¥5/¥22 起，比同样原生多模态的 glm-5.3-flash（¥0.8/¥2.8）贵 6–8 倍
+    //   - glm-4.6：已不在官方定价页上，且旧价高于 glm-5.3-flash
+    // 更早的 2026-10-07 清理：glm-5.1-reasoning / glm-5.1-air 官方从未存在（取自第三方标注，同 v0.4.198
+    // 的 `glm-5.2[1m]`），glm-5 / glm-5.1 已被取代。
   ]
 };
 var REASONING_EFFORT_MODELS = /^glm-5/i;
@@ -4585,7 +4634,10 @@ var KimiProvider = class _KimiProvider extends OpenAICompatibleProvider {
    */
   static TEMP_1_MODELS = /* @__PURE__ */ new Set([
     "kimi-k3",
+    "kimi-k2.7-code",
+    "kimi-k2.7-code-highspeed",
     "kimi-k2.6",
+    // 已 404 的旧 code 仍留在这里：config 自愈迁移之外的显式 `/model kimi-k2.5` 不该先撞 temperature 400
     "kimi-k2.5",
     "kimi-k2-thinking",
     "kimi-k2-thinking-turbo"
@@ -4606,7 +4658,7 @@ var KimiProvider = class _KimiProvider extends OpenAICompatibleProvider {
   info = {
     id: "kimi",
     displayName: "Kimi (Moonshot AI)",
-    defaultModel: "kimi-k2.6",
+    defaultModel: "kimi-k2.7-code",
     apiKeyEnvVar: "AICLI_API_KEY_KIMI",
     requiresApiKey: true,
     baseUrl: this.defaultBaseUrl,
@@ -4616,39 +4668,33 @@ var KimiProvider = class _KimiProvider extends OpenAICompatibleProvider {
       //   - contextWindow 1M（1048576）；code 就是 `kimi-k3`（UI 名 "K3 · Max" ≠ API code）
       //   - temperature 固定 1.0（已并入 TEMP_1_MODELS）
       //   - vision 走 base64 / ms://file-id（不收公网 URL）——本版未接图片输入路径，仅文本/工具
-      // 默认仍保持 kimi-k2.6（K3 输出 $15/MTok 偏贵，按需 /model kimi-k3）。
+      // 默认用 kimi-k2.7-code（K3 输出 $15/MTok，约为它的 4 倍，按需 /model kimi-k3）。
       {
         id: "kimi-k3",
         displayName: "Kimi K3",
         contextWindow: 1048576,
         supportsStreaming: true
       },
-      // Kimi K2.6 / K2.5 系列（最新一代，推荐）
-      {
-        id: "kimi-k2.6",
-        displayName: "Kimi K2.6",
-        contextWindow: 262144,
-        supportsStreaming: true
-      },
-      {
-        id: "kimi-k2.5",
-        displayName: "Kimi K2.5",
-        contextWindow: 262144,
-        supportsStreaming: true
-      },
-      // Kimi K2 Thinking 系列（支持多步推理 + 工具调用）
+      // Kimi K2.7 Code（官方 platform.kimi.com/docs/guide/kimi-k2-7-code-quickstart，2026-10-18 核对）：
+      //   - 256K 上下文；虽名为 Coding 模型，官方写明也支持对话与 Agent 任务
+      //   - 思考默认开启且**关不掉**（关闭报错）——本 provider 不发 thinking 字段，正好不踩
+      //   - temperature 固定 1.0（已并入 TEMP_1_MODELS）；tool_choice 只收 auto / none
+      //   - 价格与 K2.6 相同（$0.95/$4.00），各项基准提升 10–30%，故取代 K2.6 成为默认
       {
-        id: "kimi-k2-thinking",
-        displayName: "Kimi K2 Thinking",
+        id: "kimi-k2.7-code",
+        displayName: "Kimi K2.7 Code",
         contextWindow: 262144,
         supportsStreaming: true
       },
+      // 同一模型的高速版：输出约 5–6 倍速，价格翻倍。
       {
-        id: "kimi-k2-thinking-turbo",
-        displayName: "Kimi K2 Thinking Turbo",
+        id: "kimi-k2.7-code-highspeed",
+        displayName: "Kimi K2.7 Code Highspeed (2x price, ~5x speed)",
         contextWindow: 262144,
         supportsStreaming: true
       }
+      // 2026-10-18 下架：kimi-k2.6 与 K2.7 Code 同价、已被取代（端点仍可用，`/model` 照收）；
+      // kimi-k2.5 / kimi-k2-thinking / kimi-k2-thinking-turbo 端点已 404，由 config 自愈迁移到 kimi-k2.7-code。
     ]
   };
   /**
@@ -4760,11 +4806,10 @@ var MINIMAX_CONFIG = {
   apiKeyEnvVar: "AICLI_API_KEY_MINIMAX",
   baseUrl: "https://api.minimaxi.com/v1",
   models: [
-    { id: "MiniMax-M3", displayName: "MiniMax M3\uFF08\u65D7\u8230\uFF0C1M \u4E0A\u4E0B\u6587\uFF09", contextWindow: 1e6, supportsStreaming: true },
-    { id: "MiniMax-M2.7", displayName: "MiniMax M2.7", contextWindow: 1e6, supportsStreaming: true },
-    { id: "MiniMax-M2.7-highspeed", displayName: "MiniMax M2.7 Highspeed", contextWindow: 1e6, supportsStreaming: true },
-    { id: "MiniMax-M2.5", displayName: "MiniMax M2.5", contextWindow: 1e6, supportsStreaming: true },
-    { id: "MiniMax-M2.5-highspeed", displayName: "MiniMax M2.5 Highspeed", contextWindow: 1e6, supportsStreaming: true }
+    { id: "MiniMax-M3", displayName: "MiniMax M3\uFF08\u65D7\u8230\uFF0C1M \u4E0A\u4E0B\u6587\uFF09", contextWindow: 1e6, supportsStreaming: true }
+    // 2026-10-18 按「老模型没有价格优势就下架」清理（官方定价页 platform.minimax.cn/docs/guides/pricing）：
+    // M3「永久五折」后 ¥2.1/¥8.4，与 M2.7 同价、比 M2.x highspeed 便宜一半；M2.5 / M2.1 / M2 官方已归入
+    // 「历史模型」。端点仍可用，`/model` 照收，pricing.ts 保留费率供历史回算。
   ]
 };
 var MiniMaxProvider = createStaticProviderClass(MINIMAX_CONFIG);
@@ -4773,14 +4818,20 @@ var MiniMaxProvider = createStaticProviderClass(MINIMAX_CONFIG);
 var QWEN_CONFIG = {
   id: "qwen",
   displayName: "Qwen (Alibaba Cloud)",
-  defaultModel: "qwen-plus",
+  defaultModel: "qwen3.8-flash",
   apiKeyEnvVar: "AICLI_API_KEY_QWEN",
   baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1",
+  // 2026-10-18 按百炼官方文档更新（help.aliyun.com/zh/model-studio/models、/model-pricing、/deep-thinking）。
+  // **未经真实端点实测**（本机无 Qwen key）。三款 qwen3.x 默认开启思考、支持非流式调用、纯文本输入；
+  // 思考与非思考输出同价。上下文按官方价格阶梯的上限 1M 记。
   models: [
-    { id: "qwen-plus", displayName: "Qwen Plus", contextWindow: 131072, supportsStreaming: true },
-    { id: "qwen-max", displayName: "Qwen Max", contextWindow: 131072, supportsStreaming: true },
-    { id: "qwen-turbo", displayName: "Qwen Turbo", contextWindow: 131072, supportsStreaming: true },
-    { id: "qwen-coder-plus", displayName: "Qwen Coder Plus", contextWindow: 131072, supportsStreaming: true }
+    { id: "qwen3.8-max", displayName: "Qwen3.8 Max\uFF08\u65D7\u8230\uFF0C1M\uFF0C\xA512/\xA536\uFF09", contextWindow: 1e6, supportsStreaming: true },
+    { id: "qwen3.7-plus", displayName: "Qwen3.7 Plus\uFF081M\uFF0C\u2264256K \xA52/\xA58\uFF09", contextWindow: 1e6, supportsStreaming: true },
+    { id: "qwen3.8-flash", displayName: "Qwen3.8 Flash\uFF081M\uFF0C\xA50.8/\xA52.7\uFF09", contextWindow: 1e6, supportsStreaming: true },
+    // 老模型里唯一有价格优势的：≤128K ¥0.15/¥1.5，最便宜。默认不开思考。
+    { id: "qwen-flash", displayName: "Qwen Flash\uFF08\u6700\u4FBF\u5B9C\uFF0C\u2264128K \xA50.15/\xA51.5\uFF09", contextWindow: 1e6, supportsStreaming: true }
+    // 下架（端点若仍可用，`/model` 照收）：qwen-max ¥2.4/¥9.6 比 qwen3.7-plus 还贵；qwen-plus 价格不如
+    // qwen-flash、能力不如 qwen3.8-flash；qwen-turbo / qwen-coder-plus 已不在官方价格页上。
   ]
 };
 var QwenProvider = createStaticProviderClass(QWEN_CONFIG);
@@ -17834,7 +17885,7 @@ Cache:  write=${cacheCreate} read=${cacheRead}` : "";
 Cost:   ${formatCost(cost)}` : "";
   let memoryLine = "";
   try {
-    const { countPendingMemories: countPendingMemories2 } = await import("./persistent-memory-OCDFR4BH.js");
+    const { countPendingMemories: countPendingMemories2 } = await import("./persistent-memory-3XTSZ3QR.js");
     const pending = countPendingMemories2(ctx.config.getConfigDir());
     if (pending > 0) memoryLine = `
 Memory: \u26A0 ${pending} pending approval \u2014 see the Memory panel or /memory`;
@@ -18822,7 +18873,7 @@ async function handleTest(args, ctx) {
       const isCommand = argStr.includes(" ") || /^(mvn|gradle|npm|pytest|cargo|go)\b/.test(argStr);
       testArgs = isCommand ? { command: argStr } : { filter: argStr };
     }
-    const runTests = ctx.runTests ?? (await import("./run-tests-LTYJACRX.js")).executeTests;
+    const runTests = ctx.runTests ?? (await import("./run-tests-WQE3XTUL.js")).executeTests;
     const report = await runTests(testArgs);
     ctx.send({ type: "info", message: report });
   } catch (err) {
@@ -19213,7 +19264,7 @@ async function handleMemory(args, ctx) {
   }
   if (intent.kind === "status") {
     const { getChatIndexStatus: getChatIndexStatus2 } = await import("./chat-index-RLVLQRRC.js");
-    const { listMemoryEntries: listMemoryEntries2, isMemoryExpired: isMemoryExpired2, formatMemoryForPrompt: formatMemoryForPrompt2 } = await import("./persistent-memory-OCDFR4BH.js");
+    const { listMemoryEntries: listMemoryEntries2, isMemoryExpired: isMemoryExpired2, formatMemoryForPrompt: formatMemoryForPrompt2 } = await import("./persistent-memory-3XTSZ3QR.js");
     const memories = listMemoryEntries2(configDir, { includeExpired: true, includeRejected: true });
     const injected = formatMemoryForPrompt2(configDir, process.cwd(), ctx.config.get("memory")?.maxChars);
     const view = buildMemoryStatusView({
@@ -19236,7 +19287,7 @@ async function handleMemory(args, ctx) {
     return;
   }
   if (intent.kind === "recall") {
-    const { searchPersistentMemories: searchPersistentMemories2, touchMemoryReferences: touchMemoryReferences2 } = await import("./persistent-memory-OCDFR4BH.js");
+    const { searchPersistentMemories: searchPersistentMemories2, touchMemoryReferences: touchMemoryReferences2 } = await import("./persistent-memory-3XTSZ3QR.js");
     const { searchChatMemory: searchChatMemory2 } = await import("./chat-index-RLVLQRRC.js");
     const persistentHits = searchPersistentMemories2(configDir, intent.query, process.cwd(), 3);
     if (persistentHits.length > 0) {
@@ -19357,7 +19408,7 @@ async function handleMemory(args, ctx) {
     ctx.handleMemoryManage(intent.action, args[1], intent.action === "expire" ? args[2] : void 0);
   } else if (intent.kind === "export") {
     if (args[1] === "md") {
-      const { exportMemoryEntries: exportMemoryEntries2 } = await import("./persistent-memory-OCDFR4BH.js");
+      const { exportMemoryEntries: exportMemoryEntries2 } = await import("./persistent-memory-3XTSZ3QR.js");
       ctx.send({
         type: "export_data",
         format: "md",
@@ -20363,7 +20414,7 @@ No tools match "${filter}".
           const { join: join26 } = await import("path");
           const { existsSync: existsSync32 } = await import("fs");
           const { getGitRoot: getGitRoot2 } = await import("./git-context-VNCA2GU7.js");
-          const { MCP_PROJECT_CONFIG_NAME: MCP_PROJECT_CONFIG_NAME2 } = await import("./constants-BYWEJLAI.js");
+          const { MCP_PROJECT_CONFIG_NAME: MCP_PROJECT_CONFIG_NAME2 } = await import("./constants-J5SJ232X.js");
           const { approveProject, hashMcpFile } = await import("./project-trust-6CJB5ZMX.js");
           const cwd = process.cwd();
           const projectRoot = getGitRoot2(cwd) ?? cwd;
@@ -21282,7 +21333,7 @@ function createReviewCommands() {
       usage: "/test [command|filter]",
       async execute(args, ctx) {
         try {
-          const { executeTests } = await import("./run-tests-LTYJACRX.js");
+          const { executeTests } = await import("./run-tests-WQE3XTUL.js");
           const argStr = args.join(" ").trim();
           let testArgs = {};
           if (argStr) {

package/dist/{hub-GRHYVLGV.js → hub-PPG4N6QQ.js}

@@ -158,7 +158,7 @@ ${content}`);
   }
 }
 async function runTaskMode(config, providers, configManager, topic) {
-  const { TaskOrchestrator } = await import("./task-orchestrator-URFLWSPI.js");
+  const { TaskOrchestrator } = await import("./task-orchestrator-CI7EDT25.js");
   const orchestrator = new TaskOrchestrator(config, providers, configManager);
   let interrupted = false;
   const onSigint = () => {

package/dist/index.js

@@ -24,12 +24,12 @@ import {
   setupProxy,
   snapshotFile,
   stripRoutingTags
-} from "./chunk-LSSOPLKJ.js";
+} from "./chunk-LLXHA2LA.js";
 import {
   CostTracker,
   getPricing,
   setCostCurrency
-} from "./chunk-BRZZ3QAI.js";
+} from "./chunk-2BAFLDDL.js";
 import {
   ToolExecutor,
   ToolRegistry,
@@ -47,7 +47,7 @@ import {
   setMaxOutputCap,
   spawnAgentContext,
   undoStack
-} from "./chunk-MPCQG7LV.js";
+} from "./chunk-M5B6G6AY.js";
 import "./chunk-HLWUDRBO.js";
 import {
   NO_THINKING,
@@ -60,7 +60,7 @@ import {
   countPendingMemories,
   detectStaleLegacyMarkdownEdit,
   formatMemoryForPrompt
-} from "./chunk-XF2S7LNM.js";
+} from "./chunk-DX2U6QK4.js";
 import "./chunk-RWRZVQCB.js";
 import "./chunk-B3MPYI4E.js";
 import "./chunk-M6HLYG33.js";
@@ -81,22 +81,22 @@ import {
 import {
   loadContextFiles,
   writeCrashLog
-} from "./chunk-ABIR2TGW.js";
+} from "./chunk-265M6ZLH.js";
 import {
   getActivePluginAssets,
   getPendingHookTrust,
   installFlushOnExit,
   mergePluginHooks,
   runLifecycleHooks
-} from "./chunk-SS3PG7LP.js";
+} from "./chunk-HMQA2R3O.js";
 import "./chunk-RS7SKYTS.js";
 import {
   ProviderRegistry,
   listBuiltInProviderMetadata
-} from "./chunk-I5KLH7TL.js";
+} from "./chunk-GEC3MEVQ.js";
 import {
   ConfigManager
-} from "./chunk-2CNJ25U7.js";
+} from "./chunk-QU74ONFF.js";
 import {
   CONTENT_ONLY_STREAM_REMINDER,
   TEE_FINAL_USER_NUDGE,
@@ -134,7 +134,7 @@ import {
   isInterrupted,
   requestInterrupt,
   resetInterrupt
-} from "./chunk-DRNRQ6SD.js";
+} from "./chunk-GWEBO4RF.js";
 import {
   AGENTIC_BEHAVIOR_GUIDELINE,
   AUTHOR,
@@ -152,7 +152,7 @@ import {
   SKILLS_DIR_NAME,
   VERSION,
   buildUserIdentityPrompt
-} from "./chunk-6UDKGZ3D.js";
+} from "./chunk-AJITHMRE.js";
 import "./chunk-SXIJLSVF.js";
 
 // src/index.ts
@@ -1079,6 +1079,7 @@ function getVisionModelHint(model, provider) {
   if (bare.includes("vision") || bare.includes("vl") || bare.includes("5v") || // glm-5v-turbo
   bare.endsWith(".6v") || // glm-4.6v
   bare === "glm-5.3-flash" || // GLM-5 系列首个原生多模态（名字里没有 v，得单独认）
+  bare === "glm-5.3-flashx" || // 同为原生多模态（2026-10-18 实测 base64 图片可识别）
   // DeepSeek V4.1 Flash（2026-09）原生多模态。放在这条通用规则里而不是只写进下面的
   // deepseek 分支，是为了让 OpenRouter 的 `deepseek/deepseek-flash` 也一并放行——
   // 按 provider 分支写就会重演 v0.4.295「同一个坑换个前缀又漏一次」。
@@ -1091,7 +1092,7 @@ function getVisionModelHint(model, provider) {
     return null;
   }
   if (provider === "zhipu") {
-    return `model "${model}" does not support images. Use /model glm-5.3-flash (or glm-4.6v)`;
+    return `model "${model}" does not support images. Use /model glm-5.3-flash (or glm-5.3-flashx)`;
   }
   if (provider === "deepseek") {
     if (bare.startsWith("deepseek-v4-flash")) return null;
@@ -4117,7 +4118,7 @@ program.command("web").description("Start Web UI server with browser-based chat
     console.error("Error: Invalid port number. Must be between 1 and 65535.");
     process.exit(1);
   }
-  const { startWebServer, WebStartupError } = await import("./server-2NZN22XP.js");
+  const { startWebServer, WebStartupError } = await import("./server-5NQSYQ2T.js");
   try {
     await startWebServer({ port, host: options.host, externalOrigin: options.externalOrigin });
   } catch (err) {
@@ -4294,11 +4295,11 @@ program.command("sessions").description("List recent conversation sessions").opt
   console.log(footer + "\n");
 });
 program.command("usage").description("Show token + cost usage grouped by provider/model (cross-session)").option("--days <n>", "Only the last N days (inclusive of today)").option("--month <ym>", "Only a specific month, format YYYY-MM (e.g. 2026-06)").option("--json", "Output as JSON (for scripting)").action(async (options) => {
-  const { runUsageCli } = await import("./usage-IP6G4Q2K.js");
+  const { runUsageCli } = await import("./usage-AJYZ5A7P.js");
   await runUsageCli(options);
 });
 program.command("doctor").description("Health check: API keys, config, MCP, recent crashes, tool usage, disk usage").option("--json", "Output as JSON (for scripting)").option("--reset-stats", "Reset accumulated tool usage statistics").action(async (options) => {
-  const { runDoctorCli } = await import("./doctor-cli-LYZSHYR6.js");
+  const { runDoctorCli } = await import("./doctor-cli-BHS7G2XY.js");
   const argv = process.argv.slice(2);
   await runDoctorCli({
     json: !!options.json || argv.includes("--json"),
@@ -4307,7 +4308,7 @@ program.command("doctor").description("Health check: API keys, config, MCP, rece
 });
 program.command("batch <action> [arg] [arg2]").description("Anthropic Message Batches: submit | list | status <id> | results <id> [out] | cancel <id>").option("--dry-run", "Parse and validate input without submitting (submit only)").action(async (action, arg, arg2, options) => {
   try {
-    const batch = await import("./batch-NNYRIS4U.js");
+    const batch = await import("./batch-GXITPLER.js");
     switch (action) {
       case "submit":
         if (!arg) {
@@ -4350,7 +4351,7 @@ program.command("batch <action> [arg] [arg2]").description("Anthropic Message Ba
   }
 });
 program.command("mcp-serve").description("Start an MCP server over STDIO, exposing aicli's built-in tools to Claude Desktop / Cursor / other MCP clients").option("--allow-destructive", "Allow bash / run_interactive / task_create (always destructive in MCP mode)").option("--allow-outside-cwd", "Allow tool path arguments to escape the sandbox root \u2014 disabled by default").option("--tools <list>", "Comma-separated whitelist of tools to expose (default: all eligible tools)").option("--cwd <path>", "Working directory AND sandbox root (default: current directory)").action(async (options) => {
-  const { startMcpServer } = await import("./server-ZTFB3KMS.js");
+  const { startMcpServer } = await import("./server-J2B26DZK.js");
   await startMcpServer({
     allowDestructive: !!options.allowDestructive,
     allowOutsideCwd: !!options.allowOutsideCwd,
@@ -4363,7 +4364,7 @@ program.command("pr <action> [target]").description("PR review helpers: review |
     console.error(`Unknown pr action: ${action}. Use review | security-review | summarize.`);
     process.exit(1);
   }
-  const { runPrCli } = await import("./pr-LDUIDU27.js");
+  const { runPrCli } = await import("./pr-M6ZY4CEK.js");
   const result = await runPrCli({
     action,
     target,
@@ -4383,8 +4384,8 @@ program.command("pr <action> [target]").description("PR review helpers: review |
   process.exit(result.exitCode);
 });
 program.command("ci").description("Headless PR review (code + security) \u2014 reads git/gh diff, optionally posts to PR. Designed for GitHub Actions.").option("--pr <num>", "PR number or URL; diff fetched via gh pr diff <value>", (v) => /^#?\d+$/.test(v) ? parseInt(v.replace(/^#/, ""), 10) : v).option("--base <ref>", "Base ref for git diff <ref>...HEAD (ignored when --pr set)").option("--post", "Post review as a PR comment (requires gh CLI + GH_TOKEN, needs --pr)").option("--no-update", "Always create a new comment instead of updating the previous aicli review").option("--skip-code", "Skip the code review section").option("--skip-security", "Skip the security review section").option("--detailed", "Use the detailed code-review prompt").option("--max-diff <n>", "Max diff chars sent to the model (default 30000)", (v) => parseInt(v, 10)).option("--provider <id>", "Override provider (default: config.defaultProvider)").option("--model <id>", "Override model").option("--dry-run", "Print result to stdout instead of posting (overrides --post)").option("--json", "Output a structured JSON report").option("--markdown", "Output Markdown report (default)").option("--sarif", "Output SARIF 2.1.0 for code scanning artifacts").option("--fail-on-security-high", "Fail when the security review contains HIGH/CRITICAL findings (default)").option("--no-fail-on-security-high", "Do not fail on security HIGH/CRITICAL findings").option("--fail-on-test-failure", "Fail when the report mentions test failures (default)").option("--no-fail-on-test-failure", "Do not fail on test failure findings").option("--fail-on-lint-failure", "Fail when the report mentions lint/typecheck failures (default)").option("--no-fail-on-lint-failure", "Do not fail on lint/typecheck failure findings").action(async (options) => {
-  const { runCi } = await import("./ci-ODKA57IG.js");
-  const { formatCiResult } = await import("./ci-format-QDC4XKYZ.js");
+  const { runCi } = await import("./ci-6K3JNHB4.js");
+  const { formatCiResult } = await import("./ci-format-2L3RFJVH.js");
   const formats = [options.json, options.markdown, options.sarif].filter(Boolean).length;
   if (formats > 1) {
     console.error("Choose only one output format: --json, --markdown, or --sarif.");
@@ -4543,7 +4544,7 @@ program.command("hub [topic]").description("Start multi-agent hub (discuss / bra
     config.get("customProviders"),
     config.getConfigDir()
   );
-  const { startHub } = await import("./hub-GRHYVLGV.js");
+  const { startHub } = await import("./hub-PPG4N6QQ.js");
   await startHub(
     {
       topic: topic ?? "",

package/dist/{persistent-memory-OCDFR4BH.js → persistent-memory-3XTSZ3QR.js}

@@ -27,11 +27,11 @@ import {
   touchMemoryReferences,
   updateMemoryApproval,
   updateMemoryEntry
-} from "./chunk-LKSMXCFL.js";
+} from "./chunk-2SYDIWRU.js";
 import "./chunk-FSC6KEWU.js";
 import "./chunk-2XMUCC42.js";
 import "./chunk-CGLYJRFA.js";
-import "./chunk-IUC3DRGB.js";
+import "./chunk-3YVZJHTF.js";
 export {
   addMemoryEntry,
   backupMemoryStore,

package/dist/{persistent-memory-O6GWHGJH.js → persistent-memory-O4T2OTVV.js}

@@ -28,11 +28,11 @@ import {
   touchMemoryReferences,
   updateMemoryApproval,
   updateMemoryEntry
-} from "./chunk-XF2S7LNM.js";
+} from "./chunk-DX2U6QK4.js";
 import "./chunk-3FEZSLGI.js";
 import "./chunk-HOSJZMQS.js";
 import "./chunk-BMLQPSHP.js";
-import "./chunk-6UDKGZ3D.js";
+import "./chunk-AJITHMRE.js";
 export {
   addMemoryEntry,
   backupMemoryStore,

package/dist/{pr-LDUIDU27.js → pr-M6ZY4CEK.js}

@@ -2,8 +2,8 @@
 import {
   countSeverity,
   runCi
-} from "./chunk-TYQODNI7.js";
-import "./chunk-YDH5ZWOQ.js";
+} from "./chunk-LUJ4JVMI.js";
+import "./chunk-CCFCPAKN.js";
 import {
   buildReviewPrompt,
   buildSecurityReviewPrompt,
@@ -14,16 +14,16 @@ import {
 } from "./chunk-6YZEZFX5.js";
 import {
   ProviderRegistry
-} from "./chunk-I5KLH7TL.js";
+} from "./chunk-GEC3MEVQ.js";
 import {
   ConfigManager
-} from "./chunk-2CNJ25U7.js";
+} from "./chunk-QU74ONFF.js";
 import "./chunk-E2W2MI6H.js";
 import "./chunk-TU3L3PW7.js";
 import "./chunk-BMLQPSHP.js";
 import {
   VERSION
-} from "./chunk-6UDKGZ3D.js";
+} from "./chunk-AJITHMRE.js";
 
 // src/cli/pr.ts
 import { execFileSync } from "child_process";

package/dist/{run-tests-YTRU22LY.js → run-tests-H3DIUCDB.js}

@@ -3,8 +3,8 @@ import {
   executeTests,
   runTestProcess,
   runTestsTool
-} from "./chunk-DRNRQ6SD.js";
-import "./chunk-6UDKGZ3D.js";
+} from "./chunk-GWEBO4RF.js";
+import "./chunk-AJITHMRE.js";
 import "./chunk-SXIJLSVF.js";
 export {
   executeTests,