dsh-llm-workbuddy 0.1.13 → 0.1.15

package/CHANGELOG.md

@@ -4,6 +4,48 @@
 每次发版请同步 `package.json` 的 `version` 并打一个 `git tag`（如 `v0.1.13`），
 在 GitHub 创建 Release 时本文件即为更新说明来源。
 
+## [0.1.15] - 2026-09-10
+
+### Fixed
+- **刷新模型看不到新上架模型**：内置目录此前同时充当白名单，代理
+  `/v1/models` 里凡不在目录中的模型一律被丢弃，所以平台新上架的模型
+  （如 `deepseek-v4.1-flash`）刷新多少次都不会出现。现改为**黑名单**
+  `RETIRED_MODEL_IDS`：只隐藏实测不可用（上游 `service info not found`）的
+  旧模型 id，其余照单显示。顺带放出 7 个别名仍在服务、却被旧白名单误藏的模型
+  （`minimax-m2.7`、`glm-5.0-turbo`、`hy3-preview`、`deepseek-v3-1-lkeap`、
+  `deepseek-v3-0324-lkeap`、`deepseek-r1-0528-lkeap`、`hunyuan-2.0-instruct`）。
+- 内置目录新增 `deepseek-v4.1-flash`，并在代理不可达时作为兜底条目。
+- 🔍 诊断的「模型出字」误报修复：探针用 `hy3` + `max_tokens: 4` 发一次请求，
+  而推理模型会把这 4 个 token 全花在思考上、`content` 返回空串，于是
+  "代理正常"被误判成「代理返回了空响应」。现承认"有计费 completion token +
+  有 choice"即为正常。
+
+### Added
+- **模型名后显示积分消耗倍率**：读取代理 `/v1/models` 的 `credits` 字段
+  （`"x0.06 credits"`）并拼成 `Deepseek-V4-Flash ×0.06`；实时值优先，内置目录
+  `credits` 兜底。平台未声明倍率的模型只显示模型名——不会用 0 或猜测值冒充。
+- 内置目录补齐各模型 `credits`（取自 WorkBuddy 应用 `product.json`）。
+
+### Changed（代理侧，`.workbuddy-src` / workbuddy2api）
+- `/v1/models` 透出 `credits` 字段（`model_to_codex_format`），倍率不再由各客户端
+  各自维护。
+- 官方应用 `product.json` 与本地 `models_config.json` 合并时倍率单独回填：官方
+  条目为空则沿用本地倍率，避免官方合并把倍率抹平。
+- `models_config.json` 补充 `deepseek-v4.1-flash`（1M 上下文 / 50k 输出）。
+
+## [0.1.14] - 2026-08-28
+
+### Changed
+- 状态胶囊精简：外面只保留一个 **⚙️ 设置** 按钮，点击展开菜单显示原来的
+  四个操作（🔍 诊断 / 📊 用量 / 🎁 签到 / 🔄 刷新模型），每个都带明确 title。
+- 浮层（诊断 / 用量）统一加 **右上角 ✕ 关闭按钮**，并支持**点击浮层外区域**与
+  **按 Esc** 关闭，不再只能"再点一次菜单项"才收起。
+- 菜单交互修复（与本次功能一并发布）：
+  - 点击设置按钮可正常展开/收起（解决 `render()` 重建 DOM 触发外部点击误关的问题，
+    给设置按钮加 `stopPropagation()`）。
+  - 点击菜单项后立即收起菜单（此前只改状态不重渲染，要等下次轮询才消失）。
+  - 点击页面空白处亦可收起。
+
 ## [0.1.13] - 2026-08-28
 
 ### Added
@@ -37,6 +79,8 @@
 ---
 
 <!-- 历史版本锚点（便于生成 Release 时对比区间） -->
+[0.1.15]: https://github.com/zdk119746/dsh-llm-workbuddy/compare/v0.1.14...v0.1.15
+[0.1.14]: https://github.com/zdk119746/dsh-llm-workbuddy/compare/v0.1.13...v0.1.14
 [0.1.13]: https://github.com/zdk119746/dsh-llm-workbuddy/compare/v0.1.12...v0.1.13
 [0.1.12]: https://github.com/zdk119746/dsh-llm-workbuddy/compare/v0.1.11...v0.1.12
 [0.1.11]: https://github.com/zdk119746/dsh-llm-workbuddy/releases/tag/v0.1.11

package/README.md

@@ -213,13 +213,17 @@ Web 登录后终端脚本也读得到同一份会话。
 - 启动时在 `document.body` 末尾挂一个 `position:fixed` 的胶囊（右下角）。
 - 每 **5 秒** `GET /api/workbuddy/status`；点「登录」后加快到每 **2 秒**轮询、
   最多 30 次，直到 `authenticated:true`。
-- 胶囊里有个 **🔍 诊断** 按钮：点它 `POST /api/workbuddy/diagnose`，弹出一个面板
+- 胶囊外面只显示一个 **⚙️ 设置** 按钮（带 title「WorkBuddy 设置」）。点击展开菜单，
+  内含四个带明确 title 的操作：**🔍 诊断 / 📊 用量 / 🎁 签到 / 🔄 刷新模型**；
+  点击页面空白或选中某项后菜单自动收起。
+- **🔍 诊断**：点它 `POST /api/workbuddy/diagnose`，弹出一个面板
   显示**真实健康状态**（登录、会话文件、代理进程、登录令牌、模型能否出字），
   发现问题时附带**可复制的重启命令**（一键复制到终端执行）。
-- 胶囊里有个 **📊 用量** 按钮：点它 `GET /api/workbuddy/usage`，弹出一个面板
+- **📊 用量**：点它 `GET /api/workbuddy/usage`，弹出一个面板
   显示**今日/累计 token 用量**、**积分消耗**（上游每次返回的 `credit` 累加）与
   **按模型明细**（数据来自本地台账 `$DSH_HOME/llm-workbuddy/usage.jsonl`）。
   注：代理不暴露余额/剩余积分接口，只能统计**已消耗**积分，无法显示账户剩余。
+- 诊断 / 用量浮层**右上角都有 ✕ 关闭按钮**，也可点击浮层外区域或按 Esc 关闭。
 - 状态映射：
   - `authenticated && proxyUp` → 🟢 绿，显示 `WorkBuddy · <昵称>`
   - 否则 → 🔴 红，显示「登录」按钮；`proxyUp` 为 false 时额外提示 `代理未运行`
@@ -280,16 +284,29 @@ DSH 的 `dsh.client` 机制只要求 `package.json` 里：
     # discovery: true                   # 实时拉取代理的 /v1/models（30s 缓存）
     # models: [...]                     # 静态目录（代理不可达时的兜底）
     #   每个条目可带 reasoningEffort: 'low'|'medium'|'high'（默认推理等级）
+    #   以及 credits: 'x0.06 credits'（显示为模型名后的 ×0.06）
     # loginScript: ''                   # 登录脚本绝对/相对路径；默认用包内 login_workbuddy.py
     # sessionFile: ~/.codebuddy-session.json  # 会话文件路径；默认同上
 ```
 
 模型列表默认取插件内置目录；代理可达时改为实时拉取 `/v1/models`（支持
 `{"models": [...]}` / `{"data": [...]}` 两种返回），未列出的模型 id 仍可原样
-传递。**内置目录充当白名单**：代理返回的已下架/不可用模型（如 glm-4.6v、
-glm-5.0、minimax-m2.5 等，上游返回 `service info not found`）会被过滤，不显示
-在 UI 中。每个模型的推理等级（`reasoningEffort`）默认来自内置目录，代理可达时
-优先沿用目录值（代理 `/v1/models` 当前只上报 `High`，不区分模型）。
+传递。**内置目录不再是白名单**：平台新上架的模型刷新后即出现在选择器里，
+无需等插件发版；只有已下架/不可用的模型 id（`RETIRED_MODEL_IDS`，上游返回
+`service info not found`）会被隐藏。每个模型的推理等级（`reasoningEffort`）
+默认来自内置目录，代理可达时优先沿用目录值（代理 `/v1/models` 当前只上报
+`High`，不区分模型）。
+
+**积分倍率**：模型名后面会拼上平台声明的积分倍率，例如
+`Deepseek-V4-Flash ×0.06`。倍率来自代理 `/v1/models` 的 `credits` 字段
+（`"x0.06 credits"`，平台写在 WorkBuddy 应用 `product.json` /
+`models_config.json` 里），实时值优先、内置目录兜底；平台没声明的模型
+（如个别新模型尚未写入 `product.json`）只显示模型名，不会猜一个数字。
+
+**刷新模型**：⚙️ 设置 → 🔄 刷新模型 会清掉插件的 30s 发现缓存并重新拉取
+代理列表。注意代理自身的模型清单来自本机 WorkBuddy 应用的 `product.json`
+与 `models_config.json`——两者都没有的模型，刷新也不会出现，需先更新应用或
+把模型补进代理配置。
 
 ---
 

package/lib/client.js

@@ -73,13 +73,34 @@ function runWidget() {
     ".wb-btn{margin-left:4px;border:0;border-radius:6px;background:#2563eb;color:#fff;" +
     "padding:3px 8px;cursor:pointer;font:inherit;}" +
     ".wb-btn:disabled{opacity:.6;cursor:default;}" +
-    ".wb-warn{margin-left:6px;color:#f59e0b;font-size:11px;}";
+    ".wb-warn{margin-left:6px;color:#f59e0b;font-size:11px;}" +
+    ".wb-settle{position:relative;}" +
+    ".wb-setbtn{margin-left:4px;border:0;border-radius:6px;background:#374151;color:#e5e7eb;" +
+    "padding:3px 8px;cursor:pointer;font:inherit;}" +
+    ".wb-menu{position:absolute;right:0;bottom:38px;display:flex;flex-direction:column;gap:4px;" +
+    "background:#0f172a;border:1px solid #334155;border-radius:8px;padding:6px;min-width:148px;" +
+    "box-shadow:0 6px 20px rgba(0,0,0,.5);}" +
+    ".wb-menu .wb-item{display:flex;align-items:center;gap:6px;width:100%;text-align:left;" +
+    "border:0;border-radius:6px;background:#1e293b;color:#e2e8f0;padding:6px 8px;cursor:pointer;" +
+    "font:inherit;white-space:nowrap;}" +
+    ".wb-menu .wb-item:hover{background:#334155;}" +
+    ".wb-menu .wb-item .wb-ico{font-size:13px;}" +
+    ".wb-menu .wb-item:disabled{opacity:.6;cursor:default;}" +
+    ".wb-panel{position:fixed;right:12px;bottom:44px;z-index:2147483647;width:340px;max-height:340px;overflow:auto;" +
+    "background:#0f172a;color:#e2e8f0;border:1px solid #334155;border-radius:8px;padding:10px;" +
+    "font:12px/1.5 system-ui,sans-serif;box-shadow:0 6px 20px rgba(0,0,0,.5);}" +
+    ".wb-panel-head{display:flex;align-items:center;justify-content:space-between;margin-bottom:6px;}" +
+    ".wb-panel-head .wb-ttl{font-weight:700;}" +
+    ".wb-panel .wb-close{border:0;background:transparent;color:#94a3b8;cursor:pointer;" +
+    "font-size:14px;line-height:1;padding:2px 6px;border-radius:4px;}" +
+    ".wb-panel .wb-close:hover{background:#1e293b;color:#e2e8f0;}";
   document.head.appendChild(style);
   document.body.appendChild(host);
 
   var status = null;
   var busy = false;
   var timer = null;
+  var menuOpen = false;
 
   function fmtExpiry(expiresAt) {
     if (!expiresAt) return "";
@@ -105,35 +126,60 @@ function runWidget() {
       html += '<button class="wb-btn" id="wb-login"' + disabled + ">" +
         (busy ? "登录中…" : "登录") + "</button>";
     }
-    // 诊断按钮始终显示，便于随时自查真实健康状态。
-    html += '<button class="wb-btn wb-diagnose" id="wb-diagnose" title="一键诊断">🔍</button>';
-    // 用量按钮：查看今日/累计 token 用量与模型明细。
-    html += '<button class="wb-btn wb-usage" id="wb-usage" title="用量统计">📊</button>';
-    // 签到按钮：✅ 今日已处理 / 🎁 今日未处理（点击立即领取；10 点后也会自动领取）。
-    var ck = status.checkin;
-    if (ck) {
-      var ckMsg = ck.lastResult && ck.lastResult.message ? "\n" + ck.lastResult.message : "";
-      var ckTitle = ck.handledToday ? "今日已处理签到" + ckMsg : "今日还没签到（每天 10 点后自动领取）\n点击立即领取";
-      var ckIcon = ck.handledToday && ck.lastResult && ck.lastResult.ok ? "✅" : "🎁";
-      html += '<button class="wb-btn wb-checkin" id="wb-checkin" title="' + esc(ckTitle) + '">' + ckIcon + "</button>";
+    // 只外显一个设置按钮：点击展开诊断/用量/签到/刷新模型四个操作。
+    html += '<span class="wb-settle">' +
+      '<button class="wb-setbtn" id="wb-settings" title="WorkBuddy 设置（诊断 / 用量 / 签到 / 刷新模型）">⚙️</button>';
+    if (menuOpen) {
+      // 签到按钮：✅ 今日已处理 / 🎁 今日未处理（点击立即领取；10 点后也会自动领取）。
+      var ck = status.checkin;
+      var ckTitle = "今日还没签到（每天 10 点后自动领取）\n点击立即领取";
+      var ckIcon = "🎁";
+      var ckDisabled = "";
+      if (ck) {
+        var ckMsg = ck.lastResult && ck.lastResult.message ? "\n" + ck.lastResult.message : "";
+        ckTitle = ck.handledToday ? "今日已处理签到" + ckMsg : ckTitle;
+        ckIcon = ck.handledToday && ck.lastResult && ck.lastResult.ok ? "✅" : "🎁";
+        ckDisabled = ck.handledToday ? " disabled" : "";
+      }
+      html += '<span class="wb-menu">' +
+        '<button class="wb-item" id="wb-diagnose" title="一键诊断：真实探测登录、代理、令牌与模型出字状态">🔍 诊断</button>' +
+        '<button class="wb-item" id="wb-usage" title="用量统计：查看今日/累计 token 与积分消耗、按模型明细">📊 用量</button>' +
+        '<button class="wb-item" id="wb-checkin" title="' + esc(ckTitle) + '"' + ckDisabled + '>' + ckIcon + " 签到</button>" +
+        '<button class="wb-item" id="wb-models" title="刷新模型列表：从代理重读模型并让模型选择器立即重载">🔄 刷新模型</button>' +
+        "</span>";
     }
-    // 刷新模型按钮：从代理重读模型列表（跟随 WorkBuddy 应用更新）并让模型选择器立即重载。
-    html += '<button class="wb-btn wb-models" id="wb-models" title="刷新模型列表">🔄</button>';
+    html += "</span>";
     if (!status.proxyUp) html += '<span class="wb-warn">代理未运行</span>';
     html += "</div>";
     host.innerHTML = html;
     var btn = host.querySelector("#wb-login");
     if (btn) btn.addEventListener("click", onLogin);
+    var setBtn = host.querySelector("#wb-settings");
+    if (setBtn) setBtn.addEventListener("click", function (e) {
+      // 阻止冒泡：render() 会重建 host 的 DOM，被点按钮随即脱离文档，
+      // 否则 document 的外部点击监听器会误判为"点在外面"而立刻关掉菜单。
+      e.stopPropagation();
+      menuOpen = !menuOpen;
+      render();
+    });
     var dg = host.querySelector("#wb-diagnose");
-    if (dg) dg.addEventListener("click", onDiagnose);
+    if (dg) dg.addEventListener("click", function () { menuOpen = false; render(); onDiagnose(); });
     var us = host.querySelector("#wb-usage");
-    if (us) us.addEventListener("click", onUsage);
+    if (us) us.addEventListener("click", function () { menuOpen = false; render(); onUsage(); });
     var ckBtn = host.querySelector("#wb-checkin");
-    if (ckBtn) ckBtn.addEventListener("click", onCheckin);
+    if (ckBtn) ckBtn.addEventListener("click", function () { menuOpen = false; render(); onCheckin(); });
     var mdBtn = host.querySelector("#wb-models");
-    if (mdBtn) mdBtn.addEventListener("click", onRefreshModels);
+    if (mdBtn) mdBtn.addEventListener("click", function () { menuOpen = false; render(); onRefreshModels(); });
   }
 
+  // 点击页面其它位置关闭设置菜单。
+  document.addEventListener("click", function (e) {
+    if (!menuOpen) return;
+    if (host.contains(e.target)) return;
+    menuOpen = false;
+    render();
+  });
+
   function refresh() {
     return fetch("/api/workbuddy/status")
       .then(function (r) { return r.ok ? r.json() : null; })
@@ -177,15 +223,7 @@ function runWidget() {
   function onDiagnose() {
     var panel = document.getElementById("wb-diagnose-panel");
     if (panel) { panel.remove(); return; }
-    var el = document.createElement("div");
-    el.id = "wb-diagnose-panel";
-    el.style.cssText =
-      "position:fixed;right:12px;bottom:44px;z-index:2147483647;width:340px;max-height:340px;overflow:auto;" +
-      "background:#0f172a;color:#e2e8f0;border:1px solid #334155;border-radius:8px;" +
-      "padding:10px;font:12px/1.5 system-ui,sans-serif;box-shadow:0 6px 20px rgba(0,0,0,.5);";
-    el.innerHTML = '<div class="wb-dg-title" style="font-weight:700;margin-bottom:6px">WorkBuddy 诊断中…</div>' +
-      '<div class="wb-dg-body" style="white-space:pre-wrap;word-break:break-word;color:#94a3b8">正在探测…</div>';
-    document.body.appendChild(el);
+    var el = makePanel("wb-diagnose-panel", "WorkBuddy 诊断");
 
     fetch("/api/workbuddy/diagnose", { method: "POST" })
       .then(function (r) { return r.json(); })
@@ -199,18 +237,20 @@ function runWidget() {
         lines.push("登录令牌: " + (d.health && d.health.tokenValid ? "有效" : "无效"));
         lines.push("模型出字: " + (d.chat && d.chat.chatWorking ? "正常" : "失败"));
         if (d.chat && d.chat.chatError) lines.push("出字错误: " + d.chat.chatError);
-        var html = '<div class="wb-dg-title" style="font-weight:700;margin-bottom:6px">' + title + '</div>' +
-          '<div class="wb-dg-body" style="white-space:pre-wrap;word-break:break-word">' + lines.map(esc).join("\n") + "</div>";
+        var body = el.querySelector(".wb-panel-body");
+        var html = '<div style="font-weight:700;margin-bottom:4px">' + esc(title) + '</div>' +
+          '<div>' + lines.map(esc).join("\n") + "</div>";
         if (!d.ok && d.restartCommand) {
           html += '<div style="margin-top:8px;font-weight:700">重启命令（复制到终端执行）：</div>' +
             '<pre style="background:#1e293b;border-radius:6px;padding:8px;font-size:11px;overflow:auto;white-space:pre-wrap">' +
             esc(d.restartCommand) + "</pre>" +
             '<button class="wb-btn" id="wb-copy-cmd">复制重启命令</button>';
         }
-        el.innerHTML = html;
+        body.innerHTML = html;
         var copy = document.getElementById("wb-copy-cmd");
         if (copy) {
-          copy.addEventListener("click", function () {
+          copy.addEventListener("click", function (e) {
+            e.stopPropagation();
             navigator.clipboard.writeText(d.restartCommand).then(function () {
               copy.textContent = "已复制 ✓";
               setTimeout(function () { copy.textContent = "复制重启命令"; }, 1500);
@@ -219,8 +259,8 @@ function runWidget() {
         }
       })
       .catch(function (err) {
-        el.innerHTML = '<div class="wb-dg-title" style="font-weight:700;color:#f87171">诊断失败</div>' +
-          '<div class="wb-dg-body">' + esc(String(err && err.message || err)) + "</div>";
+        el.querySelector(".wb-panel-body").innerHTML =
+          '<div style="font-weight:700;color:#f87171">诊断失败</div><div>' + esc(String(err && err.message || err)) + "</div>";
       });
   }
 
@@ -228,15 +268,7 @@ function runWidget() {
   function onUsage() {
     var panel = document.getElementById("wb-usage-panel");
     if (panel) { panel.remove(); return; }
-    var el = document.createElement("div");
-    el.id = "wb-usage-panel";
-    el.style.cssText =
-      "position:fixed;right:12px;bottom:44px;z-index:2147483647;width:340px;max-height:340px;overflow:auto;" +
-      "background:#0f172a;color:#e2e8f0;border:1px solid #334155;border-radius:8px;" +
-      "padding:10px;font:12px/1.5 system-ui,sans-serif;box-shadow:0 6px 20px rgba(0,0,0,.5);";
-    el.innerHTML = '<div class="wb-usage-title" style="font-weight:700;margin-bottom:6px">WorkBuddy 用量统计</div>' +
-      '<div class="wb-usage-body" style="color:#94a3b8">加载中…</div>';
-    document.body.appendChild(el);
+    var el = makePanel("wb-usage-panel", "WorkBuddy 用量统计");
 
     fetch("/api/workbuddy/usage")
       .then(function (r) { return r.ok ? r.json() : null; })
@@ -260,13 +292,12 @@ function runWidget() {
         var note = hasCredit
           ? "数据来源：本地用量台账（含上游每次返回的积分消耗；代理不暴露余额接口，无法显示剩余积分）"
           : "数据来源：本地 token 用量台账（" + d.total.requests + " 次累计）；当前模型为免费/折扣档，上游回报积分为 0";
-        el.innerHTML = '<div class="wb-usage-title" style="font-weight:700;margin-bottom:6px">WorkBuddy 用量统计</div>' +
-          '<div class="wb-usage-body" style="white-space:pre-wrap;word-break:break-word">' + esc(lines.join("\n")) + "</div>" +
-          '<div style="margin-top:8px;color:#64748b;font-size:11px">' + esc(note) + "</div>";
+        el.querySelector(".wb-panel-body").innerHTML =
+          esc(lines.join("\n")) + '<div style="margin-top:8px;color:#64748b;font-size:11px">' + esc(note) + "</div>";
       })
       .catch(function (err) {
-        el.innerHTML = '<div class="wb-usage-title" style="font-weight:700;color:#f87171">用量加载失败</div>' +
-          '<div class="wb-usage-body">' + esc(String(err && err.message || err)) + "</div>";
+        el.querySelector(".wb-panel-body").innerHTML =
+          '<div style="font-weight:700;color:#f87171">用量加载失败</div><div>' + esc(String(err && err.message || err)) + "</div>";
       });
   }
 
@@ -282,6 +313,35 @@ function runWidget() {
     return (Math.round(v * 100) / 100).toString();
   }
 
+  /**
+   * 创建一个可关闭的浮层（诊断/用量面板共用）。右上角带 ❌，点击浮层外区域或按
+   * Esc 均可关闭。返回的 el 由调用方填充内容；再次点击对应菜单项也可关闭
+   * （调用方已对 getElementById(id) 做了存在即 remove 的处理）。
+   * @param {string} id  浮层 DOM id（如 "wb-usage-panel"）
+   * @param {string} title 标题文字
+   * @returns {HTMLElement}
+   */
+  function makePanel(id, title) {
+    var el = document.createElement("div");
+    el.id = id;
+    el.className = "wb-panel";
+    el.innerHTML =
+      '<div class="wb-panel-head"><span class="wb-ttl">' + esc(title) + "</span>" +
+      '<button class="wb-close" title="关闭">✕</button></div>' +
+      '<div class="wb-panel-body" style="white-space:pre-wrap;word-break:break-word;color:#94a3b8">加载中…</div>';
+    document.body.appendChild(el);
+    var close = function () { el.remove(); document.removeEventListener("click", onDocClick, true); document.removeEventListener("keydown", onKey); };
+    el.querySelector(".wb-close").addEventListener("click", close);
+    var onDocClick = function (e) { if (!el.contains(e.target)) close(); };
+    var onKey = function (e) { if (e.key === "Escape") close(); };
+    // 用捕获阶段，避免被状态胶囊的 document 点击监听（菜单外部关闭）提前吞掉。
+    setTimeout(function () {
+      document.addEventListener("click", onDocClick, true);
+      document.addEventListener("keydown", onKey);
+    }, 0);
+    return el;
+  }
+
   /** 手动领取今日签到（幂等：官方对已签到返回业务拒绝，不会出错）。 */
   function onCheckin() {
     var btn = host.querySelector("#wb-checkin");

package/lib/index.js

@@ -77,26 +77,53 @@ const DISCOVERY_TIMEOUT_MS = 2_000;
  * CodeBuddy/WorkBuddy platform (from workbuddy2api's models_config.json).
  * It is advisory and replaced by the live `/v1/models` answer whenever the
  * proxy is reachable; entries not announced by the proxy are still listed.
+ *
+ * `credits` is the platform's declared multiplier (`"x0.06 credits"`), shown
+ * after the model name. The live answer wins when it carries one — this copy
+ * only covers the proxy-down / discovery-off fallback.
  */
 const DEFAULT_MODELS = [
-  { id: "deepseek-v4-pro", name: "Deepseek-V4-Pro", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
-  { id: "deepseek-v4-flash", name: "Deepseek-V4-Flash", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
-  { id: "deepseek-v3-2-volc", name: "DeepSeek-V3.2", contextWindow: 96_000, maxTokens: 32_000, reasoningEffort: "medium" },
-  { id: "glm-5.2", name: "GLM-5.2", contextWindow: 1_000_000, maxTokens: 48_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "glm-5.1", name: "GLM-5.1", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "medium" },
-  { id: "glm-5v-turbo", name: "GLM-5v-Turbo", contextWindow: 200_000, maxTokens: 64_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "kimi-k3-1", name: "Kimi-K3", contextWindow: 1_000_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "kimi-k2.7", name: "Kimi-K2.7-Code", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "kimi-k2.6", name: "Kimi-K2.6", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "kimi-k2.5", name: "Kimi-K2.5", contextWindow: 164_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "minimax-m3", name: "MiniMax-M3", contextWindow: 512_000, maxTokens: 128_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "hy3", name: "Hy3", contextWindow: 192_000, maxTokens: 64_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
-  { id: "hunyuan-2.0-thinking", name: "Hunyuan-2.0-Thinking", contextWindow: 128_000, maxTokens: 24_000, reasoningEffort: "medium" },
+  { id: "deepseek-v4.1-flash", name: "Deepseek-V4.1-Flash", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
+  { id: "deepseek-v4-pro", name: "Deepseek-V4-Pro", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high", credits: "x0.16 credits" },
+  { id: "deepseek-v4-flash", name: "Deepseek-V4-Flash", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high", credits: "x0.06 credits" },
+  { id: "deepseek-v3-2-volc", name: "DeepSeek-V3.2", contextWindow: 96_000, maxTokens: 32_000, reasoningEffort: "medium", credits: "x0.29 credits" },
+  { id: "glm-5.2", name: "GLM-5.2", contextWindow: 1_000_000, maxTokens: 48_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x0.79 credits" },
+  { id: "glm-5.1", name: "GLM-5.1", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "medium", credits: "x0.79 credits" },
+  { id: "glm-5v-turbo", name: "GLM-5v-Turbo", contextWindow: 200_000, maxTokens: 64_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x0.95 credits" },
+  { id: "kimi-k3-1", name: "Kimi-K3", contextWindow: 1_000_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x1.62 credits" },
+  { id: "kimi-k2.7", name: "Kimi-K2.7-Code", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x0.57 credits" },
+  { id: "kimi-k2.6", name: "Kimi-K2.6", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x0.52 credits" },
+  { id: "kimi-k2.5", name: "Kimi-K2.5", contextWindow: 164_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x0.45 credits" },
+  { id: "minimax-m3", name: "MiniMax-M3", contextWindow: 512_000, maxTokens: 128_000, inputModalities: ["text", "image"], reasoningEffort: "medium", credits: "x0.25 credits" },
+  { id: "hy3", name: "Hy3", contextWindow: 192_000, maxTokens: 64_000, inputModalities: ["text", "image"], reasoningEffort: "high", credits: "x0.00 credits" },
+  { id: "hunyuan-2.0-thinking", name: "Hunyuan-2.0-Thinking", contextWindow: 128_000, maxTokens: 24_000, reasoningEffort: "medium", credits: "x0.04 credits" },
   { id: "hunyuan-chat", name: "Hunyuan-Turbos", contextWindow: 200_000, maxTokens: 8_192 },
   { id: "auto", name: "Auto", contextWindow: 168_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
-  { id: "default", name: "Default", contextWindow: 200_000, maxTokens: 24_000 },
+  { id: "default", name: "Default", contextWindow: 200_000, maxTokens: 24_000, credits: "x2.00 credits" },
 ];
 
+/**
+ * Models the proxy still announces but that the upstream platform rejects with
+ * `service info not found` (each id re-probed 2026-09-10 against the live
+ * endpoint). The catalog is no longer a whitelist — a model the platform
+ * launches shows up after a refresh without a plugin update — so dead ids are
+ * hidden here by name instead.
+ */
+const RETIRED_MODEL_IDS = new Set([
+  "minimax-m2.5",
+  "glm-5.0",
+  "glm-4.7",
+  "glm-4.6",
+  "glm-4.6v",
+  "kimi-k2-thinking",
+  "deepseek-v3-1",
+  "deepseek-v3-1-volc",
+  "kimi-k2-instruct-taiji",
+  "completion-gf",
+  "default-1.1",
+  "default-1.2",
+]);
+
 /** Selectable reasoning efforts exposed to the harness UI, in display order. */
 const REASONING_EFFORTS = [
   { id: "low", name: "Low" },
@@ -156,6 +183,26 @@ function liveInputModalities(entry) {
   return valid.length > 0 ? [...new Set(valid)] : ["text"];
 }
 
+/**
+ * Parse the platform's credit declaration into the `×0.06` suffix appended to
+ * a model name. The platform writes these as `"x0.06 credits"` (product.json /
+ * models_config.json), and the proxy relays the string verbatim; a bare number
+ * or an already-compact `×1.62` also parses. Anything without a recognizable
+ * number yields `undefined`, so callers omit the suffix rather than guess —
+ * an unknown multiplier must not be rendered as 0.
+ */
+function creditsMultiplier(value) {
+  if (typeof value !== "string") return undefined;
+  const match = /^[x×*]?\s*(\d+(?:\.\d+)?)\s*(?:credits?)?$/i.exec(value.trim());
+  return match === null ? undefined : `×${match[1]}`;
+}
+
+/** Append the credit multiplier to a model's display name, when declared. */
+function withCredits(name, credits) {
+  const multiplier = creditsMultiplier(credits);
+  return multiplier === undefined ? name : `${name} ${multiplier}`;
+}
+
 
 // #region serialize
 
@@ -580,7 +627,7 @@ function modelInfo(provider, model) {
   return {
     provider,
     id: model.id,
-    name: model.name ?? model.id,
+    name: withCredits(model.name ?? model.id, model.credits),
     ...(model.description === undefined ? {} : { description: model.description }),
     inputModalities: model.inputModalities ?? ["text"],
     ...(reasoning === undefined ? {} : { reasoning }),
@@ -673,6 +720,10 @@ export class WorkBuddyAdapter extends LlmAdapter {
           description: typeof entry.description === "string" && entry.description.length > 0
             ? entry.description
             : undefined,
+          // Platform-declared credit multiplier, relayed by the proxy.
+          credits: typeof entry.credits === "string" && entry.credits.trim().length > 0
+            ? entry.credits.trim()
+            : undefined,
           modalities: liveInputModalities(entry),
           contextWindow: Number.isInteger(entry.context_window) && entry.context_window > 0
             ? entry.context_window
@@ -689,19 +740,23 @@ export class WorkBuddyAdapter extends LlmAdapter {
         if (seen.has(entry.id)) continue;
         seen.add(entry.id);
         const catalog = byId.get(entry.id);
-        // The static catalog doubles as a whitelist. The proxy still announces
-        // retired/legacy models (deepseek-v3-1, glm-4.6, glm-4.6v, kimi-k2,
-        // kimi-k2-thinking, minimax-m2.5, hunyuan-image-v3.0, ...) that the
-        // upstream rejects with `service info not found`. Drop any live entry
-        // that is not in the catalog so those never reach the UI; the catalog
-        // is the curated set that matches what the official client shows.
-        if (catalog === undefined) continue;
+        // Retired ids are hidden by name (RETIRED_MODEL_IDS) rather than by
+        // absence from the catalog: the proxy still announces models upstream
+        // rejects, but a model the platform launches must appear after a
+        // refresh without waiting for a new plugin release. The catalog is now
+        // only a metadata overlay plus a fallback for unannounced entries.
+        if (RETIRED_MODEL_IDS.has(entry.id)) continue;
         const effort = liveReasoningEffort(entry, catalog?.reasoningEffort);
         const reasoning = modelReasoningInfo(effort);
         merged.push({
           provider,
           id: entry.id,
-          name: entry.name ?? catalog?.name ?? entry.id,
+          // The proxy's multiplier tracks the platform; the catalog's is a
+          // snapshot, so the live value wins and the catalog only fills gaps.
+          name: withCredits(
+            entry.name ?? catalog?.name ?? entry.id,
+            entry.credits ?? catalog?.credits,
+          ),
           ...(catalog?.description !== undefined || entry.description !== undefined
             ? { description: catalog?.description ?? entry.description }
             : {}),
@@ -843,6 +898,9 @@ const catalogModel = z.object({
   maxTokens: z.number().step(1).min(1),
   inputModalities: z.array(z.union(MODEL_MODALITIES)).min(1).default(["text"]),
   reasoningEffort: z.string(),
+  // Platform-declared credit multiplier, e.g. "x0.06 credits"; displayed after
+  // the model name as "×0.06".
+  credits: z.string(),
 });
 
 export const Config = z.object({
@@ -890,6 +948,10 @@ function resolveModels(models) {
     if (reasoningEffort !== undefined && !REASONING_EFFORT_IDS.has(reasoningEffort)) {
       throw new Error(`dsh-llm-workbuddy: catalog model "${model.id}" reasoningEffort must be one of low|medium|high`);
     }
+    const credits = typeof model.credits === "string" ? model.credits.trim() : undefined;
+    if (credits !== undefined && credits.length > 0 && creditsMultiplier(credits) === undefined) {
+      throw new Error(`dsh-llm-workbuddy: catalog model "${model.id}" credits must read like "x0.06 credits"`);
+    }
     return {
       id: model.id,
       ...(model.name === undefined ? {} : { name: model.name }),
@@ -898,6 +960,7 @@ function resolveModels(models) {
       ...(model.maxTokens === undefined ? {} : { maxTokens: model.maxTokens }),
       inputModalities: [...inputModalities],
       ...(reasoningEffort === undefined ? {} : { reasoningEffort }),
+      ...(credits === undefined || credits.length === 0 ? {} : { credits }),
     };
   });
 }
@@ -1234,10 +1297,19 @@ async function probeChatWorking(baseURL) {
       return { chatWorking: false, chatError: `HTTP ${response.status}${text ? `: ${text.slice(0, 200)}` : ""}` };
     }
     const body = await response.json().catch(() => null);
-    const content = body?.choices?.[0]?.message?.content;
+    const choice = body?.choices?.[0];
+    const content = choice?.message?.content;
     if (typeof content === "string" && content.length > 0) {
       return { chatWorking: true };
     }
+    // A reasoning model spends the whole 4-token cap on thinking and answers
+    // with `content: ""` + `finish_reason: "length"` — that is a healthy proxy,
+    // not an empty response (probed against hy3, which is all-thinking at this
+    // cap). Any billed completion token proves the upstream round trip worked.
+    const completionTokens = body?.usage?.completion_tokens;
+    if (choice !== undefined && typeof completionTokens === "number" && completionTokens > 0) {
+      return { chatWorking: true };
+    }
     return { chatWorking: false, chatError: "代理返回了空响应" };
   } catch (error) {
     return { chatWorking: false, chatError: String(error?.message ?? error) };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-llm-workbuddy",
-  "version": "0.1.13",
+  "version": "0.1.15",
   "description": "WorkBuddy (via the local workbuddy2api proxy) LLM provider adapter for DeepSeek Harness, with a Web login-status widget",
   "type": "module",
   "main": "lib/index.js",