dsh-llm-workbuddy 0.1.6 → 0.1.7

package/README.md

@@ -1,5 +1,8 @@
 # dsh-llm-workbuddy
 
+[![npm version](https://img.shields.io/npm/v/dsh-llm-workbuddy.svg)](https://www.npmjs.com/package/dsh-llm-workbuddy)
+[![license](https://img.shields.io/npm/l/dsh-llm-workbuddy.svg)](https://github.com/zdk119746/dsh-llm-workbuddy/blob/main/LICENSE)
+
 在 DeepSeek Harness 中使用你的 **WorkBuddy / CodeBuddy** 账号模型的 LLM 适配器插件。
 
 它把 `workbuddy` 这个 provider 路由指向本地运行的
@@ -12,7 +15,8 @@
 
 1. **模型能力**：Web 界面的模型选择器（composer 模型菜单或 `/model` 命令）会多出
    一个 **WorkBuddy** 分组，模型（DeepSeek-V4、GLM-5.x、Kimi-K2.x、MiniMax-M3、
-   Hy3、Hunyuan…）随账号可用列表实时同步，点一下即可切换。
+   Hy3、Hunyuan…）随账号可用列表实时同步，点一下即可切换；支持推理的模型还会
+   显示**推理等级**选择器（低 / 中 / 高）。
 2. **Web 登录状态小组件**：在 Web GUI 右下角常驻一个状态胶囊，**实时显示登录/
    代理状态**，未登录时一键在新标签页打开 WorkBuddy 登录页，登录完成后自动变绿。
    无需再回到终端手动跑登录脚本。
@@ -269,13 +273,15 @@ DSH 的 `dsh.client` 机制只要求 `package.json` 里：
     # defaultContextWindow: 200000      # 未在目录中标明容量的模型使用
     # 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）
     # 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。
+传递。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）。
 
 ---
 
@@ -297,7 +303,9 @@ DSH 的 `dsh.client` 机制只要求 `package.json` 里：
 ## 限制
 
 - 当前为纯文本适配器：图片输入会以 `UNSUPPORTED_CONTENT` 拒绝（后续可加）。
-- 不公布 reasoning effort 选择器；模型按平台默认强度运行。
+- 推理等级（reasoning effort）：支持推理的模型（如 DeepSeek-V4、GLM、Kimi、MiniMax、
+  Hy3 等）会显示推理等级下拉（低 / 中 / 高），默认值取平台默认强度。`reasoning_effort`
+  会透传给代理；若某模型平台侧只接受平台默认、忽略该参数，则退化为平台默认强度，不影响出字。
 - 代理未运行时，模型请求会以 `TRANSPORT` 错误快速失败（连接被拒绝）；但状态
   小组件本身不依赖代理——代理挂了它仍能显示「代理未运行」并允许触发登录。
 - 登录态有效期由 WorkBuddy 云端决定；过期后胶囊变红，重新点「登录」即可，

package/lib/index.js

@@ -73,28 +73,77 @@ const DISCOVERY_TIMEOUT_MS = 2_000;
  * proxy is reachable; entries not announced by the proxy are still listed.
  */
 const DEFAULT_MODELS = [
-  { id: "deepseek-v4-pro", name: "Deepseek-V4-Pro", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"] },
-  { id: "deepseek-v4-flash", name: "Deepseek-V4-Flash", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"] },
-  { id: "glm-5.2", name: "GLM-5.2", contextWindow: 1_000_000, maxTokens: 48_000, inputModalities: ["text", "image"] },
-  { id: "glm-5.1", name: "GLM-5.1", contextWindow: 200_000, maxTokens: 48_000 },
-  { id: "glm-5.0", name: "GLM-5.0", contextWindow: 200_000, maxTokens: 48_000 },
-  { id: "glm-4.7", name: "GLM-4.7", contextWindow: 200_000, maxTokens: 48_000 },
-  { id: "glm-4.6", name: "GLM-4.6", contextWindow: 168_000, maxTokens: 32_000 },
-  { id: "glm-5v-turbo", name: "GLM-5v-Turbo", contextWindow: 200_000, maxTokens: 64_000, inputModalities: ["text", "image"] },
-  { id: "kimi-k3-1", name: "Kimi-K3", contextWindow: 1_000_000, maxTokens: 32_000, inputModalities: ["text", "image"] },
-  { id: "kimi-k2.7", name: "Kimi-K2.7-Code", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"] },
-  { id: "kimi-k2.6", name: "Kimi-K2.6", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"] },
-  { id: "kimi-k2.5", name: "Kimi-K2.5", contextWindow: 164_000, maxTokens: 32_000, inputModalities: ["text", "image"] },
-  { id: "kimi-k2-thinking", name: "Kimi-K2-Thinking", contextWindow: 164_000, maxTokens: 32_000 },
-  { id: "minimax-m3", name: "MiniMax-M3", contextWindow: 512_000, maxTokens: 128_000, inputModalities: ["text", "image"] },
-  { id: "minimax-m2.5", name: "MiniMax-M2.5", contextWindow: 200_000, maxTokens: 48_000 },
-  { id: "hy3", name: "Hy3", contextWindow: 192_000, maxTokens: 64_000, inputModalities: ["text", "image"] },
-  { id: "hunyuan-2.0-thinking", name: "Hunyuan-2.0-Thinking", contextWindow: 128_000, maxTokens: 24_000 },
+  { 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: "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-5.0", name: "GLM-5.0", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "high" },
+  { id: "glm-4.7", name: "GLM-4.7", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "high" },
+  { id: "glm-4.6", name: "GLM-4.6", contextWindow: 168_000, maxTokens: 32_000, reasoningEffort: "high" },
+  { 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: "kimi-k2-thinking", name: "Kimi-K2-Thinking", contextWindow: 164_000, maxTokens: 32_000, reasoningEffort: "medium" },
+  { id: "minimax-m3", name: "MiniMax-M3", contextWindow: 512_000, maxTokens: 128_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
+  { id: "minimax-m2.5", name: "MiniMax-M2.5", contextWindow: 200_000, maxTokens: 48_000, 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: "hunyuan-chat", name: "Hunyuan-Turbos", contextWindow: 200_000, maxTokens: 8_192 },
-  { id: "auto", name: "Auto", contextWindow: 168_000, maxTokens: 32_000, inputModalities: ["text", "image"] },
+  { 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 },
 ];
 
+/** Selectable reasoning efforts exposed to the harness UI, in display order. */
+const REASONING_EFFORTS = [
+  { id: "low", name: "低" },
+  { id: "medium", name: "中" },
+  { id: "high", name: "高" },
+];
+
+/**
+ * Adapter-owned reasoning-effort identifiers accepted by the wire. The harness
+ * `ReasoningEffortId` is opaque; we keep the catalog default and the request
+ * param in this vocabulary and map to/from the proxy as needed.
+ */
+const REASONING_EFFORT_IDS = new Set(["low", "medium", "high"]);
+
+/**
+ * Build the `reasoning` capability metadata for a model, per the dsh-llm
+ * `LlmModelReasoningInfo` contract. `effort` is the model's platform default
+ * (`high`/`medium`/`low`); absent means the model exposes no selectable
+ * reasoning (the harness then hides the selector). When the platform does not
+ * support per-request switching, callers may still omit an explicit effort and
+ * the proxy uses its own default — publishing the default keeps the UI honest
+ * about the level actually applied.
+ * @returns the `reasoning` block, or `undefined` when the model has none.
+ */
+function modelReasoningInfo(effort) {
+  if (effort === undefined || !REASONING_EFFORT_IDS.has(effort)) return undefined;
+  return {
+    efforts: REASONING_EFFORTS,
+    defaultEffort: effort,
+  };
+}
+
+/**
+ * Extract a model's default reasoning effort from a live `/v1/models` entry.
+ * The proxy reports `supported_reasoning_levels: [{ effort: 'High' }]` and
+ * `default_reasoning_level: null`, so we fall back to the catalog's
+ * `reasoningEffort` when the live answer is uninformative (as it currently is).
+ * @returns a lowercased effort id (`low`/`medium`/`high`) or `undefined`.
+ */
+function liveReasoningEffort(entry, catalogEffort) {
+  if (catalogEffort !== undefined) return catalogEffort;
+  const levels = Array.isArray(entry?.supported_reasoning_levels) ? entry.supported_reasoning_levels : [];
+  const wire = levels[0]?.effort;
+  if (typeof wire !== "string") return undefined;
+  const lower = wire.toLowerCase();
+  return lower === "low" || lower === "medium" || lower === "high" ? lower : undefined;
+}
+
+
 // #region serialize
 
 /** Join the text blocks of a message (used for user/tool-result content). */
@@ -164,6 +213,9 @@ function serializeRequest(options) {
     ...(options.temperature !== undefined ? { temperature: options.temperature } : {}),
     ...(options.maxTokens === undefined ? {} : { max_tokens: options.maxTokens }),
     ...(options.stop !== undefined ? { stop: options.stop } : {}),
+    // reasoning_effort is opaque here (low|medium|high); the proxy passes it
+    // through verbatim to the upstream /v2/chat/completions body.
+    ...(options.reasoningEffort === undefined ? {} : { reasoning_effort: options.reasoningEffort }),
   };
 }
 
@@ -363,12 +415,14 @@ async function* translate(payloads) {
 
 /** Display metadata for one catalog entry. */
 function modelInfo(provider, model) {
+  const reasoning = modelReasoningInfo(model.reasoningEffort);
   return {
     provider,
     id: model.id,
     name: model.name ?? model.id,
     ...(model.description === undefined ? {} : { description: model.description }),
     inputModalities: model.inputModalities ?? ["text"],
+    ...(reasoning === undefined ? {} : { reasoning }),
   };
 }
 
@@ -457,12 +511,15 @@ export class WorkBuddyAdapter extends LlmAdapter {
         if (seen.has(entry.id)) continue;
         seen.add(entry.id);
         const catalog = byId.get(entry.id);
+        const effort = liveReasoningEffort(entry, catalog?.reasoningEffort);
+        const reasoning = modelReasoningInfo(effort);
         merged.push({
           provider,
           id: entry.id,
           name: entry.name ?? catalog?.name ?? entry.id,
           ...(catalog?.description !== undefined ? { description: catalog.description } : {}),
           inputModalities: catalog?.inputModalities ?? ["text"],
+          ...(reasoning === undefined ? {} : { reasoning }),
         });
       }
       // Catalog entries the proxy did not announce (e.g. unauthenticated or
@@ -482,10 +539,11 @@ export class WorkBuddyAdapter extends LlmAdapter {
   resolveModel(provider, model, _signal) {
     const connection = this.config.options();
     const configured = connection.models.find((entry) => entry.id === model);
+    const base = configured === undefined
+      ? { provider, id: model, name: model, inputModalities: ["text"] }
+      : modelInfo(provider, configured);
     return Promise.resolve({
-      ...(configured === undefined
-        ? { provider, id: model, name: model, inputModalities: ["text"] }
-        : modelInfo(provider, configured)),
+      ...base,
       context: { contextWindow: configured?.contextWindow ?? connection.defaultContextWindow },
       defaultMaxTokens: configured?.maxTokens ?? connection.maxTokens,
     });
@@ -586,6 +644,7 @@ const catalogModel = z.object({
   contextWindow: z.number().step(1).min(1),
   maxTokens: z.number().step(1).min(1),
   inputModalities: z.array(z.union(MODEL_MODALITIES)).min(1).default(["text"]),
+  reasoningEffort: z.string(),
 });
 
 export const Config = z.object({
@@ -625,6 +684,10 @@ function resolveModels(models) {
     }
     if (seen.has(model.id)) throw new Error(`dsh-llm-workbuddy: duplicate catalog model "${model.id}"`);
     seen.add(model.id);
+    const reasoningEffort = model.reasoningEffort;
+    if (reasoningEffort !== undefined && !REASONING_EFFORT_IDS.has(reasoningEffort)) {
+      throw new Error(`dsh-llm-workbuddy: catalog model "${model.id}" reasoningEffort must be one of low|medium|high`);
+    }
     return {
       id: model.id,
       ...(model.name === undefined ? {} : { name: model.name }),
@@ -632,6 +695,7 @@ function resolveModels(models) {
       ...(model.contextWindow === undefined ? {} : { contextWindow: model.contextWindow }),
       ...(model.maxTokens === undefined ? {} : { maxTokens: model.maxTokens }),
       inputModalities: [...inputModalities],
+      ...(reasoningEffort === undefined ? {} : { reasoningEffort }),
     };
   });
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-llm-workbuddy",
-  "version": "0.1.6",
+  "version": "0.1.7",
   "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",
@@ -48,9 +48,9 @@
   ],
   "peerDependencies": {
     "@deepseek-ai/cordis": "^4.0.1",
-    "@deepseek-ai/dsh-llm": "^0.1.0-rc.7",
-    "@deepseek-ai/dsh-settings": "^0.1.0-rc.7",
-    "@deepseek-ai/dsh-timeout": "^0.1.0-rc.7",
+    "@deepseek-ai/dsh-llm": "^0.1.1-rc.2",
+    "@deepseek-ai/dsh-settings": "^0.1.1-rc.2",
+    "@deepseek-ai/dsh-timeout": "^0.1.1-rc.2",
     "@deepseek-ai/schemastery": "^3.18.1"
   }
 }