@goodandready/dsh-clinebot 0.3.17 → 0.3.18

package/CHANGELOG.md

@@ -5,6 +5,18 @@ All notable changes to `@goodandready/dsh-clinebot` will be documented in this f
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.3.18] - 2026-09-22
+
+### Added
+- **Reasoning Effort (思考强度) Selection (#59)**: Enabled reasoning effort selection (`Off` / `Low` / `Medium` / `High` / `Max`) across DSH composer model picker and `dsh-effort-slider` for all supported reasoning models (Qwen 3.7 Max/Plus, MiMo 2.5/Pro, GLM 5.2, Kimi K3, DeepSeek V4 Pro/Flash, MiniMax M3).
+- **Protocol Compliance**: Injected `off: null` into wire dictionary and declared `compat: { supportsReasoningEffort: true }` at provider and model levels in accordance with `@earendil-works/pi-ai` protocol specifications.
+- **Dynamic Plan Model Reasoning Heuristics**: Added automatic detection of reasoning capabilities for dynamically discovered models matching `/qwen3\.7|mimo|glm-5|kimi-k3|deepseek-v4|minimax-m3/i`.
+- **Reasoning Tooltip Localization**: Added localized explanatory tooltips (`models.reasoning_tooltip`) across English, Russian, and Chinese in the Settings UI.
+
+### Fixed
+- **Settings UI Reasoning Badge**: Fixed reasoning badge check in `models-section.js` to correctly detect object dictionary configurations instead of relying solely on array length.
+- **Context Window Capacity Clarification**: Updated catalog documentation and descriptions to clarify that the uniform 200k tokens context parameter reflects the official ClinePass upstream completion proxy gateway limit.
+
 ## [0.3.17] - 2026-09-21
 
 ### Fixed

package/README.md

@@ -108,13 +108,13 @@ graph LR
 ## ✨ Features & Module Breakdown
 
 * **`lib/models.js`**:
-  Manages the curated catalog (11 built-in models) and dynamically parses subscription plan models (`parsePlanIncludedModels`, `getAllModels`, `getDynamicModels`).
+  Manages the curated catalog (11 built-in models) with full reasoning effort mappings (`off: null`, `low`, `medium`, `high`, `max`), upstream 200k context limit declarations, and dynamic subscription plan model parsing (`parsePlanIncludedModels`, `getAllModels`, `getDynamicModels`).
 * **`lib/cline-client.js`**:
   * `fetchUsageLimits`: queries `GET /users/me/plan/usage-limits`, `GET /users/me/plan`, and `GET /users/me` with in-memory caching.
   * `sessionStats` / `recordSessionRequest`: in-memory telemetry recording requests count, tokens, latency, and timestamps.
   * `saveCredentialKey`: writes credentials directly into `~/.dsh/.credentials.yaml`.
   * `smokeChat`: tests latency via non-streaming ping and updates session metrics.
-  * `buildPiAiProvider`: builds the DSH `llm-pi-ai` structure (`api: 'openai-completions'`).
+  * `buildPiAiProvider`: builds the DSH `llm-pi-ai` structure (`api: 'openai-completions'`) with full reasoning effort protocol compliance (`compat.supportsReasoningEffort: true`).
 * **`lib/index.js`**:
   Cordis service module managing routes (including `POST /dsh-clinebot/models/sync`), quota warnings threshold evaluation, credentials, and registering the `/cline` slash command.
 * **`lib/client.js`**:

package/README.ru.md

@@ -107,7 +107,7 @@ graph LR
 
 ## ✨ Структура модулей и возможности
 
-* **`lib/models.js`**: каталог встроенных моделей ClinePass (11 моделей) и парсер моделей подписки (`parsePlanIncludedModels`, `getAllModels`, `getDynamicModels`).
+* **`lib/models.js`**: каталог встроенных моделей ClinePass (11 моделей с полной поддержкой уровней размышлений `Off` / `Low` / `Medium` / `High` / `Max` и фиксацией лимита шлюза в 200k токенов) и парсер моделей подписки (`parsePlanIncludedModels`, `getAllModels`, `getDynamicModels`).
 * **`lib/cline-client.js`**:
   * `fetchUsageLimits`: параллельный опрос `GET /users/me/plan/usage-limits`, `GET /users/me/plan` и `GET /users/me` с кэшированием в памяти.
   * `sessionStats` / `recordSessionRequest`: счетчики сессии (запросы, токены, задержка, время).

package/README.zh.md

@@ -107,7 +107,7 @@ graph LR
 
 ## ✨ 核心模块与功能
 
-* **`lib/models.js`**：管理 11 款官方精选内置模型以及套餐模型动态解析器（`parsePlanIncludedModels`, `getAllModels`, `getDynamicModels`）。
+* **`lib/models.js`**：管理 11 款官方精选内置模型（提供完整的 `Off` / `Low` / `Medium` / `High` / `Max` 思考强度映射、上游 200k 上下文容量标注）以及套餐模型动态解析器（`parsePlanIncludedModels`, `getAllModels`, `getDynamicModels`）。
 * **`lib/cline-client.js`**：
   * `fetchUsageLimits`：高效并发轮询 `GET /users/me/plan/usage-limits`、`GET /users/me/plan` 与 `GET /users/me` 并进行内存缓存。
   * `sessionStats` / `recordSessionRequest`：内存级会话度量记录器（请求次数、Token 估算、延迟、时间戳）。

package/lib/client.js

@@ -65,7 +65,8 @@ const en = {
       'quota.remaining': 'Remaining: {pct}%',
       'quota.reset_at': 'Reset: {time}',
       'models.title': '🎯 Model Picker in DSH ({enabled} of {total} active)',
-      'models.desc': 'Catalog synchronizes directly from ClinePass subscription. New plan models are enabled automatically. Uncheck models to hide them from DSH chat picker.',
+      'models.desc': 'Catalog synchronizes directly from ClinePass subscription. New plan models are enabled automatically. Uncheck models to hide them from DSH chat picker. Thinking effort is selectable in the composer model menu.',
+      'models.reasoning_tooltip': 'Supports thinking effort selection in composer model menu (Off / Low / Medium / High / Max)',
       'models.sync': '🔄 Check Plan Models',
       'models.syncing': 'Checking…',
       'models.synced_msg': 'Synchronized subscription models: {total} ({discovered} discovered from ClinePass)',
@@ -154,7 +155,8 @@ const en = {
       'quota.remaining': '剩余可用：{pct}%',
       'quota.reset_at': '重置时间：{time}',
       'models.title': '🎯 DSH 模型选择列表（已启用 {enabled}/{total}）',
-      'models.desc': '模型目录与 ClinePass 官方订阅实时同步。新订阅模型默认自动启用。取消勾选可在 DSH 对话模型列表中隐藏不需要的模型。',
+      'models.desc': '模型目录与 ClinePass 官方订阅实时同步。新订阅模型默认自动启用。取消勾选可在 DSH 对话模型列表中隐藏不需要的模型。支持在对话输入框上方切换推理思考强度（Effort）。',
+      'models.reasoning_tooltip': '支持在对话框上方模型菜单中选择思考强度（Off / Low / Medium / High / Max）',
       'models.sync': '🔄 同步订阅模型',
       'models.syncing': '正在同步…',
       'models.synced_msg': '已成功同步订阅模型：共 {total} 个（发现 {discovered} 个新模型）',
@@ -727,8 +729,8 @@ function ModelsSection({ modelsList, disabledSet, enabledCount, keyPresent, busy
               m.input?.includes('image') || m.input?.includes('vision')
                 ? React.createElement('span', { className: 'cb-badge', style: { marginLeft: '4px' } }, 'Vision')
                 : null,
-              m.reasoningEfforts?.length || m.reasoning
-                ? React.createElement('span', { className: 'cb-badge', style: { marginLeft: '4px' } }, '🧠 Reasoning')
+              (Array.isArray(m.reasoningEfforts) ? m.reasoningEfforts.length > 0 : (m.reasoningEfforts && typeof m.reasoningEfforts === 'object' ? Object.keys(m.reasoningEfforts).length > 0 : Boolean(m.reasoning)))
+                ? React.createElement('span', { className: 'cb-badge cb-badge-ok', style: { marginLeft: '4px' }, title: t('models.reasoning_tooltip') }, '🧠 Reasoning')
                 : null
             )
           )

package/lib/cline-client.js

@@ -526,24 +526,24 @@ export function buildPiAiProvider({
       provider: PROVIDER_ID,
     }
     const validEfforts = ['off', 'minimal', 'low', 'medium', 'high', 'xhigh', 'max']
+    let efforts = null
     if (Array.isArray(item.reasoningEfforts) && item.reasoningEfforts.length) {
-      const efforts = {}
+      efforts = {}
       for (const effort of item.reasoningEfforts) {
         const key = String(effort).trim().toLowerCase()
-        if (validEfforts.includes(key)) {
-          efforts[key] = key
-        }
+        if (validEfforts.includes(key)) efforts[key] = key === 'off' ? null : key
       }
-      res.reasoningEfforts = Object.keys(efforts).length ? efforts : false
     } else if (item.reasoningEfforts && typeof item.reasoningEfforts === 'object' && !Array.isArray(item.reasoningEfforts)) {
-      const efforts = {}
+      efforts = {}
       for (const [k, v] of Object.entries(item.reasoningEfforts)) {
         const key = String(k).trim().toLowerCase()
-        if (validEfforts.includes(key)) {
-          efforts[key] = typeof v === 'string' && v.trim() ? v.trim() : key
-        }
+        if (validEfforts.includes(key)) efforts[key] = key === 'off' ? (v === undefined ? null : v) : (typeof v === 'string' && v.trim() ? v.trim() : key)
       }
-      res.reasoningEfforts = Object.keys(efforts).length ? efforts : false
+    }
+    if (efforts && Object.keys(efforts).length) {
+      if (!('off' in efforts)) efforts.off = null
+      res.reasoningEfforts = efforts
+      res.compat = { supportsReasoningEffort: true }
     } else {
       res.reasoningEfforts = false
     }
@@ -555,6 +555,7 @@ export function buildPiAiProvider({
     api: 'openai-completions',
     baseURL: normalizeBaseUrl(baseUrl),
     apiKeyEnv: apiKeyEnv || DEFAULT_API_KEY_ENV,
+    compat: { supportsReasoningEffort: true },
     models: modelList,
   }
 }

package/lib/config.js

@@ -38,6 +38,7 @@ export const Config = z.object({
     input: z.array(z.string()).default(['text']),
     category: z.string().default('general'),
     isCustom: z.boolean().default(false),
+    reasoningEfforts: z.any().default(undefined),
   })).default([])
     .description('Models automatically discovered from the official ClinePass subscription plan.'),
   timeoutMs: z.number().default(DEFAULT_TIMEOUT_MS)

package/lib/models.js

@@ -2,8 +2,11 @@
  * Curated and extensible models catalog for ClineBot / ClinePass.
  *
  * Notice: GET /v1/models returns 404 on api.cline.bot, so DSH cannot rely on
- * dynamic OpenAI model discovery. This catalogue provides the official ClinePass
- * models plus support for user-defined custom models added via the Settings page.
+ * dynamic OpenAI model discovery. Furthermore, ClinePass upstream proxy enforces
+ * a uniform 200,000 token context window cap across subscription completion routes,
+ * and /users/me/plan returns only feature strings without per-model context specs.
+ * This catalogue provides the official ClinePass models (200k gateway capacity) plus
+ * support for user-defined custom models with arbitrary context length.
  */
 
 export const PROVIDER_ID = 'clinebot'
@@ -45,6 +48,7 @@ export const CLINE_MODELS = Object.freeze([
     category: 'general',
     recommended: false,
     isCustom: false,
+    reasoningEfforts: ['low', 'medium', 'high'],
   },
   {
     id: 'cline-pass/kimi-k3',
@@ -56,7 +60,7 @@ export const CLINE_MODELS = Object.freeze([
     category: 'reasoning',
     recommended: true,
     isCustom: false,
-    reasoningEfforts: ['low', 'high', 'max'],
+    reasoningEfforts: ['low', 'medium', 'high', 'max'],
   },
   {
     id: 'cline-pass/kimi-k2.7-code',
@@ -83,24 +87,26 @@ export const CLINE_MODELS = Object.freeze([
   {
     id: 'cline-pass/qwen3.7-max',
     name: 'Qwen 3.7 Max',
-    description: 'Large-scale multimodal foundation model from Alibaba Cloud.',
+    description: 'Large-scale multimodal foundation model from Alibaba Cloud with reasoning support.',
     contextLength: 200000,
     maxTokens: 8192,
     input: ['text', 'image'],
     category: 'multimodal',
     recommended: true,
     isCustom: false,
+    reasoningEfforts: ['low', 'medium', 'high', 'max'],
   },
   {
     id: 'cline-pass/qwen3.7-plus',
     name: 'Qwen 3.7 Plus',
-    description: 'Fast, capable multimodal model with strong multilingual skills.',
+    description: 'Fast, capable multimodal model with strong multilingual skills and reasoning.',
     contextLength: 200000,
     maxTokens: 8192,
     input: ['text', 'image'],
     category: 'multimodal',
     recommended: false,
     isCustom: false,
+    reasoningEfforts: ['low', 'medium', 'high', 'max'],
   },
   {
     id: 'cline-pass/minimax-m3',
@@ -117,24 +123,26 @@ export const CLINE_MODELS = Object.freeze([
   {
     id: 'cline-pass/mimo-v2.5',
     name: 'MiMo V2.5',
-    description: 'Xiaomi MiMo efficient instruction model.',
+    description: 'Xiaomi MiMo efficient instruction model with reasoning and multimodal support.',
     contextLength: 200000,
     maxTokens: 8192,
-    input: ['text'],
+    input: ['text', 'image'],
     category: 'general',
     recommended: false,
     isCustom: false,
+    reasoningEfforts: ['low', 'medium', 'high', 'max'],
   },
   {
     id: 'cline-pass/mimo-v2.5-pro',
     name: 'MiMo V2.5 Pro',
-    description: 'Xiaomi MiMo advanced agentic reasoning model.',
+    description: 'Xiaomi MiMo advanced agentic reasoning model with multimodal support.',
     contextLength: 200000,
     maxTokens: 8192,
-    input: ['text'],
+    input: ['text', 'image'],
     category: 'coding',
     recommended: false,
     isCustom: false,
+    reasoningEfforts: ['low', 'medium', 'high', 'max'],
   },
 ])
 
@@ -186,16 +194,19 @@ export function parsePlanIncludedModels(includedInput) {
     } else {
       // Dynamic fallback for newly introduced models mentioned in plan
       const idPart = name.toLowerCase().replace(/[\s_]+/g, '-')
+      const isVision = /(vision|vl|multimodal|omni|image)/i.test(name)
+      const hasReasoning = /(reason|think|r1|pro|flash|max|plus|k3|m3|glm|qwen|mimo)/i.test(name)
       matched.push({
         id: `cline-pass/${idPart}`,
         name,
         description: `Official ClinePass subscription model: ${name}`,
         contextLength: 200000,
         maxTokens: 8192,
-        input: ['text'],
-        category: 'general',
+        input: isVision ? ['text', 'image'] : ['text'],
+        category: hasReasoning ? 'reasoning' : (isVision ? 'multimodal' : 'general'),
         recommended: false,
         isCustom: false,
+        ...(hasReasoning ? { reasoningEfforts: ['low', 'medium', 'high', 'max'] } : {}),
       })
     }
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-clinebot",
-  "version": "0.3.17",
+  "version": "0.3.18",
   "description": "DeepSeek Harness companion for ClineBot / ClinePass: dynamic subscription models sync, quota exhaustion warnings, session metrics, dedicated settings page, live usage limits, and /cline slash-command.",
   "license": "MIT",
   "type": "module",