@goodandready/dsh-model-sync 0.5.14 → 0.5.16

package/README.md

@@ -80,6 +80,8 @@ graph LR
 
 * 💰 **Built-in Reference Pricing Database**: Canonical token pricing database automatically enriches models for OpenAI, Anthropic, Groq, Mistral, xAI, and DeepSeek, enabling cost-limit filtering and `@cheap` auto-resolution even when upstream APIs omit pricing metadata.
 * 🎯 **Dynamic Smart Virtual Aliases & Auto-Fallback**: Route requests to virtual aliases like `@fast`, `@cheap`, and `@best-code`, with automatic fallback chains when circuit breakers open or health probes fail.
+* 📏 **Real Context Limits & Boundary Safety (`contextWindow` / `maxOutputTokens`)**: precise separation of input and generation limits, prevention of 400 `max_tokens` errors via auto-clamping, and clear `In: 128k | Out: 8k` badges in Web UI.
+* 🧠 **Unified Reasoning & Thinking Configuration**: automatic format detection (`effort`, `budget_tokens`, `boolean`) and parameter translation with suppression of incompatible options (such as `temperature` on o1/o3).
 * 🔔 **External Diff Webhooks & Event Dispatching**: Real-time webhook notifications with HMAC-SHA256 signature verification and Cordis `model-sync/diff` events when models are added, deprecated, or prices change.
 * ⚡ **High-Performance Policy Filtering**: Fast-path evaluation in `filterModels` bypasses regular expressions and array filtering when no filtering policy is configured, guaranteeing zero-overhead discovery runs.
 * 🔄 **One-Click In-App Updates**: Built-in update checker in Settings with automatic npm cache bypass on demand (`?fresh=1` / `Cache-Control: no-cache`), ensuring instantaneous discovery of new package releases.

package/README.ru.md

@@ -80,6 +80,8 @@ graph LR
 
 * 💰 **Встроенная база эталонных цен токенов**: Автоматическое обогащение моделей тарифами для OpenAI, Anthropic, Groq, Mistral, xAI и DeepSeek, обеспечивающее работу лимитов цен и псевдонима `@cheap` даже при отсутствии метаданных в API провайдера.
 * 🎯 **Динамические «умные» псевдонимы и авто-фоллбэк**: Поддержка виртуальных алиасов (`@fast`, `@cheap`, `@best-code`) и цепочек резервирования с автопереключением при размыкании Circuit Breaker.
+* 📏 **Реальные лимиты контекста и авто-клампинг (`contextWindow` / `maxOutputTokens`)**: точное разделение входного и выходного лимита токенов, защита от ошибок 400 `max_tokens` и индикация `Вход: 128k | Выход: 8k` в Web UI.
+* 🧠 **Унификация параметров Reasoning & Thinking**: автоматическое определение формата (`effort`, `budget_tokens`, `boolean`) и параметров бюджета рассуждений с подавлением несовместимых параметров (например, `temperature` для o1/o3).
 * 🔔 **Внешние Diff Webhooks и диспетчеризация событий**: Уведомления в реальном времени с проверкой подписи HMAC-SHA256 и событиями Cordis `model-sync/diff` при добавлении моделей, их устаревании или изменении цен.
 * ⚡ **Высокопроизводительная фильтрация моделей**: Быстрый путь (`fast-path`) в `filterModels` мгновенно возвращает каталог при отсутствии активных правил политики, исключая лишнюю компиляцию regex и прогоны массивов.
 * 🔄 **Обновление плагина в один клик**: Встроенный механизм проверки обновлений в Настройках со сбросом кэша npm при ручном запросе (`?fresh=1` / `Cache-Control: no-cache`), гарантирующий получение актуальной версии.

package/README.zh.md

@@ -80,6 +80,8 @@ graph LR
 
 * 💰 **内置基准价格数据库**：自动为 OpenAI、Anthropic、Groq、Mistral、xAI 和 DeepSeek 模型补充 Token 价格数据，即使服务商接口未返回价格，也可无缝支持成本上限过滤与 `@cheap` 别名解析。
 * 🎯 **动态智能虚拟别名与故障转移**：支持 `@fast`、`@cheap`、`@best-code` 等虚拟别名及故障转移链，在熔断器打开或健康检查失败时自动切换至备用模型。
+* 📏 **真实上下文限制与边界安全 (`contextWindow` / `maxOutputTokens`)**：精确区分输入与输出令牌限制，通过自动限制防止 400 `max_tokens` 错误，并在 Web UI 中显示 `输入: 128k | 输出: 8k` 徽章。
+* 🧠 **统一的深度思考与推理配置 (Reasoning & Thinking)**：自动检测参数格式 (`effort`, `budget_tokens`, `boolean`) 并转换推理预算，自动屏蔽不兼容参数（如 o1/o3 的 `temperature`）。
 * 🔔 **外部差异通知 Webhook 与事件分发**：支持 HMAC-SHA256 签名验证的实时 Webhook 通知与 Cordis `model-sync/diff` 事件，可在新模型上线、模型废弃或价格变更时即时推送。
 * ⚡ **高性能模型策略过滤**：在未配置过滤策略时，`filterModels` 采用快速路径直接返回模型目录，避免不必要的正则表达式编译和数组遍历。
 * 🔄 **一键检查与更新**：内置设置面板更新检查，在手动刷新时通过 `?fresh=1` 或 `Cache-Control: no-cache` 绕过 npm 缓存，即时获取最新版本。

package/lib/adapter-registry.js

@@ -32,6 +32,7 @@ const PROVIDER_ADAPTERS = Object.freeze({
       id: 'id',
       name: 'name',
       contextWindow: 'context_length',
+      maxOutputTokens: 'max_tokens',
     },
     defaultModels: [
       'deepseek/deepseek-v4.1-flash',
@@ -83,7 +84,7 @@ function mappedRows(provider, rows, descriptor) {
   return rows.map((row) => {
     const mapped = { ...(row && typeof row === 'object' ? row : {}) }
     for (const [target, source] of Object.entries(fields)) {
-      if (!['id', 'name', 'description', 'contextWindow', 'maxTokens', 'pricing', 'tags'].includes(target)) continue
+      if (!['id', 'name', 'description', 'contextWindow', 'maxTokens', 'maxOutputTokens', 'pricing', 'tags'].includes(target)) continue
       const value = pathValue(row, source)
       if (value !== undefined) mapped[target] = value
     }
@@ -128,6 +129,7 @@ function parseRows(provider, payload, descriptor) {
       id: String(row.name ?? '').replace(/^models\//, ''),
       name: row.displayName ?? row.display_name ?? row.name,
       contextWindow: row.inputTokenLimit,
+      maxOutputTokens: row.outputTokenLimit,
       maxTokens: row.outputTokenLimit,
     })))
   }
@@ -194,7 +196,7 @@ export function normalizeAdapterConfig(config, profile = {}) {
   const capabilityMap = source.capabilityMap && typeof source.capabilityMap === 'object' && !Array.isArray(source.capabilityMap) ? structuredClone(source.capabilityMap) : {}
   for (const key of Object.keys(capabilityMap)) if (!CAPABILITIES.has(key)) delete capabilityMap[key]
   const fields = source.fields && typeof source.fields === 'object' && !Array.isArray(source.fields) ? structuredClone(source.fields) : {}
-  for (const key of Object.keys(fields)) if (!['id', 'name', 'description', 'contextWindow', 'maxTokens', 'pricing', 'tags'].includes(key) || typeof fields[key] !== 'string' || fields[key].length > 128) delete fields[key]
+  for (const key of Object.keys(fields)) if (!['id', 'name', 'description', 'contextWindow', 'maxTokens', 'maxOutputTokens', 'pricing', 'tags'].includes(key) || typeof fields[key] !== 'string' || fields[key].length > 128) delete fields[key]
   for (const key of Object.keys(capabilityMap)) {
     const values = Array.isArray(capabilityMap[key]) ? capabilityMap[key] : [capabilityMap[key]]
     capabilityMap[key] = values.filter((value) => typeof value === 'string' && value.length <= 128).slice(0, 8)

package/lib/client.js

@@ -295,6 +295,9 @@ window.__ModuleLoader__.load({
       saveFailed: 'Save rejected by settings service',
       scopeReadOnly: 'Settings are read-only in this environment',
       firstSeen: 'First seen',
+      contextIn: 'In',
+      contextOut: 'Out',
+      reasoningFormat: 'Reasoning',
     }
 
     const zh = {
@@ -464,6 +467,23 @@ window.__ModuleLoader__.load({
       saveFailed: '设置服务拒绝保存',
       scopeReadOnly: '当前环境为只读设置',
       firstSeen: '首次发现',
+      contextIn: '输入',
+      contextOut: '输出',
+      reasoningFormat: '推理',
+    }
+
+    
+    function formatTokens(num) {
+      if (typeof num !== "number" || !Number.isFinite(num) || num <= 0) return "?";
+      if (num >= 1000000) {
+        const m = num / 1000000;
+        return (m % 1 === 0 ? m : m.toFixed(1)) + "M";
+      }
+      if (num >= 1000) {
+        const k = num / 1000;
+        return (k % 1 === 0 ? k : k.toFixed(0)) + "k";
+      }
+      return String(num);
     }
 
     const API = '/dsh-model-sync'
@@ -1703,6 +1723,30 @@ window.__ModuleLoader__.load({
                     ` ${model.name ?? model.id} (${model.id})`,
                     isNewModel(model, row) ? React.createElement('span', { style: { marginLeft: 6, fontSize: 11, padding: '1px 5px', borderRadius: 4, background: 'var(--dsw-alias-state-business-primary)', color: 'white' } }, 'NEW') : null,
                     deprecationDays !== null ? React.createElement('span', { style: { marginLeft: 6, fontSize: 10, padding: '1px 5px', borderRadius: 4, background: 'var(--dsw-alias-state-warn-primary)', color: 'white' } }, deprecationDays > 0 ? t.deprecatesIn.replace('%d', deprecationDays) : t.deprecated) : null,
+                    (model.contextWindow || model.maxOutputTokens || model.maxTokens) ? React.createElement('span', {
+                      style: {
+                        marginLeft: 6,
+                        fontSize: 11,
+                        padding: '1px 5px',
+                        borderRadius: 4,
+                        background: 'var(--dsw-alias-bg-layer-2)',
+                        border: '1px solid var(--dsw-alias-border-l1)',
+                        color: 'var(--dsw-alias-label-secondary)',
+                      },
+                      title: `${t.contextIn || 'In'}: ${model.contextWindow || '?'} | ${t.contextOut || 'Out'}: ${model.maxOutputTokens || model.maxTokens || '?'}`
+                    }, `${t.contextIn || 'In'}: ${formatTokens(model.contextWindow)} | ${t.contextOut || 'Out'}: ${formatTokens(model.maxOutputTokens || model.maxTokens)}`) : null,
+                    (model.reasoningConfig?.supportsReasoning || model.capabilities?.reasoning) ? React.createElement('span', {
+                      style: {
+                        marginLeft: 6,
+                        fontSize: 11,
+                        padding: '1px 5px',
+                        borderRadius: 4,
+                        background: 'var(--dsw-alias-state-business-secondary, var(--dsw-alias-bg-layer-2))',
+                        color: 'var(--dsw-alias-state-business-primary)',
+                        border: '1px solid var(--dsw-alias-border-l1)',
+                      },
+                      title: model.reasoningConfig?.paramFormat ? `Reasoning format: ${model.reasoningConfig.paramFormat}` : 'Reasoning supported'
+                    }, `🧠 ${model.reasoningConfig?.paramFormat || 'reasoning'}`) : null,
                     (model.contextWindow && maxContext > 0 && model.contextWindow >= maxContext) ? React.createElement('span', { style: { marginLeft: 4, fontSize: 10, padding: '1px 4px', borderRadius: 3, background: 'var(--dsw-alias-state-warn-primary)', color: 'white' } }, '★') : null,
                     (model.pricing && minPrice !== Infinity && model.pricing.inputPerToken === minPrice) ? React.createElement('span', { style: { marginLeft: 4, fontSize: 10, padding: '1px 4px', borderRadius: 3, background: 'var(--dsw-alias-state-success-primary)', color: 'white' } }, '$') : null,
                     model.tags && model.tags.length ? React.createElement('small', { style: { marginLeft: 6, color: 'var(--dsw-alias-label-secondary)' } }, model.tags.join(', ')) : null,

package/lib/diff-notifier.js

@@ -1,3 +1,15 @@
+function getOptionalService(ctx, name) {
+  try {
+    return ctx?.[name] ?? ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+  } catch {
+    try {
+      return ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+    } catch {
+      return undefined
+    }
+  }
+}
+
 import { createHmac } from "node:crypto"
 
 export function extractDiffEvents(results, options = {}) {
@@ -220,9 +232,10 @@ export async function dispatchDiffNotifications(ctx, config, results, { fetchImp
   }
 
   try {
-    if (typeof ctx?.pluginNotify?.send === "function") {
+    const pluginNotify = getOptionalService(ctx, 'pluginNotify')
+    if (typeof pluginNotify?.send === "function") {
       const summary = `Model Sync: +${events.counts.added} added, -${events.counts.deprecated} deprecated, ~${events.counts.priceChanged} price changed`
-      await ctx.pluginNotify.send({
+      await pluginNotify.send({
         source: "dsh-model-sync",
         title: "Model Sync Catalog Changes",
         message: summary,

package/lib/http.js

@@ -1,3 +1,15 @@
+function getOptionalService(ctx, name) {
+  try {
+    return ctx?.[name] ?? ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+  } catch {
+    try {
+      return ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+    } catch {
+      return undefined
+    }
+  }
+}
+
 import { isForbiddenWebhookHost, isLoopbackHost } from './diff-notifier.js'
 import crypto from 'node:crypto'
 import { validateAliasName } from './alias-resolver.js'
@@ -43,7 +55,7 @@ export function isLoopbackAddress(ip) {
 
 export function trusted(req, { requireLoopback = false, ctx = null } = {}) {
   // 1. Core connection rejection check (#237)
-  const conn = ctx?.connection || ctx?.get?.('connection') || req?.connection
+  const conn = getOptionalService(ctx, 'connection') || req?.connection
   if (conn && typeof conn.requestRejection === 'function') {
     const rejection = conn.requestRejection(req)
     if (rejection !== undefined) return false
@@ -274,7 +286,7 @@ export function registerHttpApi(ctx, synchronizer, { getSchedulerStatus = () =>
   const makeHandler = (route) => async (req, res) => {
     try {
       // 1. Core connection check on all routes (#237)
-    const conn = ctx?.connection || ctx?.get?.('connection') || req?.connection
+    const conn = getOptionalService(ctx, 'connection') || req?.connection
     if (conn && typeof conn.requestRejection === 'function') {
       const rejection = conn.requestRejection(req)
       if (rejection !== undefined) {

package/lib/models.js

@@ -1,13 +1,23 @@
 import { lookupReferencePricing } from './pricing-database.js'
+import { lookupReferenceLimits, lookupReferenceReasoning, lookupReferenceCompat } from './specs-database.js'
 const FIELD_ALIASES = Object.freeze({
   contextWindow: [
     'contextWindow', 'context_window', 'contextLength', 'context_length',
     'maxContextTokens', 'max_context_tokens', 'maxContextLength', 'max_context_length',
     'inputTokenLimit', 'input_token_limit',
+    'top_provider.context_length', 'topProvider.contextLength',
   ],
   maxTokens: [
     'maxTokens', 'max_tokens', 'max_output_tokens', 'maxOutputTokens',
     'outputTokenLimit', 'output_token_limit',
+    'max_completion_tokens', 'maxCompletionTokens',
+    'top_provider.max_completion_tokens', 'topProvider.maxCompletionTokens',
+  ],
+  maxOutputTokens: [
+    'maxOutputTokens', 'max_output_tokens', 'maxTokens', 'max_tokens',
+    'outputTokenLimit', 'output_token_limit',
+    'max_completion_tokens', 'maxCompletionTokens',
+    'top_provider.max_completion_tokens', 'topProvider.maxCompletionTokens',
   ],
 })
 
@@ -60,8 +70,14 @@ function numericValue(value, { allowZero = false } = {}) {
 
 function firstNumber(row, fields, options) {
   for (const field of fields) {
-    const value = numericValue(row?.[field], options)
-    if (value !== undefined) return value
+    let value
+    if (field.includes('.')) {
+      value = field.split('.').reduce((acc, k) => (acc && typeof acc === 'object') ? acc[k] : undefined, row)
+    } else {
+      value = row?.[field]
+    }
+    const val = numericValue(value, options)
+    if (val !== undefined) return val
   }
   return undefined
 }
@@ -184,6 +200,13 @@ function normalizeCompat(provider, row, id) {
     source.supportsDeveloperRole = defaultSupportsDeveloperRole
   }
 
+  const refCompat = lookupReferenceCompat(provider, id)
+  if (refCompat) {
+    for (const [k, v] of Object.entries(refCompat)) {
+      if (source[k] === undefined) source[k] = v
+    }
+  }
+
   return Object.keys(source).length > 0 ? source : undefined
 }
 
@@ -298,6 +321,50 @@ function normalizeDate(val) {
   return undefined
 }
 
+function normalizeReasoningConfig(provider, row, id, capabilities) {
+  const source = row?.reasoningConfig && typeof row.reasoningConfig === 'object' ? { ...row.reasoningConfig } : {}
+  const ref = lookupReferenceReasoning(provider, id)
+  if (ref) {
+    for (const [k, v] of Object.entries(ref)) {
+      if (source[k] === undefined) source[k] = v
+    }
+  }
+  const isReasoning = capabilities?.reasoning === true ||
+    Boolean(source.supportsReasoning) ||
+    /(^|[_-])(o1|o3|r1|qwq|thinking)($|[_-])/i.test(id) ||
+    id.toLowerCase().includes('reasoner') ||
+    id.toLowerCase().includes('-r1') ||
+    id.toLowerCase().includes('thinking')
+
+  if (!isReasoning && Object.keys(source).length === 0) {
+    return undefined
+  }
+
+  source.supportsReasoning = true
+  if (!source.paramFormat) {
+    const idLower = id.toLowerCase()
+    const provLower = String(provider || '').toLowerCase()
+    if (provLower === 'openai' || /(^|[_-])(o1|o3)($|[_-])/i.test(idLower)) {
+      source.paramFormat = 'effort'
+      source.defaultEffort = source.defaultEffort || 'medium'
+      source.supportedEfforts = source.supportedEfforts || ['low', 'medium', 'high']
+    } else if (provLower === 'anthropic' || idLower.includes('claude-3-7') || idLower.includes('claude-3.7')) {
+      source.paramFormat = 'budget_tokens'
+      source.defaultBudget = source.defaultBudget || 4096
+      source.maxBudget = source.maxBudget || 64000
+    } else if (provLower === 'deepseek' && (idLower.includes('reasoner') || idLower.includes('r1'))) {
+      source.paramFormat = 'boolean'
+    } else if (idLower.includes('thinking') || idLower.includes('r1') || idLower.includes('qwq')) {
+      source.paramFormat = 'budget_tokens'
+      source.defaultBudget = source.defaultBudget || 4096
+      source.maxBudget = source.maxBudget || 32768
+    } else {
+      source.paramFormat = 'boolean'
+    }
+  }
+  return source
+}
+
 export function normalizeModel(provider, row) {
   if (typeof row === 'string') {
     row = { id: row }
@@ -306,11 +373,15 @@ export function normalizeModel(provider, row) {
   const effectiveProvider = (typeof provider === 'string' && provider.trim())
     ? provider.trim()
     : (typeof row.provider === 'string' ? row.provider.trim() : '')
-  const id = String(row.id ?? row.slug ?? row.model ?? '').trim()
+  const id = String(row.id ?? row.slug ?? row.model ?? (typeof row.name === 'string' && row.name.startsWith('models/') ? row.name.replace(/^models\//, '') : '')).trim()
   if (!id) return null
   const name = String(row.name ?? row.displayName ?? row.display_name ?? id).trim() || id
-  const contextWindow = firstNumber(row, FIELD_ALIASES.contextWindow)
-  const maxTokens = firstNumber(row, FIELD_ALIASES.maxTokens)
+  const rawContextWindow = firstNumber(row, FIELD_ALIASES.contextWindow)
+  const rawMaxOutputTokens = firstNumber(row, FIELD_ALIASES.maxOutputTokens)
+  const refLimits = lookupReferenceLimits(effectiveProvider || provider, id)
+  const contextWindow = rawContextWindow ?? refLimits?.contextWindow
+  const maxOutputTokens = rawMaxOutputTokens ?? refLimits?.maxOutputTokens
+  const maxTokens = maxOutputTokens ?? firstNumber(row, FIELD_ALIASES.maxTokens)
   const capabilities = normalizeCapabilities(row)
   const pricing = normalizePricing(row) ?? lookupReferencePricing(effectiveProvider || provider, id)
   const tags = normalizeTags(row)
@@ -318,6 +389,7 @@ export function normalizeModel(provider, row) {
   const lifecycle = normalizeLifecycle(row)
   const input = normalizeInput(row)
   const compat = normalizeCompat(effectiveProvider, row, id)
+  const reasoningConfig = normalizeReasoningConfig(effectiveProvider || provider, row, id, capabilities)
   const firstSeenAt = normalizeDate(row.firstSeenAt)
   const createdAt = normalizeDate(row.createdAt ?? row.created)
   return {
@@ -327,11 +399,13 @@ export function normalizeModel(provider, row) {
     ...(firstSeenAt === undefined ? {} : { firstSeenAt }),
     ...(createdAt === undefined ? {} : { createdAt }),
     ...(contextWindow === undefined ? {} : { contextWindow }),
+    ...(maxOutputTokens === undefined ? {} : { maxOutputTokens }),
     ...(maxTokens === undefined ? {} : { maxTokens }),
     ...(row.description ? { description: String(row.description) } : {}),
     ...(input === undefined ? {} : { input }),
     ...(compat === undefined ? {} : { compat }),
     ...(capabilities === undefined ? {} : { capabilities }),
+    ...(reasoningConfig === undefined ? {} : { reasoningConfig }),
     ...(pricing === undefined ? {} : { pricing }),
     ...(tags === undefined ? {} : { tags }),
     ...(aliases.length === 0 ? {} : { aliases }),
@@ -410,6 +484,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/deepseek-v4-flash',
     name: 'DeepSeek V4 Flash',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { vision: true, code: true, reasoning: true },
     input: ['text', 'image'],
@@ -418,6 +493,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/deepseek-v4-pro',
     name: 'DeepSeek V4 Pro',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { vision: true, code: true, reasoning: true },
     input: ['text', 'image'],
@@ -426,6 +502,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/glm-5.2',
     name: 'GLM 5.2',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { code: true },
     input: ['text'],
@@ -434,6 +511,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/kimi-k3',
     name: 'Kimi K3',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { reasoning: true },
     input: ['text'],
@@ -442,6 +520,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/kimi-k2.7-code',
     name: 'Kimi K2.7 Code',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { code: true, tools: true },
     input: ['text'],
@@ -450,6 +529,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/kimi-k2.6',
     name: 'Kimi K2.6',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { code: true },
     input: ['text'],
@@ -458,6 +538,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/qwen3.7-max',
     name: 'Qwen 3.7 Max',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { vision: true },
     input: ['text', 'image'],
@@ -466,6 +547,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/qwen3.7-plus',
     name: 'Qwen 3.7 Plus',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { vision: true },
     input: ['text', 'image'],
@@ -474,6 +556,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/minimax-m3',
     name: 'MiniMax M3',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { reasoning: true },
     input: ['text'],
@@ -482,6 +565,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/mimo-v2.5',
     name: 'MiMo V2.5',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { code: true },
     input: ['text'],
@@ -490,6 +574,7 @@ export const CLINEPASS_MODELS = Object.freeze([
     id: 'cline-pass/mimo-v2.5-pro',
     name: 'MiMo V2.5 Pro',
     contextWindow: 200000,
+    maxOutputTokens: 8192,
     maxTokens: 8192,
     capabilities: { code: true, reasoning: true },
     input: ['text'],
@@ -546,7 +631,8 @@ export function parseClinePassModels(includedInput) {
         id: `cline-pass/${idPart}`,
         name,
         contextWindow: 200000,
-        maxTokens: 8192,
+        maxOutputTokens: 8192,
+    maxTokens: 8192,
         capabilities: {},
         input: ['text'],
       })

package/lib/provider-service.js

@@ -1,4 +1,5 @@
 import { listApiKeyProviders } from './inventory.js'
+import { lookupReferenceLimits, lookupReferenceReasoning, clampOutputTokens, adaptReasoningPayload } from './specs-database.js'
 
 export function createProviderService(ctx) {
   const readSettings = () => {
@@ -20,10 +21,53 @@ export function createProviderService(ctx) {
 
   const listProviders = () => listApiKeyProviders(readLlm(), readSettings())
 
+  const getProvider = (provider) => listProviders().find((row) => row.provider === provider) ?? null
+
+  const getLimits = (provider, modelId) => {
+    const p = getProvider(provider)
+    const catalog = Array.isArray(p?.models) ? p.models : []
+    const found = catalog.find((m) => (m?.id === modelId || m?.name === modelId))
+    if (found && (found.contextWindow !== undefined || found.maxOutputTokens !== undefined || found.maxTokens !== undefined)) {
+      return {
+        contextWindow: found.contextWindow,
+        maxOutputTokens: found.maxOutputTokens ?? found.maxTokens,
+      }
+    }
+    return lookupReferenceLimits(provider, modelId) ?? null
+  }
+
+  const clampCompletionTokens = (provider, modelId, requestedTokens) => {
+    const limits = getLimits(provider, modelId)
+    return clampOutputTokens(limits?.maxOutputTokens, requestedTokens)
+  }
+
+  const getReasoningConfig = (provider, modelId) => {
+    const p = getProvider(provider)
+    const catalog = Array.isArray(p?.models) ? p.models : []
+    const found = catalog.find((m) => (m?.id === modelId || m?.name === modelId))
+    if (found?.reasoningConfig) {
+      return found.reasoningConfig
+    }
+    return lookupReferenceReasoning(provider, modelId) ?? null
+  }
+
+  const adaptReasoning = (provider, modelId, options = {}) => {
+    const p = getProvider(provider)
+    const catalog = Array.isArray(p?.models) ? p.models : []
+    const found = catalog.find((m) => (m?.id === modelId || m?.name === modelId)) || {
+      provider,
+      id: modelId,
+      reasoningConfig: getReasoningConfig(provider, modelId),
+    }
+    return adaptReasoningPayload(found, options)
+  }
+
   return Object.freeze({
     listProviders,
-    getProvider(provider) {
-      return listProviders().find((row) => row.provider === provider) ?? null
-    },
+    getProvider,
+    getLimits,
+    clampCompletionTokens,
+    getReasoningConfig,
+    adaptReasoningPayload: adaptReasoning,
   })
 }

package/lib/reconcile.js

@@ -30,6 +30,9 @@ export function reconcileModels(provider, current, discovered, {
         ...(row.compat || fresh.compat ? {
           compat: { ...(fresh.compat ?? {}), ...(row.compat ?? {}) },
         } : {}),
+        ...(row.reasoningConfig || fresh.reasoningConfig ? {
+          reasoningConfig: { ...(fresh.reasoningConfig ?? {}), ...(row.reasoningConfig ?? {}) },
+        } : {}),
         ...(fresh.input?.length ? { input: fresh.input } : row.input ? { input: row.input } : {}),
       }
       next.push(merged)

package/lib/specs-database.js

@@ -0,0 +1,223 @@
+function cleanId(id) {
+  if (typeof id !== 'string') return ''
+  return id
+    .toLowerCase()
+    .trim()
+    .replace(/^models\//, '')
+    .replace(/:latest$/, '')
+    .replace(/-latest$/, '')
+    .replace(/-(202\d{5}|202\d-\d{2}-\d{2}|\d{4})$/, '')
+}
+
+export const REFERENCE_SPECS = Object.freeze({
+  openai: {
+    'gpt-4o': { contextWindow: 128000, maxOutputTokens: 16384 },
+    'gpt-4o-mini': { contextWindow: 128000, maxOutputTokens: 16384 },
+    'o1': {
+      contextWindow: 200000,
+      maxOutputTokens: 100000,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'effort', defaultEffort: 'medium', supportedEfforts: ['low', 'medium', 'high'] },
+      compat: { supportsTemperature: false, supportsMaxTokens: false },
+    },
+    'o1-preview': {
+      contextWindow: 128000,
+      maxOutputTokens: 32768,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'effort', defaultEffort: 'medium', supportedEfforts: ['low', 'medium', 'high'] },
+      compat: { supportsTemperature: false, supportsMaxTokens: false },
+    },
+    'o1-mini': {
+      contextWindow: 128000,
+      maxOutputTokens: 65536,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'effort', defaultEffort: 'medium', supportedEfforts: ['low', 'medium', 'high'] },
+      compat: { supportsTemperature: false, supportsMaxTokens: false },
+    },
+    'o3-mini': {
+      contextWindow: 200000,
+      maxOutputTokens: 100000,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'effort', defaultEffort: 'medium', supportedEfforts: ['low', 'medium', 'high'] },
+      compat: { supportsTemperature: false, supportsMaxTokens: false },
+    },
+    'gpt-4-turbo': { contextWindow: 128000, maxOutputTokens: 4096 },
+    'gpt-4': { contextWindow: 8192, maxOutputTokens: 4096 },
+    'gpt-3.5-turbo': { contextWindow: 16385, maxOutputTokens: 4096 },
+    'chatgpt-4o-latest': { contextWindow: 128000, maxOutputTokens: 16384 },
+  },
+  anthropic: {
+    'claude-3-7-sonnet': {
+      contextWindow: 200000,
+      maxOutputTokens: 64000,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'budget_tokens', defaultBudget: 4096, maxBudget: 64000 },
+    },
+    'claude-3-5-sonnet': { contextWindow: 200000, maxOutputTokens: 8192 },
+    'claude-3-5-haiku': { contextWindow: 200000, maxOutputTokens: 8192 },
+    'claude-3-opus': { contextWindow: 200000, maxOutputTokens: 4096 },
+    'claude-3-haiku': { contextWindow: 200000, maxOutputTokens: 4096 },
+  },
+  deepseek: {
+    'deepseek-chat': { contextWindow: 64000, maxOutputTokens: 8192 },
+    'deepseek-reasoner': {
+      contextWindow: 64000,
+      maxOutputTokens: 8192,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'boolean' },
+    },
+    'deepseek-coder': { contextWindow: 16384, maxOutputTokens: 4096 },
+  },
+  google: {
+    'gemini-2.0-flash': { contextWindow: 1048576, maxOutputTokens: 8192 },
+    'gemini-2.0-flash-thinking-exp': {
+      contextWindow: 1048576,
+      maxOutputTokens: 65536,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'budget_tokens', defaultBudget: 8192, maxBudget: 65536 },
+    },
+    'gemini-2.0-flash-lite': { contextWindow: 1048576, maxOutputTokens: 8192 },
+    'gemini-1.5-pro': { contextWindow: 2097152, maxOutputTokens: 8192 },
+    'gemini-1.5-flash': { contextWindow: 1048576, maxOutputTokens: 8192 },
+    'gemini-1.5-flash-8b': { contextWindow: 1048576, maxOutputTokens: 8192 },
+  },
+  groq: {
+    'llama-3.3-70b-versatile': { contextWindow: 131072, maxOutputTokens: 32768 },
+    'llama-3.3-70b-specdec': { contextWindow: 8192, maxOutputTokens: 8192 },
+    'llama-3.1-70b-versatile': { contextWindow: 131072, maxOutputTokens: 32768 },
+    'llama-3.1-8b-instant': { contextWindow: 131072, maxOutputTokens: 8192 },
+    'llama3-70b-8192': { contextWindow: 8192, maxOutputTokens: 8192 },
+    'llama3-8b-8192': { contextWindow: 8192, maxOutputTokens: 8192 },
+    'mixtral-8x7b-32768': { contextWindow: 32768, maxOutputTokens: 32768 },
+    'gemma2-9b-it': { contextWindow: 8192, maxOutputTokens: 8192 },
+    'deepseek-r1-distill-llama-70b': {
+      contextWindow: 131072,
+      maxOutputTokens: 32768,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'budget_tokens', defaultBudget: 4096, maxBudget: 32768 },
+    },
+    'deepseek-r1-distill-qwen-32b': {
+      contextWindow: 131072,
+      maxOutputTokens: 32768,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'budget_tokens', defaultBudget: 4096, maxBudget: 32768 },
+    },
+    'qwen-2.5-32b': { contextWindow: 131072, maxOutputTokens: 8192 },
+    'qwen-2.5-coder-32b': { contextWindow: 131072, maxOutputTokens: 8192 },
+    'qwen-qwq-32b': {
+      contextWindow: 131072,
+      maxOutputTokens: 32768,
+      reasoningConfig: { supportsReasoning: true, paramFormat: 'budget_tokens', defaultBudget: 4096, maxBudget: 32768 },
+    },
+  },
+  xai: {
+    'grok-2': { contextWindow: 131072, maxOutputTokens: 8192 },
+    'grok-2-mini': { contextWindow: 131072, maxOutputTokens: 8192 },
+    'grok-2-vision': { contextWindow: 131072, maxOutputTokens: 8192 },
+    'grok-beta': { contextWindow: 131072, maxOutputTokens: 8192 },
+  },
+  mistral: {
+    'mistral-large': { contextWindow: 128000, maxOutputTokens: 8192 },
+    'mistral-small': { contextWindow: 128000, maxOutputTokens: 8192 },
+    'codestral': { contextWindow: 256000, maxOutputTokens: 8192 },
+    'ministral-8b': { contextWindow: 128000, maxOutputTokens: 8192 },
+    'ministral-3b': { contextWindow: 128000, maxOutputTokens: 8192 },
+    'pixtral-large': { contextWindow: 128000, maxOutputTokens: 8192 },
+    'pixtral-12b': { contextWindow: 128000, maxOutputTokens: 8192 },
+  },
+})
+
+export function lookupReferenceSpec(provider, modelId) {
+  if (typeof modelId !== 'string' || !modelId.trim()) return undefined
+  const provKey = String(provider || '').toLowerCase().trim()
+  const rawId = modelId.toLowerCase().trim()
+  const cleaned = cleanId(rawId)
+
+  // 1. Direct provider match
+  const table = REFERENCE_SPECS[provKey]
+  if (table) {
+    if (table[rawId]) return { ...table[rawId] }
+    if (table[cleaned]) return { ...table[cleaned] }
+    return undefined
+  }
+
+  // 2. Slash prefix (e.g. "anthropic/claude-3-7-sonnet" or "openai/o3-mini")
+  const slashIdx = rawId.indexOf('/')
+  if (slashIdx > 0) {
+    const subProv = rawId.slice(0, slashIdx)
+    const subId = rawId.slice(slashIdx + 1)
+    const subCleaned = cleanId(subId)
+    const subTable = REFERENCE_SPECS[subProv]
+    if (subTable) {
+      if (subTable[subId]) return { ...subTable[subId] }
+      if (subTable[subCleaned]) return { ...subTable[subCleaned] }
+    }
+  }
+
+  // 3. Fallback across all tables for aggregators (OpenRouter, ClineBot, CommandCode)
+  for (const [, t] of Object.entries(REFERENCE_SPECS)) {
+    if (t[rawId]) return { ...t[rawId] }
+    if (t[cleaned]) return { ...t[cleaned] }
+  }
+
+  return undefined
+}
+
+export function lookupReferenceLimits(provider, modelId) {
+  const spec = lookupReferenceSpec(provider, modelId)
+  if (!spec) return undefined
+  const res = {}
+  if (spec.contextWindow !== undefined) res.contextWindow = spec.contextWindow
+  if (spec.maxOutputTokens !== undefined) res.maxOutputTokens = spec.maxOutputTokens
+  return Object.keys(res).length > 0 ? res : undefined
+}
+
+export function lookupReferenceReasoning(provider, modelId) {
+  const spec = lookupReferenceSpec(provider, modelId)
+  return spec?.reasoningConfig ? structuredClone(spec.reasoningConfig) : undefined
+}
+
+export function lookupReferenceCompat(provider, modelId) {
+  const spec = lookupReferenceSpec(provider, modelId)
+  return spec?.compat ? structuredClone(spec.compat) : undefined
+}
+
+export function clampOutputTokens(modelOrLimit, requestedTokens) {
+  const limit = typeof modelOrLimit === 'number'
+    ? modelOrLimit
+    : Number(modelOrLimit?.maxOutputTokens ?? modelOrLimit?.maxTokens)
+  if (!Number.isFinite(limit) || limit <= 0) {
+    return { clamped: false, value: requestedTokens, limit: undefined }
+  }
+  const requested = Number(requestedTokens)
+  if (!Number.isFinite(requested) || requested <= 0) {
+    return { clamped: false, value: requestedTokens, limit }
+  }
+  if (requested > limit) {
+    return { clamped: true, value: limit, original: requested, limit }
+  }
+  return { clamped: false, value: requested, limit }
+}
+
+export function adaptReasoningPayload(model, userOptions = {}) {
+  const warnings = []
+  const payload = {}
+  const config = model?.reasoningConfig
+  const format = config?.paramFormat || 'none'
+  const compat = model?.compat || {}
+
+  if (format === 'effort') {
+    const effort = userOptions.reasoningEffort || config?.defaultEffort || 'medium'
+    payload.reasoning_effort = effort
+    if (compat.supportsTemperature === false && userOptions.temperature !== undefined && userOptions.temperature !== 1) {
+      warnings.push(`temperature ${userOptions.temperature} is not supported by model ${model.id}; omitted`)
+    }
+    if (compat.supportsMaxTokens === false && userOptions.maxTokens !== undefined) {
+      payload.max_completion_tokens = userOptions.maxTokens
+    }
+  } else if (format === 'budget_tokens') {
+    const maxBudget = config?.maxBudget || 64000
+    const defaultBudget = config?.defaultBudget || 4096
+    const budget = Math.min(Math.max(1, Number(userOptions.thinkingBudget || defaultBudget)), maxBudget)
+    if (model?.provider === 'google' || String(model?.id || '').startsWith('gemini')) {
+      payload.thinking_config = { thinking_budget: budget }
+    } else {
+      payload.thinking = { type: 'enabled', budget_tokens: budget }
+    }
+  } else if (format === 'boolean') {
+    payload.thinking = true
+  }
+
+  return { payload, warnings, format }
+}

package/lib/synchronizer-helpers.js

@@ -38,6 +38,18 @@ export function reliabilityOptions(config, options) {
   }
 }
 
+export function getOptionalService(ctx, name) {
+  try {
+    return ctx?.[name] ?? ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+  } catch {
+    try {
+      return ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+    } catch {
+      return undefined
+    }
+  }
+}
+
 export function settingsOf(ctx) {
   try {
     return ctx?.settings ?? ctx?.reflect?.get?.('settings') ?? (typeof ctx?.get === 'function' ? ctx.get('settings') : undefined)

package/lib/updater.js

@@ -1,3 +1,15 @@
+function getOptionalService(ctx, name) {
+  try {
+    return ctx?.[name] ?? ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+  } catch {
+    try {
+      return ctx?.reflect?.get?.(name) ?? (typeof ctx?.get === 'function' ? ctx.get(name) : undefined)
+    } catch {
+      return undefined
+    }
+  }
+}
+
 import { spawn } from 'node:child_process'
 import { existsSync, readFileSync } from 'node:fs'
 import { readFile } from 'node:fs/promises'
@@ -315,7 +327,7 @@ export function registerPluginUpdater(ctx, options = {}) {
   const handler = async (request, response) => {
     try {
       // 1. Core connection rejection check on all updater routes (#237)
-      const conn = ctx?.connection || ctx?.get?.('connection') || request?.connection
+      const conn = getOptionalService(ctx, 'connection') || request?.connection
       if (conn && typeof conn.requestRejection === 'function') {
         const rejection = conn.requestRejection(request)
         if (rejection !== undefined) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-model-sync",
-  "version": "0.5.14",
+  "version": "0.5.16",
   "description": "Automatic model catalog synchronization for API-key DeepSeek Harness providers.",
   "keywords": [
     "deepseek-harness",