@mangoc/dsh-sub2api-sync 0.2.5 → 0.2.7

package/README.md

@@ -122,7 +122,8 @@ Routes take effect on the next request. The route suffix is a SHA-256 truncation
 ## Model Experience
 
 - **Context blocks (short form)**: one provider per sub2api group, one relay key per provider, model catalog discovered from the instance; per-model `compat` inherits the user template. Reasoning efforts (`reasoningEfforts`) follow an explicit template when set (including `false`), and otherwise auto-fill per model from the models.dev enrichment.
-- **Reasoning-effort auto-fill**: the models.dev `reasoning_options` effort enumeration maps onto pi-ai's `reasoningEfforts` (`none` → `off: null`, same-name passthrough otherwise). Merging is first-party first: a vendor-official entry (openai/anthropic/google/zhipuai/zai/deepseek/moonshotai/xai/alibaba, vendor-governed cloud hosting) wins outright — including when it carries no effort data (budget-only claude/gemini), which silences aggregator-invented vocabularies; with no first-party entry, aggregator sources intersect. Toggle/budget-only reasoning (claude, gemini, deepseek-v3.2, kimi-k2.x) has no effort-string wire shape and stays template/`compat`-owned.
+- **Reasoning-effort auto-fill**: the models.dev `reasoning_options` effort enumeration maps onto pi-ai's `reasoningEfforts` (`none` → `off: null`, same-name passthrough otherwise). Merging is first-party first: a vendor-official entry (openai/anthropic/google/zhipuai/zai/deepseek/moonshotai/xai/alibaba, vendor-governed cloud hosting) wins outright — including when it carries no effort data (budget-only older Claude/Gemini entries), which silences aggregator-invented vocabularies; with no first-party entry, aggregator sources intersect. Toggle/budget-only reasoning (older Claude, Gemini, deepseek-v3.2, kimi-k2.x) has no effort-string wire shape and stays template/`compat`-owned.
+- **Native Claude reasoning efforts**: for confirmed adaptive-thinking Claude models with catalog effort metadata (such as Opus 4.6/4.7/4.8 and Sonnet 4.6), syncing over `anthropic-messages` also writes per-model `compat.forceAdaptiveThinking: true`. This SDK switch sends Claude's native `thinking: { type: "adaptive" }` and `output_config: { effort: "max" }` when Max is selected: no OpenAI protocol bridge and no Max-to-token-budget downgrade. Explicit template compat values (including `false`) win; disabled reasoning, legacy/unknown models, missing catalog effort metadata, and other protocols do not auto-enable it. Update the plugin and sync again to repair existing managed routes; subsequent syncs remain idempotent. The usage log records the declared effort, not a measurement of actual thinking.
 - **Display-name enrichment**: the models.dev `name` lands on each pi-ai model entry's `name` field — the display name the model selector and the model-catalog editor (显示名称 column) show. The models.dev " (latest)" family marker is stripped, a name repeating the id is dropped, merging is first-party first (official name overrides, aggregator names fill absences), and a name can be an id's only enriched fact.
 - Model ids come verbatim from the instance; the provider display name is the `providerNameTemplate` rendering of the key name (fallback `Group <id>`).
 

package/README.zh.md

@@ -92,7 +92,8 @@ dsh plugin --profile web add ./packages/llm/sub2api-sync
 
 - 每分组一个 provider、一把 relay key；模型目录来自实例发现；逐模型 `compat` 继承用户模板；推理档位 `reasoningEfforts` 模板显式设置优先，未设置时按模型自动填充（见下）；
 - **模型元数据补全（C4）**：同步时按模型 id 从 models.dev 目录补全**显示名称**（`name`，即模型目录编辑器「显示名称」列）/`contextWindow`/`maxTokens`/图片输入/推理档位——`enrichModelMeta` 开启（缺省）时联网拉取（24h 进程内缓存），失败自动降级到**随包内置快照**（`data/models-dev.snapshot.json`，`pnpm --filter @mangoc/dsh-sub2api-sync models:refresh` 重新生成）；关闭则只用内置快照（离线确定）；两者都未命中的模型保持路由级缺省 128000/8192/纯文本；
-- **推理档位自动填充**：取 models.dev 条目的 `reasoning_options` 中 `effort` 枚举，映射为 pi-ai 的 `reasoningEfforts`（`none` → `off: null`，其余同名直传）。合并按 **first-party 优先**：官方厂商条目（openai/anthropic/google/zhipuai/zai/deepseek/moonshotai/xai/alibaba 等及官方云托管）存在即胜出——官方条目无 effort 数据（如 claude/gemini 的 budget-only）同样压制聚合商的虚报词表；无官方条目时聚合商来源取**交集**。优先级：**模板显式设置（含 `false`）> 每模型自动填充 > 不写字段**；显示名 = `providerNameTemplate` 渲染的 key 名。
+- **推理档位自动填充**：取 models.dev 条目的 `reasoning_options` 中 `effort` 枚举，映射为 pi-ai 的 `reasoningEfforts`（`none` → `off: null`，其余同名直传）。合并按 **first-party 优先**：官方厂商条目（openai/anthropic/google/zhipuai/zai/deepseek/moonshotai/xai/alibaba 等及官方云托管）存在即胜出——官方条目无 effort 数据（如旧版 Claude/Gemini 的 budget-only 条目）同样压制聚合商的虚报词表；无官方条目时聚合商来源取**交集**。优先级：**模板显式设置（含 `false`）> 每模型自动填充 > 不写字段**；显示名 = `providerNameTemplate` 渲染的 key 名。
+- **Claude 原生推理档位**：对已确认支持 adaptive thinking、且目录有 effort 元数据的 Claude 模型（如 Opus 4.6/4.7/4.8、Sonnet 4.6），仅在 `anthropic-messages` 协议下逐模型自动补齐 `compat.forceAdaptiveThinking: true`。该开关让 SDK 直接发送 Claude 原生的 `thinking: { type: "adaptive" }` 与 `output_config: { effort: "max" }`（以选择 Max 为例），不经过 OpenAI 协议转换，也不将 Max 降成 token budget。用户模板中同名 compat 字段（包括显式 `false`）优先；关闭 reasoning、旧模型、未知别名/新版本、缺少目录 effort 元数据或其他协议都不会自动开启。更新插件并重新同步即可修复已有受管路由，重复同步不重复写入；无需修改 sub2api 的记录值。注意：用量日志的 effort 是请求声明档位，不是实际思考量的测量值。
 - **显示名称富化**：取 models.dev 条目的 `name` 写入 pi-ai 模型条目的 `name` 字段（模型选择器与「已自定义模型目录」编辑器的显示名）；剥离 models.dev 的 " (latest)" 家族标记，名称与模型 id 相同（无信息量）时不写；合并同样 first-party 优先（官方名称覆盖，聚合商名称仅补缺），纯名称条目也参与富化。
 
 ## Known Limitations and Deferred Work
@@ -104,7 +105,7 @@ dsh plugin --profile web add ./packages/llm/sub2api-sync
 - **客户端半部只允许基线外部模块**（React / Cordis / ui-primitives / ui-slots / client-store / ui-dockkit）——客户端代码不得值导入 `config.ts`（会把 schemastery 拽进浏览器 bundle，动态模块表无此行，加载即静默失败）；`test/client-bundle.test.ts` 是产物守卫。
 - 实例前置 Cloudflare 时，需在 zone 放行 `/api/v1/*` 的宿主侧调用（WAF skip 规则）。
 - 元数据补全会向 models.dev 发起一次公开目录请求（不带任何凭据与对话内容）；`enrichModelMeta: false` 可完全离线（仅内置快照）。
-- 推理档位自动填充只覆盖 models.dev 提供 `effort` 枚举的模型（OpenAI gpt-5 系/o3、GLM 5.x、Qwen3.8、DeepSeek V4 系等）；`toggle`/`budget_tokens` 型推理（claude 系、gemini 系、deepseek-v3.2、kimi-k2.x 等）的线上形态不是 effort 字符串，`reasoningEfforts` 无法表达，仍由模板/`compat` 管辖。
+- 推理档位自动填充只覆盖 models.dev 提供 `effort` 枚举的模型（OpenAI gpt-5 系/o3、GLM 5.x、Qwen3.8、DeepSeek V4 系等）；`toggle`/`budget_tokens` 型推理（旧版 Claude、Gemini、deepseek-v3.2、kimi-k2.x 等）的线上形态不是 effort 字符串，`reasoningEfforts` 无法表达，仍由模板/`compat` 管辖。
 - 用量查询（`GET /v1/usage`）是已决策的非目标。
 - 与社区插件 `GodD6366/dsh-sub2api` 同装：路由名不冲突（本插件 `sub2api-<10-hex>` 摘要形态），各自管理条目与凭据引用。
 

package/lib/client.cjs

@@ -793,8 +793,13 @@ const CSS = `
 
 /* The options panel: fixed-positioned from the trigger (escapes the modal
    card's overflow clip), width-matched to it, search pinned on top, options
-   scrolling below. */
+   scrolling below. position: fixed is load-bearing: MenuSurface defaults to
+   position: relative, and without this rule the panel becomes a flex item
+   inside the form row while the viewport-coordinate top/left from
+   useAnchoredPosition act as relative offsets — the panel displaces across
+   the dialog and stays clipped by the card (the blank floating box bug). */
 .s2s-groupPanel {
+  position: fixed;
   display: flex;
   flex-direction: column;
   box-sizing: border-box;
@@ -1396,8 +1401,8 @@ function GroupPicker({ groups, value, placeholder, searchPlaceholder, emptyLabel
 			ref: panelRef,
 			className: "s2s-groupPanel",
 			role: "listbox",
-			style: position === null ? { visibility: "hidden" } : {
-				...position,
+			style: {
+				...position === null ? { visibility: "hidden" } : position,
 				...triggerWidth > 0 ? { width: `${triggerWidth}px` } : {},
 				maxHeight: `${maxHeight}px`
 			},

package/lib/client.js

@@ -799,8 +799,13 @@ const CSS = `
 
 /* The options panel: fixed-positioned from the trigger (escapes the modal
    card's overflow clip), width-matched to it, search pinned on top, options
-   scrolling below. */
+   scrolling below. position: fixed is load-bearing: MenuSurface defaults to
+   position: relative, and without this rule the panel becomes a flex item
+   inside the form row while the viewport-coordinate top/left from
+   useAnchoredPosition act as relative offsets — the panel displaces across
+   the dialog and stays clipped by the card (the blank floating box bug). */
 .s2s-groupPanel {
+  position: fixed;
   display: flex;
   flex-direction: column;
   box-sizing: border-box;
@@ -1402,8 +1407,8 @@ function GroupPicker({ groups, value, placeholder, searchPlaceholder, emptyLabel
 			ref: panelRef,
 			className: "s2s-groupPanel",
 			role: "listbox",
-			style: position === null ? { visibility: "hidden" } : {
-				...position,
+			style: {
+				...position === null ? { visibility: "hidden" } : position,
 				...triggerWidth > 0 ? { width: `${triggerWidth}px` } : {},
 				maxHeight: `${maxHeight}px`
 			},

package/lib/index.js

@@ -300,8 +300,38 @@ const THINKING_LEVELS = [
 	"xhigh",
 	"max"
 ];
+/**
+* Native adaptive-thinking entries confirmed by pi-ai 1.1.0's Anthropic
+* catalog. Keep this allowlist conservative: an effort enum alone is NOT proof
+* of adaptive support (Opus 4.5 has native effort but still uses budget thinking,
+* and aggregators publish effort enums even for older Claude models).
+* Unknown releases/aliases stay explicit-template owned rather than guessed.
+*/
+const ADAPTIVE_CLAUDE_MODELS = new Set([
+	"claude-opus-4-6",
+	"claude-opus-4-7",
+	"claude-opus-4-8",
+	"claude-opus-5",
+	"claude-opus-5-5",
+	"claude-sonnet-4-6",
+	"claude-sonnet-5",
+	"claude-sonnet-5-5",
+	"claude-haiku-5-5",
+	"claude-fable-5",
+	"claude-fable-5-1"
+]);
+/** Pair catalog effort metadata with its native Anthropic wire mode. */
+function modelCompat(input, model) {
+	const template = input.template?.compat;
+	const canonicalId = model.id.toLowerCase().replaceAll(".", "-").replace(/(?:-(?:fast|think|thinking|\d{8})|@(?:default|eu))$/, "");
+	const adaptive = input.protocol === "anthropic-messages" && input.template?.reasoningEfforts !== false && model.reasoningEfforts !== undefined && ADAPTIVE_CLAUDE_MODELS.has(canonicalId);
+	if (!adaptive && template === undefined) return undefined;
+	return {
+		...adaptive ? { forceAdaptiveThinking: true } : {},
+		...template
+	};
+}
 function buildProviderProfile(input) {
-	const compat = input.template?.compat;
 	const reasoning = input.template?.reasoningEfforts;
 	const override = input.retryPolicy ?? {};
 	const retryPolicy = {
@@ -319,15 +349,18 @@ function buildProviderProfile(input) {
 		apiKeyEnv: input.apiKeyEnv,
 		api: input.protocol,
 		baseURL: baseURLForProtocol(input.instance, input.protocol),
-		models: input.models.map((model) => ({
-			id: model.id,
-			...model.name !== undefined ? { name: model.name } : {},
-			...model.contextWindow !== undefined ? { contextWindow: model.contextWindow } : {},
-			...model.maxTokens !== undefined ? { maxTokens: model.maxTokens } : {},
-			...model.imageInput === true ? { input: ["text", "image"] } : {},
-			...compat !== undefined ? { compat: { ...compat } } : {},
-			...reasoning === false ? { reasoningEfforts: false } : reasoning !== undefined ? { reasoningEfforts: { ...reasoning } } : model.reasoningEfforts !== undefined ? { reasoningEfforts: { ...model.reasoningEfforts } } : {}
-		})),
+		models: input.models.map((model) => {
+			const compat = modelCompat(input, model);
+			return {
+				id: model.id,
+				...model.name !== undefined ? { name: model.name } : {},
+				...model.contextWindow !== undefined ? { contextWindow: model.contextWindow } : {},
+				...model.maxTokens !== undefined ? { maxTokens: model.maxTokens } : {},
+				...model.imageInput === true ? { input: ["text", "image"] } : {},
+				...compat !== undefined ? { compat } : {},
+				...reasoning === false ? { reasoningEfforts: false } : reasoning !== undefined ? { reasoningEfforts: { ...reasoning } } : model.reasoningEfforts !== undefined ? { reasoningEfforts: { ...model.reasoningEfforts } } : {}
+			};
+		}),
 		...input.template?.reasoning !== undefined ? { reasoning: input.template.reasoning } : {},
 		defaultContextWindow: input.defaultContextWindow ?? 128e3,
 		defaultMaxTokens: input.defaultMaxTokens ?? 8192,

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@mangoc/dsh-sub2api-sync",
   "description": "Sync one sub2api instance's groups, keys, and models into dsh-llm-pi-ai provider routes: panel-driven login (password / OAuth), one relay key one provider.",
-  "version": "0.2.5",
+  "version": "0.2.7",
   "license": "MIT",
   "author": "Shuai-S",
   "homepage": "https://github.com/Shuai-S/dsh-sub2api-sync#readme",