dsh-model 0.22.3 → 0.23.2

package/README.en.md

@@ -2,9 +2,9 @@
 
 # dsh-model
 
-**One address for all your models: subscriptions, free tiers and local models, with daily check-in credits claimed for you.**
+**📅 Daily check-in credits claimed for you · 🔁 one address that relays every model · 🧩 free and subscription models, your pick**
 
-One OpenAI-compatible endpoint, one key, dozens of models. dsh, Cline, Continue, Cherry Studio, scripts: anything that speaks the OpenAI API just works.
+Install once and all three are taken care of. dsh gets a management page; without dsh, the command line does everything.
 
 <p>
   <a href="LICENSE"><img src="https://img.shields.io/github/license/AIcivilization/dsh-model" alt="MIT license"></a>
@@ -19,22 +19,45 @@ English · [中文](README.md)
 
 </div>
 
-## Why
+## 📅 1. Daily check-in: credits land on their own
 
-You may have several at once: Codex from ChatGPT Plus, a Kimi coding plan, free WorkBuddy credits, free Gemini and Groq keys, and Ollama on your machine. Each has its own API and sign-in, every tool needs its own setup, and when a quota runs out you switch by hand.
+**The most hands-off part: no more remembering to claim the daily check-in credits.**
 
-dsh-model turns them into **one address and one kind of key**:
+| Product | How it connects | Every day |
+|---|---|---|
+| **Trae** | Detected automatically if Trae is installed and signed in; nothing to do | Check-in credits (200 a day in testing) |
+| **Qoder** | Paste a personal access token once (China or international) | That day's check-in credits |
+| **WorkBuddy (China)** | Sign in once | Check-in credits |
+
+- **Automatic by default**: done at a random 5–55 minutes after midnight Beijing time; failures retry an hour later; if the computer was off, it catches up within a minute of starting.
+- **You can see it**: one line at the top of the settings page, "Checked in today: …", with a one-click "Check in all" for anything missed, plus a one-time notice in the corner of dsh.
+- **Check-in only, credits untouched**: they stay on your account for use in each product's own app. Command line: `dsh-model checkin`.
+
+## 🔁 2. Relay: one address, one key, every model
 
-- 🔌 **Connect once, use everywhere**: every model shows up in dsh, grouped by source; any other tool uses the same address and key.
-- 🆓 **Get the most from free tiers**: free Gemini / Groq / OpenRouter keys, local models and WorkBuddy credits; every source is tagged Free / Free · limited / Paid.
-- 📅 **Daily check-in, credits claimed for you**: WorkBuddy (China), Qoder and Trae are checked in automatically every day.
-- 🔁 **Rate limits are not the end**: build a combo (say Codex → Kimi → WorkBuddy) and requests move to the next model on a rate limit or used-up quota; a model offered by several sources also switches automatically.
-- 📊 **Know where you stand**: plan usage with reset countdowns, remaining credits and when they expire, success rate and speed per model, recent failures with advice, and a notice when quota runs low.
-- 🔒 **Secure by default**: a key is always required, one per device, revocable at any time; every sign-in happens on your own machine.
-- 🖥️ **Laptop or server**: Windows, macOS, Linux; on a VPS (with [dsh-vps](https://github.com/AIcivilization/dsh-vps)) it opens an https address you can use from anywhere.
-- 🧹 **Installs in one command, uninstalls cleanly**: dsh's config is restored byte for byte.
+Your model sources each have their own API and sign-in. dsh-model puts them behind **one OpenAI-compatible address**:
+
+- **Shows up in dsh**: every model, grouped by source, with nothing to configure.
+- **Works in other tools**: Cline, Continue, Cherry Studio, scripts: anything that speaks the OpenAI API uses the same address and key.
+- **Rate limits are not the end**: build a combo (say Codex → Kimi → Gemini) and the next model takes over on a rate limit, used-up quota or error.
+- **Know where you stand**: plan usage with reset countdowns, success rate and speed per model, recent failures with what to do, and a notice when quota runs low.
+- **Secure by default**: a key is always required, one per device, revocable any time; every sign-in happens on your own machine.
+- **Laptop or server**: Windows, macOS, Linux; on a VPS (with [dsh-vps](https://github.com/AIcivilization/dsh-vps)) it opens an https address you can use from anywhere. Uninstall restores dsh's config byte for byte.
+
+## 🧩 3. Models: free or subscription, connect what you have
+
+| Kind | Sources |
+|---|---|
+| **Free (suggested)** | **SiliconFlow** (reachable from China; some models free after identity verification), **Gemini** (if you can reach Google; the best free models) |
+| Free · limited | Groq, OpenRouter (`:free` models) |
+| Local | Ollama, LM Studio: appear once started, no sign-in |
+| Subscription | Codex (ChatGPT Plus / Pro), Kimi Code, Grok, Muse, Devin: sign in from the management page |
+| Pay as you go | DeepSeek official, OpenCode Zen, any OpenAI-compatible service |
+
+Every source is tagged Free / Free · limited / Paid; see [Sources and cost](#sources-and-cost) below. dsh-model only connects; it provides no quota of its own.
+
+> WorkBuddy: since 2026-10 it recognises conversations sent by dsh and blocks them, so dsh-model uses it for check-in only and does not connect its models by default.
 
-No dsh? It still works: the command line manages everything; dsh just adds a management page.
 
 ## Quick start
 
@@ -48,7 +71,7 @@ On a dsh-vps server, use `sudo dsh-model setup`. On Windows, run it in a normal
 `setup` is safe to re-run. It:
 
 1. downloads and verifies the engine and runs it on `127.0.0.1:8317/v1`, with key auth enforced;
-2. connects WorkBuddy (dsh-model's own bridge). Setup asks for no keys: turn on OpenCode Zen and other key-based sources from the management page when you need them;
+2. starts dsh-model's daemon (the management page, check-in and sign-ins run on it). Setup asks for no keys: for a free source we suggest **SiliconFlow** (reachable from China) or **Gemini** (if you can reach Google), added from the management page. WorkBuddy is used for check-in only; its models are not connected to dsh by default;
 3. installs the management page into dsh and syncs the models into dsh;
 4. on a VPS, also opens a public address `https://<dsh's domain>:9443/v1` (or `<IP>.sslip.io` when there is no domain).
 
@@ -56,12 +79,12 @@ On a dsh-vps server, use `sudo dsh-model setup`. On Windows, run it in a normal
 
 | Section | What it holds |
 |---|---|
+| **Daily check-in** | WorkBuddy (China), Qoder and Trae are checked in automatically every day (or with one click); the card shows what you got today |
 | **Endpoint** | Local and public addresses, each with a Copy button |
 | **Sources** | One switch per provider: on connects (and signs in if needed), off pauses and keeps the sign-in. Each row is tagged Free / Free · limited / Paid and shows plan usage with reset countdowns, or remaining credits and the next batch to expire; a used-up plan shows when it comes back |
 | **Models** | Pick which models of each source appear in dsh (5 suggested per source by default); the rest stay available through the endpoint; a "Test" button per model |
 | **Combos** | Put several models in order and it becomes one model in dsh: the first is used, and the next takes over on a rate limit, used-up quota or error. Other tools use the model name `combo/<id>` |
 | **Auto switch** | When a model is offered by several sources, dsh gets an "Auto switch (multiple sources)" group that moves to another source when one is rate-limited or failing |
-| **Daily check-in** | WorkBuddy (China), Qoder and Trae are checked in automatically every day (or with one click); the card shows what you got today |
 | **Recent failures** | Failed requests in the last 24 hours: which model, why, and what to do (no conversation content kept) |
 | **API keys** | OpenAI-style, one per device. Requests, success rate, latency and speed over 24 hours; copy, rotate, revoke |
 | **Model stats** | Success rate, latency and speed per model |

package/README.md

@@ -2,9 +2,9 @@
 
 # dsh-model
 
-**一个地址，用上你所有的模型：订阅、免费额度、本机模型，签到积分自动领。**
+**📅 每天自动签到领积分 · 🔁 一个地址中转所有模型 · 🧩 免费和订阅模型随你接**
 
-一个 OpenAI 兼容地址，一把 key，几十个模型。dsh、Cline、Continue、Cherry Studio、脚本……凡是支持 OpenAI 接口的软件都能直接用。
+装一次，三件事都替你做好。dsh 里有图形管理页，没有 dsh 也能用命令行。
 
 <p>
   <a href="LICENSE"><img src="https://img.shields.io/github/license/AIcivilization/dsh-model" alt="MIT license"></a>
@@ -19,22 +19,45 @@
 
 </div>
 
-## 为什么用它
+## 📅 1. 每日签到：积分自动到账
 
-你手上可能有好几样：ChatGPT Plus 带的 Codex、Kimi 的编程套餐、WorkBuddy 的免费积分、Gemini 和 Groq 的免费 key，还有本机的 Ollama。每家接口和登录方式都不一样，每个软件都得分别配置，额度用完了还要手动换。
+**最省心的一件事：每天的签到积分，不用再记着去领。**
 
-dsh-model 把它们合成**一个地址、一种 key**：
+| 产品 | 怎么接入 | 每天 |
+|---|---|---|
+| **Trae** | 本机装了 Trae 并登录过，自动识别，什么都不用做 | 签到送积分（实测每天 200） |
+| **Qoder** | 粘贴一次个人访问令牌（国内版、国际版都行） | 领当天的签到积分 |
+| **WorkBuddy 国内版** | 登录一次 | 签到送积分 |
+
+- **自动签到默认打开**：北京时间零点后随机延后 5–55 分钟签好；失败的一小时后重试；电脑关机错过了，开机一分钟内补签。
+- **结果看得见**：设置页顶部一行「今日已完成签到：…」，没签上的可以「一键签到」；签完在 dsh 右下角提示一次。
+- **只签到，不碰积分**：积分留在你的账号里，照常在各家官方软件里用。命令行：`dsh-model checkin`。
+
+## 🔁 2. 中转：一个地址、一把 key，用上所有模型
 
-- 🔌 **一处接入，处处可用**：dsh 里自动出现全部模型，按来源分组；其他软件填上同一个地址和 key 就行。
-- 🆓 **把免费的用足**：Gemini / Groq / OpenRouter 免费 key、本机模型、WorkBuddy 积分都能接；每个来源标出「免费 / 免费·额度少 / 需付费」。
-- 📅 **每日签到，自动领积分**：WorkBuddy 国内版、Qoder、Trae 每天自动签到，不用记。
-- 🔁 **不怕限流**：自己排一个组合（如 Codex → Kimi → WorkBuddy），一个限流或额度用完就自动换下一个；同一个模型在几个来源都有时也会自动切换。
-- 📊 **心里有数**：套餐用量和重置倒计时、剩余积分和到期时间、每个模型的成功率和速度、最近失败的原因和建议；额度快用完时提醒。
-- 🔒 **默认安全**：始终要求 key，每台设备一把，可随时吊销；登录全在本机完成，不经过第三方。
-- 🖥️ **电脑和服务器都行**：Windows、macOS、Linux；装在 VPS 上（配合 [dsh-vps](https://github.com/AIcivilization/dsh-vps)）自动开一个 https 地址，电脑、手机随处用。
-- 🧹 **装得上，卸得净**：一条命令安装；卸载时把 dsh 的配置逐字节还原。
+你手上的模型来源，各家接口、登录方式都不一样。dsh-model 把它们合成**一个 OpenAI 兼容地址**：
+
+- **dsh 里自动出现**：全部模型按来源分组，不用手动配置。
+- **其他软件也能用**：Cline、Continue、Cherry Studio、脚本……凡是支持 OpenAI 接口的，填同一个地址和 key 就行。
+- **不怕限流**：自己排一个组合（如 Codex → Kimi → Gemini），一个限流、额度用完或出错，就自动换下一个。
+- **心里有数**：套餐用量和重置倒计时、每个模型的成功率和速度、最近失败的原因和该怎么办；额度快用完时提醒。
+- **默认安全**：始终要求 key，每台设备一把，可随时吊销；登录全在本机完成。
+- **电脑和服务器都行**：Windows、macOS、Linux；装在 VPS 上（配合 [dsh-vps](https://github.com/AIcivilization/dsh-vps)）自动开一个 https 地址，电脑、手机随处用。卸载时把 dsh 的配置逐字节还原。
+
+## 🧩 3. 模型提供：免费的、订阅的都能接
+
+| 类型 | 来源 |
+|---|---|
+| **免费（推荐）** | **硅基流动**（国内直连，实名后部分模型免费）、**Gemini**（能上外网，免费模型里质量最好） |
+| 免费 · 额度少 | Groq、OpenRouter（`:free` 模型） |
+| 本机 | Ollama、LM Studio：启动后自动出现，不用登录 |
+| 订阅 | Codex（ChatGPT Plus / Pro）、Kimi Code、Grok、Muse、Devin：在管理页里登录即可 |
+| 按量付费 | DeepSeek 官方、OpenCode Zen、任意 OpenAI 兼容服务 |
+
+每个来源都标出「免费 / 免费·额度少 / 需付费」，详见下面的[来源与费用](#来源与费用)。dsh-model 只负责转接，不提供额度。
+
+> WorkBuddy：2026-10 起它会认出 dsh 发出的对话并拦下，所以 dsh-model 只用它来签到，模型默认不接进 dsh。
 
-没有 dsh 也能用：命令行就能管理全部功能，dsh 里多一个图形管理页而已。
 
 ## 快速开始
 
@@ -48,7 +71,7 @@ dsh-model setup
 `setup` 可以重复执行，它会依次：
 
 1. 下载并校验引擎，在本机 `127.0.0.1:8317/v1` 上运行，强制 key 鉴权；
-2. 接入 WorkBuddy（dsh-model 自己的 bridge）。安装时不问任何 key，OpenCode Zen 等要 key 的来源，需要时在管理页里打开；
+2. 启动 dsh-model 的守护进程（管理页、签到、登录都靠它）。安装时不问任何 key：免费来源推荐**硅基流动**（国内直连）或 **Gemini**（能上外网），装好后在管理页里填 key。WorkBuddy 只用来签到，模型默认不接进 dsh；
 3. 把管理页装进 dsh，把模型同步进 dsh；
 4. 在 VPS 上，额外开通对外地址 `https://<dsh 的域名>:9443/v1`（没有域名时用 `<IP>.sslip.io`）。
 
@@ -56,12 +79,12 @@ dsh-model setup
 
 | 区块 | 内容 |
 |---|---|
+| **每日签到** | WorkBuddy 国内版、Qoder、Trae 每天自动签到（也可一键签到），卡片显示今天领了多少 |
 | **访问地址** | 本机地址和外网地址，各带「复制」按钮 |
-| **来源** | 每家一个开关：打开就接入（没登录会弹出登录框），关闭只停用、保留登录。每行标出「免费 / 免费·额度少 / 需付费」，下面显示套餐用量和重置倒计时，或剩余积分和最近到期的一批；额度用完会标出多久后恢复 |
+| **来源** | 每家一个开关：打开就接入（没登录会弹出登录框），关闭只停用、保留登录。每行标出「免费 / 免费·额度少 / 需付费」；没登录的、关掉的收进折叠列表。下面显示套餐用量和重置倒计时，或剩余积分和最近到期的一批；额度用完会标出多久后恢复 |
 | **选模型** | 每个来源选哪些模型显示在 dsh 里，默认每家推荐 5 个，其余的经统一地址照样可用；每个模型旁边有「测一下」 |
 | **组合切换** | 把几个模型按顺序排成一个组合，在 dsh 里就是一个模型：先用第一个，限流、额度用完或出错就换下一个。其他软件用模型名 `combo/<id>` |
 | **自动切换** | 同一个模型在几个来源都有时，dsh 里多一组「自动切换（多来源）」，一个来源限流或出错就自动换另一个 |
-| **每日签到** | WorkBuddy 国内版、Qoder、Trae 每天自动签到（也可一键签到），卡片显示今天领了多少 |
 | **最近失败** | 最近 24 小时失败的请求：哪个模型、为什么、该怎么办（不记录对话内容） |
 | **API key** | OpenAI 格式，每台设备一把。显示 24 小时内的请求数、成功率、延迟和速度，可复制、轮换、吊销 |
 | **模型统计** | 按模型显示成功率、延迟和速度 |

package/dist/daemon/core.js

@@ -30,7 +30,7 @@ import { Mgmt } from '../engine/mgmt.js';
 import { DshModelError, isDshModelError } from '../errors.js';
 import { L } from '../i18n.js';
 import { refreshOpencodeModels, removeOpencodeKey, saveOpencodeKey, validateKeyShape, verifyOpencodeKey } from '../integrations/opencode.js';
-import { riskNotice as workbuddyRiskNotice } from '../integrations/workbuddy.js';
+import { blockedNotice as workbuddyBlockedNotice, riskNotice as workbuddyRiskNotice } from '../integrations/workbuddy.js';
 import { DSH_KEY_NAME, addKey, findActive, loadKeys, revokeKey, rotateKey, saveKeys } from '../keys.js';
 import { applyKeyChange, dshCandidates, dshKey, loadAll, saveAll, syncAll } from '../ops.js';
 import { defaultPick, pickedFor } from '../dsh/pick.js';
@@ -301,6 +301,9 @@ export class Daemon {
                 return { enabled: false, riskNotice: upstreamRiskNotice(getUpstream(def.id)) };
             return { enabled: false, login: await this.startLogin(def) };
         }
+        // WorkBuddy 会拦 dsh 的对话：从关着打开时先提醒一次
+        if (def.kind === 'workbuddy' && !state.enabled && !opts.acceptRisk)
+            return { enabled: false, riskNotice: workbuddyBlockedNotice() };
         if (def.kind === 'engine') {
             const m = await this.mgmt();
             for (const c of credsFor(await m.credentials(), def))

package/dist/integrations/workbuddy.js

@@ -37,7 +37,31 @@ async function waitCatalogs(ctx, since, expected, timeoutMs = 45_000) {
     return cats;
 }
 /** 检测 App → 首次打印说明 → bridge 配置与服务 → 等目录。返回目录（调用方随后 syncAll） */
+export const WORKBUDDY_SOURCES = ['workbuddy', 'workbuddy-ai'];
+/** WorkBuddy 会拦 dsh 发出的对话时的提醒（新装默认关、从页面打开前都显示） */
+export function blockedNotice() {
+    return L('WorkBuddy 现在会认出 dsh 发出的对话并拦下（「请求被拦截」），在 dsh 里基本用不了；「测一下」能通也不代表能用。dsh-model 不会去绕开这个限制，继续调用有封号风险。签到不受影响。仍要把它的模型接进 dsh 吗？', 'WorkBuddy now recognises conversations sent by dsh and blocks them ("请求被拦截"), so it barely works in dsh; a passing Test does not mean it works. dsh-model will not try to get around this, and continuing risks the account. Check-in is not affected. Connect its models to dsh anyway?');
+}
+/**
+ * 只签到（设计 §14.18）：第一次装时把 WorkBuddy 的模型关掉，签到照常（签到不看来源开关）。
+ * 以前装过的（已有 riskNoticeAt）保持原样；只做一次，之后用户自己打开的不改回去。
+ */
+export function applyWorkbuddyDefault(all) {
+    if (all.config.workbuddyDefault)
+        return;
+    if (all.config.bridge?.riskNoticeAt) {
+        all.config.workbuddyDefault = 'kept';
+        return;
+    }
+    const off = new Set(all.config.disabledSources ?? []);
+    for (const id of WORKBUDDY_SOURCES)
+        off.add(id);
+    all.config.disabledSources = [...off];
+    all.config.workbuddyDefault = 'checkin-only';
+    info(L('WorkBuddy 只用来签到：它现在会拦 dsh 的对话，所以模型默认不接进 dsh（要用可在「来源」里打开）', 'WorkBuddy is used for check-in only: it now blocks conversations from dsh, so its models are not connected by default (turn it on under Sources if you want)'));
+}
 export async function enableWorkbuddy(ctx, all) {
+    applyWorkbuddyDefault(all);
     const variants = await availableVariants(ctx.paths.home);
     if (!variants.length) {
         skip(ctx.platform === 'darwin'

package/package.json

@@ -1,8 +1,10 @@
 {
   "name": "dsh-model",
-  "version": "0.22.3",
-  "description": "One OpenAI-compatible address for all your models: subscriptions (Codex, Kimi…), free tiers (WorkBuddy, Gemini, Groq, OpenRouter), local models (Ollama, LM Studio), daily check-in credits, auto failover. Works with DeepSeek Harness (dsh) and any OpenAI-compatible tool. 一个地址用上你所有的模型。",
+  "version": "0.23.2",
+  "description": "Daily check-in credits claimed for you (Trae, Qoder, WorkBuddy China), one OpenAI-compatible address that relays every model, and free or subscription model sources (SiliconFlow, Gemini, Codex, Kimi, Ollama…). For DeepSeek Harness (dsh) and any OpenAI-compatible tool. 每天自动签到领积分 · 一个地址中转所有模型 · 免费和订阅模型随你接。",
   "keywords": [
+    "siliconflow",
+    "checkin",
     "dsh-plugin",
     "dsh-plugins",
     "dsh",

package/plugin/client.js

@@ -462,8 +462,9 @@ window.__ModuleLoader__.load({
         try { await api('POST', `/sources/${s.id}/logout`); await reload() } catch (e) { setErr(e.message) }
       }
       // 已登录的照常列出；没登录的收进一个可折叠的列表（默认收起；一个都没登录时默认展开）
-      const signedIn = visible.filter((s) => s.loggedIn)
-      const rest = visible.filter((s) => !s.loggedIn)
+      // 顶部只放正在用的（已登录且打开）；没登录的、关掉的都收进折叠列表
+      const signedIn = visible.filter((s) => s.loggedIn && s.enabled)
+      const rest = visible.filter((s) => !(s.loggedIn && s.enabled))
       const [restPref, setRestPref] = useState(readShowRest)
       const openRest = restPref ?? signedIn.length === 0
       const toggleRest = (v) => {
@@ -509,7 +510,7 @@ window.__ModuleLoader__.load({
         showRisky ? h('p', { style: { ...S.note, color: C.warn } }, L('高风险来源（Claude、Antigravity）：服务商有封禁第三方使用订阅的先例，账号可能被封。仅在你清楚风险时使用。', 'High-risk sources (Claude, Antigravity): the providers have banned third-party use of subscriptions before; your account may be suspended. Use only if you accept the risk.')) : null,
         signedIn.length ? h('div', { style: S.card }, signedIn.map(row)) : h('div', { style: { ...S.note, margin: '0 0 8px' } }, L('还没有接入任何来源：展开下面的列表挑一个。', 'No source connected yet: open the list below and pick one.')),
         rest.length ? h('button', { type: 'button', style: { ...S.btn, marginTop: 10 }, 'aria-expanded': openRest, onClick: () => toggleRest(!openRest) },
-          openRest ? L(`▾ 收起未登录的来源（${rest.length} 个）`, `▾ Hide signed-out sources (${rest.length})`) : L(`▸ 未登录的来源（${rest.length} 个）`, `▸ Signed-out sources (${rest.length})`)) : null,
+          openRest ? L(`▾ 收起没在用的来源（${rest.length} 个：未登录或已关闭）`, `▾ Hide sources not in use (${rest.length}: signed out or off)`) : L(`▸ 没在用的来源（${rest.length} 个：未登录或已关闭）`, `▸ Sources not in use (${rest.length}: signed out or off)`)) : null,
         openRest && rest.length ? h('div', { style: { ...S.card, marginTop: 8 } }, rest.map(row)) : null)
     }
 
@@ -1089,9 +1090,15 @@ window.__ModuleLoader__.load({
     /** 刚装好、接入的来源少于两个时显示：先接一个免费来源 → 模型已经挑好 → 去 dsh 里用 */
     function Guide({ sources, onConnect }) {
       const [done, setDone] = useState(readGuideDone)
-      const connected = sources.filter((s) => s.loggedIn && s.enabled)
-      if (done || connected.length >= 2) return null
-      const free = sources.filter((s) => !s.risky && !(s.loggedIn && s.enabled) && s.pricing && s.pricing.tier !== 'paid')
+      // WorkBuddy 只签到（它拦 dsh 的对话），不算"已接入"
+      const connected = sources.filter((s) => s.loggedIn && s.enabled && s.kind !== 'workbuddy')
+      if (done || connected.length >= 1) return null
+      // 只推两家：国内直连的硅基流动、能上外网的 Gemini
+      const PICKS = [
+        { id: 'siliconflow', why: () => L('国内直连，不用代理；实名认证后部分模型免费', 'Reachable from China without a proxy; some models are free after identity verification') },
+        { id: 'gemini', why: () => L('能上外网就选它：免费模型里质量最好，每天有次数上限', 'If you can reach Google: the best free models, with a daily limit') },
+      ]
+      const picks = PICKS.map((p) => ({ ...p, s: sources.find((s) => s.id === p.id) })).filter((p) => p.s)
       const hide = () => {
         setDone(true)
         try { window.localStorage.setItem(GUIDE_KEY, '1') } catch { /* 无痕模式：只在本次生效 */ }
@@ -1106,9 +1113,11 @@ window.__ModuleLoader__.load({
             h('button', { type: 'button', style: S.btn, onClick: hide }, L('我知道了，不再显示', 'Got it, hide this'))),
           step(1, L('接入一个免费来源', 'Connect a free source'),
             h('div', null,
-              h('div', { style: S.meta }, L('先接一个就能用。标「免费·额度少」的要先去对方网站拿一把免费 key。', 'One is enough to start. "Free · limited" ones need a free key from their website first.')),
-              h('div', { style: { display: 'flex', flexWrap: 'wrap', gap: 6, marginTop: 6 } },
-                free.slice(0, 8).map((s) => h('button', { key: s.id, type: 'button', style: S.btn, onClick: () => onConnect(s) }, `${s.label}${s.pricing.tier === 'free' ? L(' · 免费', ' · free') : ''}`))))),
+              h('div', { style: S.meta }, L('接一个就能用：去对方网站领一把免费 key，点下面的按钮粘贴进来。', 'One is enough: get a free key from their website, then click below and paste it.')),
+              picks.map((p) => h('div', { key: p.id, style: { ...S.line, marginTop: 6, flexWrap: 'wrap' } },
+                h('button', { type: 'button', style: S.btn, onClick: () => onConnect(p.s) }, p.s.label),
+                h('span', { style: { ...S.meta, ...S.grow } }, p.why()),
+                p.s.subscribeUrl ? h('a', { href: p.s.subscribeUrl, target: '_blank', rel: 'noopener noreferrer', style: { color: C.accent, fontSize: 12 } }, L('去领 key ↗', 'Get a key ↗')) : null)))),
           step(2, L('模型已经替你挑好', 'Models are picked for you'),
             h('div', { style: S.meta }, L('每个来源默认挑 5 个放进 dsh（免费的优先）。想换就在下面「来源」里点「选模型」。', 'Each source puts 5 suggested models into dsh (free ones first). Change them with "Models" under Sources below.'))),
           step(3, L('去用', 'Use it'),