dsh-provider-hub 0.1.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 DDDMUC
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,81 @@
+# dsh-provider-hub
+
+[English](#english) | 中文
+
+为 DSH 扩大可用的服务商：**选一个服务商，粘贴 API Key，点启用**——插件把官方 `llm-pi-ai` 路由与凭据一次性写好，立即生效、无需重启、无需手改 `settings.yaml`。
+
+内置 20 家预设（OpenAI 兼容协议 + Anthropic Messages 协议），另有「自定义服务商」入口可接入任意 OpenAI 兼容端点。
+
+## 安装
+
+```sh
+# 通过插件管理器（推荐）
+dsh plugin --profile web add dsh-provider-hub
+
+# 或本地开发挂载
+dsh plugin --profile web add link:/path/to/dsh-provider-hub
+```
+
+重启 `dsh web`，打开 **设置 → Models**，页面底部出现「服务商预设」卡片。
+
+## 使用
+
+1. 点预设行右侧的「获取 Key」去对应控制台申请 Key（模型列表链接在其旁边）
+2. 把 Key 粘贴进输入框，点「启用」
+3. 状态徽标会显示 `运行中`（宿主已注册该路由）/ `已写入` / `缺 Key`
+4. 回到聊天输入框的模型选择器，新服务商的模型已经可选
+
+移除：点行内「移除」。**已保存的 Key 会保留在凭据服务里**，重新启用时不需要再粘贴；如需彻底清除，在官方 Models 页删除该路由后用凭据接口清除对应引用。
+
+## 自定义服务商
+
+「自定义服务商」支持任何 OpenAI 兼容端点（`openai-completions`）或 Anthropic Messages 端点（`anthropic-messages`）：填路由 ID、Base URL（到 `/v1` 为止）、协议、模型 ID（逗号/换行分隔）与 Key 即可。
+
+## 工作原理
+
+写入全部走 DSH 官方服务，插件不直接碰配置文件：
+
+- `credentials.set(apiKeyEnv, key)` —— Key 存进凭据服务（与手写 profile 用的是同一个引用）
+- `settings.mutate('llm-pi-ai', [set providers.<route>], revision)` —— 写入官方 pi-ai 命名空间的用户层，官方 schema 校验、适配器热注册
+
+`core/` 目录是纯 JS（无 Node 内建、无 DSH 依赖、无 I/O）：服务商目录、profile 构建与校验。宿主半、浏览器半（本插件）、以及后续的桌面端外壳共用同一份核心代码。
+
+## 预设清单
+
+OpenAI、OpenRouter、Anthropic、Google Gemini（OpenAI 兼容）、xAI Grok、Mistral、Groq、Together、Fireworks、DeepInfra、Novita、硅基流动、月之暗面 Kimi（国内/国际）、智谱 GLM（国内）、Z.ai GLM（国际）、MiniMax、阿里云百炼 Qwen、百度千帆、腾讯混元、火山方舟（豆包）。内置 DeepSeek 官方与 StepFun 已有官方适配器，本插件不重复提供。
+
+## 注意
+
+- 预设里的模型规格（上下文/输出/档位）是**可编辑的起手默认值**；模型 ID 与限额以各服务商文档为准，可在官方 Models 页增删改
+- 声明了 `reasoningEfforts` 的模型才会有思考档位选择器；未声明的模型按无档位处理
+- 移除只下线路由，不删除 Key（见上）
+
+---
+
+## English
+
+Expand the providers DSH can use: **pick a provider, paste the API key, press Enable**. The plugin writes the official `llm-pi-ai` route and its credential in one step - effective immediately, no restart, no hand-editing `settings.yaml`.
+
+20 built-in presets (OpenAI-compatible and Anthropic Messages protocols) plus a custom-endpoint form for any OpenAI-compatible gateway.
+
+### Install
+
+```sh
+dsh plugin --profile web add dsh-provider-hub
+```
+
+Restart `dsh web`, open **Settings → Models**; the "Provider presets" card is at the bottom.
+
+### Usage
+
+Paste the key (each row links to the console and the model list), press **Enable**, then pick the provider's models in the composer's model selector. **Remove** takes the route offline but keeps the stored key, so re-enabling never needs the key again.
+
+### How it works
+
+All writes go through official services - the plugin never touches config files directly: `credentials.set(apiKeyEnv, key)` stores the key, and `settings.mutate('llm-pi-ai', [set providers.<route>], revision)` writes the user layer of the official pi-ai namespace (schema-validated, hot-registered).
+
+`core/` is pure JS (no Node built-ins, no DSH imports, no I/O): the provider catalog plus profile building and validation, shared by the host half, this plugin's browser half, and any future desktop shell.
+
+### Notes
+
+Preset model specs are editable starting defaults; model ids and limits follow each vendor's docs. Only models declaring `reasoningEfforts` get the thinking-level selector. Removing a route keeps its credential.

package/cordis.patch.yml

@@ -0,0 +1,13 @@
+# dsh-provider-hub bundle patch: inserts the plugin row into the profile roster.
+#
+# The node half serves the loopback-only /dsh-provider-hub/* routes; the
+# browser half (exports "./client") mounts the presets card on the official
+# Models settings page footer.
+#
+# The entry name is the BARE package name on purpose: the client-module scanner
+# treats it as a package specifier and resolves `<name>/package.json`, so a
+# subpath entry would be cached as "not a client package" and the browser half
+# would silently never load.
+- insert:
+    - id: dsh-provider-hub
+      name: 'dsh-provider-hub'

package/core/coverage.js

@@ -0,0 +1,31 @@
+// dsh-provider-hub - core per-app coverage.
+//
+// The catalog is app-agnostic: every shell shows the same presets minus the
+// ones its target application already covers natively. Nothing is deleted -
+// covered presets stay in the catalog and stay reachable in the UI behind a
+// reveal toggle, because a native entry can still be missing for an account
+// (see StepFun's Step Plan endpoint) or carry a different offering.
+//
+// Coverage is matched on the preset id plus optional vendor aliases
+// (`nativeIds`), so a preset built for one wire endpoint still hides behind
+// the native provider that serves the same vendor (gemini -> google,
+// moonshot -> moonshotai-cn).
+
+/**
+ * Partition presets against one app's native provider ids.
+ * @param presets - catalog entries (`id`, optional `nativeIds` aliases).
+ * @param nativeIds - provider ids the target app already offers.
+ * @returns `{ visible, covered }`; covered entries carry the id that matched.
+ */
+export function coverageOf(presets, nativeIds) {
+  const native = new Set(Array.isArray(nativeIds) ? nativeIds : [])
+  const visible = []
+  const covered = []
+  for (const preset of presets) {
+    const aliases = Array.isArray(preset.nativeIds) ? preset.nativeIds : []
+    const matched = [preset.id, ...aliases].find((id) => native.has(id))
+    if (matched === undefined) visible.push(preset.id)
+    else covered.push({ id: preset.id, as: matched })
+  }
+  return { visible, covered }
+}

package/core/profile.js

@@ -0,0 +1,144 @@
+// dsh-provider-hub - core profile building and validation.
+//
+// Pure functions over plain data: turn one preset (or one custom form) into
+// the exact `llm-pi-ai` provider profile the official adapter accepts, and
+// refuse drafts the adapter would refuse later. The rules mirror the adapter's
+// own resolution: a route key is a dict key, an endpoint is http(s) root
+// without credentials/query/fragment, a model id is a non-empty string, and a
+// declared reasoning level other than `off` needs a non-empty wire value.
+
+import { OPENAI_COMPAT, SUPPORTED_APIS, customEnvName, findPreset } from './providers.js'
+
+/** Route keys the official loader and this UI agree on. */
+export const ROUTE_RE = /^[a-z0-9][a-z0-9-]{0,62}$/
+
+/** Whether a string is a usable route key. */
+export function isRouteKey(value) {
+  return typeof value === 'string' && ROUTE_RE.test(value)
+}
+
+/**
+ * Validate one http(s) base URL the way the adapter does.
+ * @param value - candidate endpoint root.
+ * @returns an error message, or undefined when acceptable.
+ */
+export function baseUrlError(value) {
+  if (typeof value !== 'string' || value.trim() === '') return 'base URL 不能为空'
+  let url
+  try {
+    url = new URL(value.trim())
+  } catch {
+    return 'base URL 不是合法 URL'
+  }
+  if (url.protocol !== 'https:' && url.protocol !== 'http:') return 'base URL 必须是 http(s)'
+  if (url.username || url.password) return 'base URL 不能带账号密码'
+  if (url.search || url.hash) return 'base URL 不能带查询参数或片段'
+  return undefined
+}
+
+/** Validate a reasoning-efforts dict the way the adapter resolves it. */
+export function effortsError(efforts) {
+  if (efforts === undefined || efforts === false) return undefined
+  if (typeof efforts !== 'object' || efforts === null) return 'reasoningEfforts 必须是对象或 false'
+  const entries = Object.entries(efforts)
+  if (!entries.some(([level]) => level !== 'off')) return 'reasoningEfforts 至少要声明一个 off 以外的档位'
+  for (const [level, wire] of entries) {
+    if (wire === null) {
+      if (level !== 'off') return `档位 ${level} 需要发送值`
+      continue
+    }
+    if (typeof wire !== 'string' || wire.length === 0) return `档位 ${level} 的发送值不能为空`
+  }
+  return undefined
+}
+
+/** Strip undefined fields so the persisted profile stays clean. */
+function compact(object) {
+  return Object.fromEntries(Object.entries(object).filter(([, value]) => value !== undefined))
+}
+
+/** Sanitize one model entry from a preset or the custom form. */
+export function sanitizeModel(entry) {
+  if (!entry || typeof entry.id !== 'string' || entry.id.trim() === '') throw new Error('模型 ID 不能为空')
+  const model = compact({
+    id: entry.id.trim(),
+    name: typeof entry.name === 'string' && entry.name.trim() !== '' ? entry.name.trim() : undefined,
+    contextWindow: Number.isFinite(entry.contextWindow) && entry.contextWindow > 0 ? Math.floor(entry.contextWindow) : undefined,
+    maxTokens: Number.isFinite(entry.maxTokens) && entry.maxTokens > 0 ? Math.floor(entry.maxTokens) : undefined,
+    input: Array.isArray(entry.input) && entry.input.length > 0 ? [...entry.input] : undefined,
+    reasoningEfforts: entry.reasoningEfforts,
+  })
+  return model
+}
+
+/**
+ * The dict key one request targets: a known preset id, or a validated custom route.
+ * @param input - `{ presetId }` or `{ route }`.
+ * @returns the `llm-pi-ai.providers.<route>` key.
+ * @throws {Error} when the route cannot be a dict key.
+ */
+export function routeOf(input) {
+  if (input && input.presetId !== undefined) {
+    const preset = findPreset(input.presetId)
+    if (!preset) throw new Error(`未知的预设：${input.presetId}`)
+    return preset.id
+  }
+  const route = String(input?.route ?? '').trim()
+  if (!isRouteKey(route)) throw new Error('路由 ID 只能用小写字母、数字和连字符，且以字母或数字开头')
+  return route
+}
+
+/**
+ * Build the provider profile for one request shape.
+ * @param input - `{ presetId }` or `{ route, displayName?, baseURL, api, modelIds }`.
+ * @returns the `llm-pi-ai.providers.<route>` value.
+ * @throws {Error} when the draft cannot be served.
+ */
+export function buildProfile(input) {
+  const preset = input && input.presetId !== undefined ? findPreset(input.presetId) : undefined
+  if (input && input.presetId !== undefined && !preset) throw new Error(`未知的预设：${input.presetId}`)
+
+  const route = routeOf(input)
+  const baseURL = String((preset ? preset.baseURL : input.baseURL) ?? '').trim()
+  const urlError = baseUrlError(baseURL)
+  if (urlError) throw new Error(urlError)
+  const api = preset ? preset.api : input.api
+  if (!SUPPORTED_APIS.includes(api)) throw new Error(`协议必须是 ${SUPPORTED_APIS.join(' 或 ')}`)
+
+  let models
+  if (preset) {
+    models = preset.models.map((entry) => sanitizeModel(entry))
+  } else {
+    const ids = String(input.modelIds ?? '')
+      .split(/[\n,]/)
+      .map((part) => part.trim())
+      .filter((part) => part !== '')
+    if (ids.length === 0) throw new Error('至少填一个模型 ID')
+    const seen = new Set()
+    models = ids.map((id) => {
+      if (seen.has(id)) throw new Error(`模型 ID 重复：${id}`)
+      seen.add(id)
+      return sanitizeModel({ id, name: id, contextWindow: 131072, maxTokens: 32768, input: ['text'] })
+    })
+  }
+  for (const model of models) {
+    const error = effortsError(model.reasoningEfforts)
+    if (error) throw new Error(`${model.id}: ${error}`)
+  }
+
+  const displayName = preset
+    ? preset.name
+    : typeof input.displayName === 'string' && input.displayName.trim() !== ''
+      ? input.displayName.trim()
+      : route
+  const env = preset ? preset.env : customEnvName(route)
+
+  return compact({
+    displayName,
+    api,
+    baseURL,
+    apiKeyEnv: env,
+    compat: api === 'openai-completions' ? { ...OPENAI_COMPAT, ...(input.compat ?? {}) } : undefined,
+    models,
+  })
+}

package/core/providers.js

@@ -0,0 +1,342 @@
+// dsh-provider-hub - core provider catalog.
+//
+// Pure data + pure helpers. No Node built-ins, no DSH imports, no I/O: the
+// host half, the browser half and (later) a desktop shell all consume this
+// same module, so a preset fixed here is fixed everywhere.
+//
+// Every model entry is a STARTING DEFAULT: contextWindow / maxTokens / input /
+// reasoningEfforts are editable afterwards in the official Models settings
+// page (or by the dsh-client-ui-model-capabilities panel when it is mounted).
+// Model ids drift faster than this file does; the docs link beside each preset
+// is the source of truth.
+
+/** Route-level OpenAI-completions switches shared by gateway presets. */
+export const OPENAI_COMPAT = {
+  supportsStore: false,
+  supportsDeveloperRole: false,
+  supportsReasoningEffort: true,
+  supportsUsageInStreaming: true,
+  supportsFinishReason: true,
+  maxTokensField: 'max_tokens',
+  requiresToolResultName: false,
+  requiresAssistantAfterToolResult: false,
+  requiresThinkingAsText: false,
+  requiresReasoningContentOnAssistantMessages: false,
+}
+
+/** Wire protocols a preset may name (the pi-ai adapter's table). */
+export const SUPPORTED_APIS = ['openai-completions', 'anthropic-messages']
+
+const text = ['text']
+const textImage = ['text', 'image']
+
+/** Reasoning levels as `{ level: wireValue }`; wire values are what is sent. */
+const efforts = (...levels) => Object.fromEntries(levels.map((level) => [level, level]))
+const LOW_HIGH = efforts('low', 'high')
+const LOW_MEDIUM_HIGH = efforts('low', 'medium', 'high')
+
+/**
+ * The curated catalog. `env` is the credential reference the profile writes as
+ * `apiKeyEnv`; the key itself is stored by the consumer's secret store (DSH's
+ * credentials service for the plugin, the OS keychain for a desktop shell).
+ */
+export const PRESETS = [
+  {
+    id: 'openai',
+    name: 'OpenAI',
+    docs: 'https://platform.openai.com/docs/models',
+    keyUrl: 'https://platform.openai.com/api-keys',
+    env: 'OPENAI_API_KEY',
+    baseURL: 'https://api.openai.com/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'gpt-5.1', name: 'GPT-5.1', contextWindow: 400000, maxTokens: 128000, input: textImage, reasoningEfforts: LOW_MEDIUM_HIGH },
+      { id: 'gpt-5.1-mini', name: 'GPT-5.1 mini', contextWindow: 400000, maxTokens: 128000, input: textImage, reasoningEfforts: LOW_MEDIUM_HIGH },
+      { id: 'gpt-4.1', name: 'GPT-4.1', contextWindow: 1047576, maxTokens: 32768, input: textImage },
+    ],
+  },
+  {
+    id: 'openrouter',
+    name: 'OpenRouter',
+    docs: 'https://openrouter.ai/models',
+    keyUrl: 'https://openrouter.ai/keys',
+    env: 'OPENROUTER_API_KEY',
+    baseURL: 'https://openrouter.ai/api/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'anthropic/claude-sonnet-4.5', name: 'Claude Sonnet 4.5', contextWindow: 200000, maxTokens: 64000, input: textImage, reasoningEfforts: LOW_MEDIUM_HIGH },
+      { id: 'deepseek/deepseek-chat', name: 'DeepSeek Chat', contextWindow: 163840, maxTokens: 65536, input: text },
+      { id: 'qwen/qwen3-max', name: 'Qwen3 Max', contextWindow: 262144, maxTokens: 32768, input: text },
+      { id: 'google/gemini-2.5-pro', name: 'Gemini 2.5 Pro', contextWindow: 1048576, maxTokens: 65536, input: textImage },
+    ],
+  },
+  {
+    id: 'anthropic',
+    name: 'Anthropic (Messages API)',
+    docs: 'https://docs.anthropic.com/en/docs/about-claude/models',
+    keyUrl: 'https://console.anthropic.com/settings/keys',
+    env: 'ANTHROPIC_API_KEY',
+    baseURL: 'https://api.anthropic.com',
+    api: 'anthropic-messages',
+    models: [
+      { id: 'claude-sonnet-4-5', name: 'Claude Sonnet 4.5', contextWindow: 200000, maxTokens: 64000, input: textImage },
+      { id: 'claude-opus-4-1', name: 'Claude Opus 4.1', contextWindow: 200000, maxTokens: 32000, input: textImage },
+      { id: 'claude-haiku-4-5', name: 'Claude Haiku 4.5', contextWindow: 200000, maxTokens: 64000, input: textImage },
+    ],
+  },
+  {
+    id: 'gemini',
+    nativeIds: ['google'],
+    name: 'Google Gemini (OpenAI 兼容)',
+    docs: 'https://ai.google.dev/gemini-api/docs/openai',
+    keyUrl: 'https://aistudio.google.com/apikey',
+    env: 'GEMINI_API_KEY',
+    baseURL: 'https://generativelanguage.googleapis.com/v1beta/openai',
+    api: 'openai-completions',
+    models: [
+      { id: 'gemini-2.5-pro', name: 'Gemini 2.5 Pro', contextWindow: 1048576, maxTokens: 65536, input: textImage },
+      { id: 'gemini-2.5-flash', name: 'Gemini 2.5 Flash', contextWindow: 1048576, maxTokens: 65536, input: textImage },
+    ],
+  },
+  {
+    id: 'xai',
+    name: 'xAI Grok',
+    docs: 'https://docs.x.ai/docs/models',
+    keyUrl: 'https://console.x.ai',
+    env: 'XAI_API_KEY',
+    baseURL: 'https://api.x.ai/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'grok-4', name: 'Grok 4', contextWindow: 262144, maxTokens: 32768, input: textImage, reasoningEfforts: LOW_HIGH },
+      { id: 'grok-4-fast', name: 'Grok 4 Fast', contextWindow: 2097152, maxTokens: 32768, input: textImage, reasoningEfforts: LOW_HIGH },
+      { id: 'grok-code-fast-1', name: 'Grok Code Fast 1', contextWindow: 262144, maxTokens: 32768, input: text },
+    ],
+  },
+  {
+    id: 'mistral',
+    name: 'Mistral',
+    docs: 'https://docs.mistral.ai/getting-started/models/models_overview/',
+    keyUrl: 'https://console.mistral.ai/api-keys',
+    env: 'MISTRAL_API_KEY',
+    baseURL: 'https://api.mistral.ai/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'mistral-large-latest', name: 'Mistral Large', contextWindow: 131072, maxTokens: 32768, input: textImage },
+      { id: 'mistral-medium-latest', name: 'Mistral Medium', contextWindow: 131072, maxTokens: 32768, input: textImage },
+      { id: 'devstral-medium-latest', name: 'Devstral Medium', contextWindow: 131072, maxTokens: 32768, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+    ],
+  },
+  {
+    id: 'groq',
+    name: 'Groq',
+    docs: 'https://console.groq.com/docs/models',
+    keyUrl: 'https://console.groq.com/keys',
+    env: 'GROQ_API_KEY',
+    baseURL: 'https://api.groq.com/openai/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'llama-3.3-70b-versatile', name: 'Llama 3.3 70B', contextWindow: 131072, maxTokens: 32768, input: text },
+      { id: 'openai/gpt-oss-120b', name: 'GPT-OSS 120B', contextWindow: 131072, maxTokens: 32768, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+    ],
+  },
+  {
+    id: 'together',
+    name: 'Together AI',
+    docs: 'https://docs.together.ai/docs/serverless-models',
+    keyUrl: 'https://api.together.ai/settings/api-keys',
+    env: 'TOGETHER_API_KEY',
+    baseURL: 'https://api.together.xyz/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'deepseek-ai/DeepSeek-V3.1', name: 'DeepSeek V3.1', contextWindow: 131072, maxTokens: 32768, input: text },
+      { id: 'Qwen/Qwen3-Coder-480B-A35B-Instruct', name: 'Qwen3 Coder 480B', contextWindow: 262144, maxTokens: 32768, input: text },
+    ],
+  },
+  {
+    id: 'fireworks',
+    name: 'Fireworks AI',
+    docs: 'https://fireworks.ai/models',
+    keyUrl: 'https://fireworks.ai/account/api-keys',
+    env: 'FIREWORKS_API_KEY',
+    baseURL: 'https://api.fireworks.ai/inference/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'accounts/fireworks/models/deepseek-v3p1', name: 'DeepSeek V3.1', contextWindow: 163840, maxTokens: 32768, input: text },
+      { id: 'accounts/fireworks/models/qwen3-coder-480b-a35b-instruct', name: 'Qwen3 Coder 480B', contextWindow: 262144, maxTokens: 16384, input: text },
+    ],
+  },
+  {
+    id: 'deepinfra',
+    name: 'DeepInfra',
+    docs: 'https://deepinfra.com/models',
+    keyUrl: 'https://deepinfra.com/dash/api_keys',
+    env: 'DEEPINFRA_API_KEY',
+    baseURL: 'https://api.deepinfra.com/v1/openai',
+    api: 'openai-completions',
+    models: [
+      { id: 'deepseek-ai/DeepSeek-V3.1', name: 'DeepSeek V3.1', contextWindow: 163840, maxTokens: 32768, input: text },
+      { id: 'meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8', name: 'Llama 4 Maverick', contextWindow: 1048576, maxTokens: 32768, input: textImage },
+    ],
+  },
+  {
+    id: 'novita',
+    name: 'Novita AI',
+    docs: 'https://novita.ai/model-api/product/llm-api',
+    keyUrl: 'https://novita.ai/settings/key-management',
+    env: 'NOVITA_API_KEY',
+    baseURL: 'https://api.novita.ai/v3/openai',
+    api: 'openai-completions',
+    models: [
+      { id: 'deepseek/deepseek-v3.1', name: 'DeepSeek V3.1', contextWindow: 131072, maxTokens: 32768, input: text },
+      { id: 'qwen/qwen3-coder-480b-a35b-instruct', name: 'Qwen3 Coder 480B', contextWindow: 262144, maxTokens: 32768, input: text },
+    ],
+  },
+  {
+    id: 'siliconflow',
+    name: '硅基流动 SiliconFlow',
+    docs: 'https://docs.siliconflow.cn/cn/userguide/introduction',
+    keyUrl: 'https://cloud.siliconflow.cn/account/ak',
+    env: 'SILICONFLOW_API_KEY',
+    baseURL: 'https://api.siliconflow.cn/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'deepseek-ai/DeepSeek-V3.2-Exp', name: 'DeepSeek V3.2', contextWindow: 163840, maxTokens: 65536, input: text, reasoningEfforts: LOW_HIGH },
+      { id: 'Qwen/Qwen3-235B-A22B-Thinking-2507', name: 'Qwen3 235B Thinking', contextWindow: 262144, maxTokens: 65536, input: text },
+      { id: 'zai-org/GLM-4.6', name: 'GLM-4.6', contextWindow: 200000, maxTokens: 128000, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+    ],
+  },
+  {
+    id: 'moonshot',
+    nativeIds: ['moonshotai-cn'],
+    name: '月之暗面 Kimi（国内）',
+    docs: 'https://platform.moonshot.cn/docs',
+    keyUrl: 'https://platform.moonshot.cn/console/api-keys',
+    env: 'MOONSHOT_API_KEY',
+    baseURL: 'https://api.moonshot.cn/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'kimi-k2-thinking', name: 'Kimi K2 Thinking', contextWindow: 262144, maxTokens: 65536, input: text, reasoningEfforts: LOW_HIGH },
+      { id: 'kimi-latest', name: 'Kimi Latest', contextWindow: 262144, maxTokens: 32768, input: textImage },
+    ],
+  },
+  {
+    id: 'moonshot-global',
+    nativeIds: ['moonshotai'],
+    name: 'Moonshot Kimi（国际）',
+    docs: 'https://platform.moonshot.ai/docs',
+    keyUrl: 'https://platform.moonshot.ai/console/api-keys',
+    env: 'MOONSHOT_GLOBAL_API_KEY',
+    baseURL: 'https://api.moonshot.ai/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'kimi-k2-thinking', name: 'Kimi K2 Thinking', contextWindow: 262144, maxTokens: 65536, input: text, reasoningEfforts: LOW_HIGH },
+      { id: 'kimi-latest', name: 'Kimi Latest', contextWindow: 262144, maxTokens: 32768, input: textImage },
+    ],
+  },
+  {
+    id: 'zhipu',
+    name: '智谱 GLM（国内）',
+    docs: 'https://docs.bigmodel.cn/cn/guide/models/text/glm-4.6',
+    keyUrl: 'https://bigmodel.cn/usercenter/apikeys',
+    env: 'ZHIPU_API_KEY',
+    baseURL: 'https://open.bigmodel.cn/api/paas/v4',
+    api: 'openai-completions',
+    models: [
+      { id: 'glm-4.6', name: 'GLM-4.6', contextWindow: 200000, maxTokens: 128000, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+      { id: 'glm-4.5-air', name: 'GLM-4.5 Air', contextWindow: 131072, maxTokens: 98304, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+    ],
+  },
+  {
+    id: 'zai',
+    name: 'Z.ai GLM（国际）',
+    docs: 'https://docs.z.ai/guides/llm/glm-4.6',
+    keyUrl: 'https://z.ai/manage-apikey/apikey-list',
+    env: 'ZAI_API_KEY',
+    baseURL: 'https://api.z.ai/api/paas/v4',
+    api: 'openai-completions',
+    models: [
+      { id: 'glm-4.6', name: 'GLM-4.6', contextWindow: 200000, maxTokens: 128000, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+      { id: 'glm-4.5-air', name: 'GLM-4.5 Air', contextWindow: 131072, maxTokens: 98304, input: text, reasoningEfforts: LOW_MEDIUM_HIGH },
+    ],
+  },
+  {
+    id: 'minimax',
+    name: 'MiniMax',
+    docs: 'https://platform.minimaxi.com/document/guides/chat-model/V2',
+    keyUrl: 'https://platform.minimaxi.com/user-center/basic-information/interface-key',
+    env: 'MINIMAX_API_KEY',
+    baseURL: 'https://api.minimaxi.com/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'MiniMax-M2', name: 'MiniMax M2', contextWindow: 204800, maxTokens: 131072, input: text },
+      { id: 'MiniMax-Text-01', name: 'MiniMax Text 01', contextWindow: 1000192, maxTokens: 131072, input: text },
+    ],
+  },
+  {
+    id: 'dashscope',
+    name: '阿里云百炼 Qwen',
+    docs: 'https://help.aliyun.com/zh/model-studio/models',
+    keyUrl: 'https://bailian.console.aliyun.com/?apiKey=1',
+    env: 'DASHSCOPE_API_KEY',
+    baseURL: 'https://dashscope.aliyuncs.com/compatible-mode/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'qwen3-max', name: 'Qwen3 Max', contextWindow: 262144, maxTokens: 65536, input: text },
+      { id: 'qwen3-coder-plus', name: 'Qwen3 Coder Plus', contextWindow: 1048576, maxTokens: 65536, input: text },
+      { id: 'qwen-plus', name: 'Qwen Plus', contextWindow: 1000000, maxTokens: 32768, input: text },
+    ],
+  },
+  {
+    id: 'qianfan',
+    name: '百度千帆（文心）',
+    docs: 'https://cloud.baidu.com/doc/qianfan-api/s/3m7of64lb',
+    keyUrl: 'https://console.bce.baidu.com/iam/#/iam/apikey/list',
+    env: 'QIANFAN_API_KEY',
+    baseURL: 'https://qianfan.baidubce.com/v2',
+    api: 'openai-completions',
+    models: [
+      { id: 'ernie-5.0', name: '文心 5.0', contextWindow: 131072, maxTokens: 32768, input: text },
+      { id: 'ernie-x1.1', name: '文心 X1.1', contextWindow: 65536, maxTokens: 16384, input: text },
+    ],
+  },
+  {
+    id: 'hunyuan',
+    name: '腾讯混元',
+    docs: 'https://cloud.tencent.com/document/product/1729/111007',
+    keyUrl: 'https://console.cloud.tencent.com/hunyuan/api-key',
+    env: 'HUNYUAN_API_KEY',
+    baseURL: 'https://api.hunyuan.cloud.tencent.com/v1',
+    api: 'openai-completions',
+    models: [
+      { id: 'hunyuan-turbos-latest', name: '混元 TurboS', contextWindow: 131072, maxTokens: 16384, input: text },
+      { id: 'hunyuan-t1-latest', name: '混元 T1', contextWindow: 65536, maxTokens: 16384, input: text },
+    ],
+  },
+  {
+    id: 'ark',
+    name: '火山方舟（豆包）',
+    docs: 'https://www.volcengine.com/docs/82379/1330310',
+    keyUrl: 'https://console.volcengine.com/ark/region:ark+cn-beijing/apiKey',
+    env: 'ARK_API_KEY',
+    baseURL: 'https://ark.cn-beijing.volces.com/api/v3',
+    api: 'openai-completions',
+    models: [
+      { id: 'doubao-seed-1-6-250615', name: '豆包 Seed 1.6', contextWindow: 262144, maxTokens: 32768, input: textImage },
+      { id: 'doubao-seed-1-6-thinking-250715', name: '豆包 Seed 1.6 Thinking', contextWindow: 262144, maxTokens: 32768, input: textImage },
+    ],
+  },
+]
+
+/** Look one preset up by id. */
+export function findPreset(id) {
+  return PRESETS.find((preset) => preset.id === id)
+}
+
+/** The credential reference for a custom (preset-less) route. */
+export function customEnvName(route) {
+  const slug = String(route ?? '')
+    .toUpperCase()
+    .replace(/[^A-Z0-9]+/g, '_')
+    .replace(/^_+|_+$/g, '')
+  return `${slug || 'CUSTOM'}_API_KEY`
+}