feat(desktop): add LiteLLM Gateway provider preset

.changeset/litellm-gateway-preset.md

@@ -0,0 +1,7 @@
+---
+"@open-codesign/shared": patch
+"@open-codesign/i18n": patch
+"@open-codesign/desktop": patch
+---
+
+Add a LiteLLM Gateway preset to Settings → Add provider. It pre-fills an OpenAI-compatible endpoint at `http://localhost:4000/v1`, starts in keyless mode for proxies without a master key, and explains how to switch to a master/virtual key. LiteLLM is not bundled.

apps/desktop/src/renderer/src/components/AddCustomProviderModal.test.tsx

@@ -24,6 +24,27 @@ describe('AddCustomProviderModal', () => {
     expect(html).not.toMatch(/type="checkbox"[^>]*checked/);
   });
 
+  it('starts keyless with the preset hint when a preset opts out of API keys', () => {
+    const html = renderToStaticMarkup(
+      <AddCustomProviderModal
+        onSave={() => undefined}
+        onClose={() => undefined}
+        initialValues={{
+          name: 'LiteLLM Gateway',
+          baseUrl: 'http://localhost:4000/v1',
+          wire: 'openai-chat',
+          requiresApiKey: false,
+          hint: 'litellm-hint',
+        }}
+      />,
+    );
+
+    expect(html).toContain('litellm-hint');
+    expect(html).toContain('value="http://localhost:4000/v1"');
+    expect(html).toMatch(/type="checkbox"[^>]*checked/);
+    expect(html).toMatch(/type="password"[^>]*disabled/);
+  });
+
   it('hides the compatibility warning when editing a locked builtin endpoint', () => {
     const html = renderToStaticMarkup(
       <AddCustomProviderModal

apps/desktop/src/renderer/src/components/AddCustomProviderModal.tsx

@@ -19,6 +19,9 @@ interface Props {
     baseUrl?: string;
     wire?: WireApi;
     defaultModel?: string;
+    requiresApiKey?: boolean;
+    /** Preset-specific setup note shown above the form. */
+    hint?: string;
   };
   /**
    * Edit-mode: pre-fill every field from an existing provider and save via
@@ -143,7 +146,9 @@ export function AddCustomProviderModal({
   const [name, setName] = useState(editTarget?.name ?? initialValues?.name ?? '');
   const [baseUrl, setBaseUrl] = useState(editTarget?.baseUrl ?? initialValues?.baseUrl ?? '');
   const [apiKey, setApiKey] = useState('');
-  const [requiresApiKey, setRequiresApiKey] = useState(editTarget?.requiresApiKey !== false);
+  const [requiresApiKey, setRequiresApiKey] = useState(
+    (editTarget ?? initialValues)?.requiresApiKey !== false,
+  );
   const [defaultModel, setDefaultModel] = useState(
     editTarget?.defaultModel ?? initialValues?.defaultModel ?? '',
   );
@@ -436,6 +441,12 @@ export function AddCustomProviderModal({
           </button>
         </div>
 
+        {!isEdit && initialValues?.hint !== undefined && (
+          <p className="text-[var(--text-xs)] leading-5 text-[var(--color-text-secondary)]">
+            {initialValues.hint}
+          </p>
+        )}
+
         {!lockEndpoint && (
           <Field label={t('settings.providers.custom.wire')}>
             <div className="flex gap-3 flex-wrap">

apps/desktop/src/renderer/src/components/settings/ModelsTab.tsx

@@ -1,6 +1,7 @@
 import { useT } from '@open-codesign/i18n';
 import type { WireApi } from '@open-codesign/shared';
 import {
+  getPresetById,
   isSupportedOnboardingProvider,
   PROVIDER_SHORTLIST as SHORTLIST,
 } from '@open-codesign/shared';
@@ -170,6 +171,7 @@ interface AddProviderMenuProps {
   onAddOllama: () => void;
   onAddCustom: () => void;
   onAddCliProxyApi: () => void;
+  onAddLiteLLM: () => void;
 }
 
 function AddProviderMenu({
@@ -182,6 +184,7 @@ function AddProviderMenu({
   onAddOllama,
   onAddCustom,
   onAddCliProxyApi,
+  onAddLiteLLM,
 }: AddProviderMenuProps) {
   const t = useT();
   const rootRef = useRef<HTMLDivElement>(null);
@@ -254,6 +257,13 @@ function AddProviderMenu({
       disabled: false,
       onClick: onAddCliProxyApi,
     },
+    {
+      key: 'litellm',
+      label: t('settings.providers.litellm.presetName'),
+      desc: t('settings.providers.litellm.presetDescription'),
+      disabled: false,
+      onClick: onAddLiteLLM,
+    },
   ];
 
   return (
@@ -347,6 +357,8 @@ export function ModelsTab() {
         baseUrl: string;
         wire: WireApi;
         defaultModel?: string;
+        requiresApiKey?: boolean;
+        hint?: string;
       }
     | undefined
   >(undefined);
@@ -1036,6 +1048,18 @@ export function ModelsTab() {
               });
               setShowAddCustom(true);
             }}
+            onAddLiteLLM={() => {
+              setShowAddMenu(false);
+              setCustomProviderPreset({
+                name: 'LiteLLM Gateway',
+                baseUrl: getPresetById('litellm')?.baseUrl ?? '',
+                wire: 'openai-chat',
+                defaultModel: '',
+                requiresApiKey: false,
+                hint: t('settings.providers.litellm.hint'),
+              });
+              setShowAddCustom(true);
+            }}
           />
         </div>
 

packages/i18n/src/locales/en.json

@@ -690,6 +690,11 @@
         "connectionOk": "Connection OK",
         "connectionFailed": "Connection failed"
       },
+      "litellm": {
+        "presetName": "LiteLLM Gateway",
+        "presetDescription": "Self-hosted OpenAI-compatible gateway, API key optional",
+        "hint": "Point this at a LiteLLM proxy you already run (LiteLLM is not bundled). The default URL is http://localhost:4000/v1; edit it for remote deployments. Keyless mode is on for proxies without a master key; turn it off and paste your master or virtual key if the gateway requires auth. For a localhost URL, allow private-network testing below."
+      },
       "cliProxyApi": {
         "presetName": "CLIProxyAPI",
         "presetDescription": "Local proxy that wraps Claude/Codex/Gemini OAuth subscriptions",

packages/i18n/src/locales/es.json

@@ -514,6 +514,11 @@
         "connectionOk": "Conexión OK",
         "connectionFailed": "Conexión fallida"
       },
+      "litellm": {
+        "presetName": "LiteLLM Gateway",
+        "presetDescription": "Gateway autoalojado compatible con OpenAI, clave API opcional",
+        "hint": "Apunta a un proxy LiteLLM que ya tengas en ejecución (LiteLLM no viene incluido). La URL predeterminada es http://localhost:4000/v1; edítala para despliegues remotos. El modo sin clave está activado para proxies sin master key; desactívalo y pega tu master key o virtual key si el gateway requiere autenticación. Para una URL localhost, permite abajo las pruebas en red privada."
+      },
       "cliProxyApi": {
         "presetName": "CLIProxyAPI",
         "presetDescription": "Proxy local que envuelve suscripciones OAuth de Claude/Codex/Gemini",

packages/i18n/src/locales/pt-BR.json

@@ -491,6 +491,11 @@
       },
       "addKey": "Adicionar chave",
       "missingKey": "Chave ausente",
+      "litellm": {
+        "presetName": "LiteLLM Gateway",
+        "presetDescription": "Gateway auto-hospedado compatível com OpenAI, chave de API opcional",
+        "hint": "Aponte para um proxy LiteLLM que você já executa (o LiteLLM não vem incluído). A URL padrão é http://localhost:4000/v1; edite-a para implantações remotas. O modo sem chave vem ativado para proxies sem master key; desative-o e cole sua master key ou virtual key se o gateway exigir autenticação. Para uma URL localhost, permita abaixo os testes em rede privada."
+      },
       "cliProxyApi": {
         "presetName": "CLIProxyAPI",
         "presetDescription": "Proxy local que encapsula assinaturas OAuth do Claude/Codex/Gemini",

packages/i18n/src/locales/zh-CN.json

@@ -690,6 +690,11 @@
         "connectionOk": "连接正常",
         "connectionFailed": "连接失败"
       },
+      "litellm": {
+        "presetName": "LiteLLM Gateway",
+        "presetDescription": "自托管的 OpenAI 兼容网关，API Key 可选",
+        "hint": "请指向你已在运行的 LiteLLM 代理（应用不内置 LiteLLM）。默认地址为 http://localhost:4000/v1，远程部署请修改 URL。默认开启无 Key 模式，适用于未配置 master key 的代理；若网关需要鉴权，请关闭该选项并粘贴 master key 或 virtual key。使用 localhost 地址时，请在下方允许测试本地/内网地址。"
+      },
       "cliProxyApi": {
         "presetName": "CLIProxyAPI",
         "presetDescription": "本地反代，将 Claude/Codex/Gemini 的订阅账号包装成统一 API",

packages/shared/src/proxy-presets.test.ts

@@ -36,6 +36,7 @@ describe('PROXY_PRESETS', () => {
     expect(ids).toContain('requesty');
     expect(ids).toContain('siliconflow');
     expect(ids).toContain('one-api');
+    expect(ids).toContain('litellm');
     expect(ids).toContain('custom');
   });
 
@@ -50,6 +51,12 @@ describe('PROXY_PRESETS', () => {
     expect(preset?.baseUrl).toBe('https://api.openai.com/v1');
   });
 
+  it('LiteLLM preset targets the default local proxy over the OpenAI wire', () => {
+    const litellm = PROXY_PRESETS.find((p) => p.id === 'litellm');
+    expect(litellm?.provider).toBe('openai');
+    expect(litellm?.baseUrl).toBe('http://localhost:4000/v1');
+  });
+
   it('custom preset has empty baseUrl', () => {
     const custom = PROXY_PRESETS.find((p) => p.id === 'custom');
     expect(custom?.baseUrl).toBe('');

packages/shared/src/proxy-presets.ts

@@ -59,6 +59,13 @@ export const PROXY_PRESETS = [
     baseUrl: 'http://localhost:3000/v1',
     notes: 'Edit URL to your deployment',
   },
+  {
+    id: 'litellm',
+    label: 'LiteLLM Gateway',
+    provider: 'openai',
+    baseUrl: 'http://localhost:4000/v1',
+    notes: 'Self-hosted OpenAI-compatible gateway; key optional',
+  },
   {
     id: 'cli-proxy-api',
     label: 'CLIProxyAPI',