revert: remove dedicated LiteLLM provider

apps/desktop/src/renderer/settings/provider-brand-marks.tsx

@@ -156,7 +156,6 @@ export function ProviderBrandMark({ type }: { type: ProviderType }): ReactElemen
       return <Claude />;
     case 'openai':
     case 'codex-subscription':
-    case 'litellm':
     case 'openai-compatible':
       return <OpenAI />;
     case 'google':

apps/desktop/src/renderer/settings/provider-display.tsx

@@ -41,8 +41,6 @@ export function providerDisplay(type: ProviderType): { name: string; description
       return { name: 'MiniMax', description: 'MiniMax · Anthropic 兼容', badge: 'API' };
     case 'MiniMax-cn':
       return { name: 'MiniMax 中国站', description: 'MiniMax 中国站 · Anthropic 兼容', badge: 'API' };
-    case 'litellm':
-      return { name: 'LiteLLM', description: 'AI gateway proxy for 100+ LLM providers.', badge: 'Gateway' };
     case 'ollama':
       return { name: 'Ollama', description: '本机运行 · 离线可用', badge: 'Local' };
     case 'openai-compatible':

packages/core/src/__tests__/llm-connections.test.ts

@@ -25,7 +25,7 @@ import {
 } from '../llm-connections.js';
 
 describe('provider compatibility contract', () => {
-  it('keeps persisted provider ids and existing provider ordering stable', () => {
+  it('exposes only supported first-class provider ids in stable order', () => {
     assert.deepEqual(Object.keys(PROVIDER_DEFAULTS), [
       'anthropic',
       'kimi-coding-plan',
@@ -37,7 +37,6 @@ describe('provider compatibility contract', () => {
       'MiniMax',
       'MiniMax-cn',
       'siliconflow',
-      'litellm',
       'ollama',
       'openai-compatible',
       'claude-subscription',
@@ -70,7 +69,6 @@ describe('provider compatibility contract', () => {
       'openai',
       'google',
       'ollama',
-      'litellm',
       'openai-compatible',
     ]);
   });
@@ -283,19 +281,6 @@ describe('provider URL defaults', () => {
     assert.equal(PROVIDER_DEFAULTS['codex-subscription'].description, 'ChatGPT/Codex account OAuth path for OpenAI Responses models.');
   });
 
-  it('defines LiteLLM as an OpenAI-protocol custom gateway with localhost default', () => {
-    const litellm = PROVIDER_DEFAULTS['litellm'];
-    assert.equal(litellm.label, 'LiteLLM');
-    assert.equal(litellm.protocol, 'openai');
-    assert.equal(litellm.category, 'custom');
-    assert.equal(litellm.baseUrl, 'http://localhost:4000/v1');
-    assert.equal(litellm.authKind, 'api_key');
-    assert.equal(litellm.backendKind, 'ai-sdk');
-    assert.equal(litellm.status, 'ready');
-    assert.equal(litellm.catalogBadge, 'Gateway');
-    assert.deepEqual(litellm.fallbackModels, []);
-  });
-
   it('keeps Kimi Coding Plan separate from Moonshot API key access', () => {
     assert.equal(PROVIDER_DEFAULTS['kimi-coding-plan'].baseUrl, 'https://api.kimi.com/coding/v1');
     assert.equal(PROVIDER_DEFAULTS['kimi-coding-plan'].signupUrl, 'https://www.kimi.com/code/console');
@@ -347,19 +332,6 @@ describe('persistedBaseUrl', () => {
     assert.equal(persistedBaseUrl('google', 'https://my-gemini-proxy.example.com/v1beta'), 'https://my-gemini-proxy.example.com/v1beta');
   });
 
-  it('collapses litellm default localhost URL to undefined (no override to persist)', () => {
-    assert.equal(
-      persistedBaseUrl('litellm', 'http://localhost:4000/v1'),
-      undefined,
-      'litellm default must not be persisted as an override',
-    );
-  });
-
-  it('persists a custom litellm proxy URL as a real override', () => {
-    const custom = 'https://litellm.company.internal/v1';
-    assert.equal(persistedBaseUrl('litellm', custom), custom);
-  });
-
   it('persists a custom override for openai-compatible (whose default is the empty string)', () => {
     // openai-compatible is the one provider with no canonical default — any
     // non-empty value the user supplies is a real override and must persist.

packages/core/src/provider-registry.ts

@@ -260,23 +260,6 @@ const providerRegistry = {
     catalogOrder: 7,
     recommendedOrder: 1,
   },
-  litellm: {
-    label: 'LiteLLM',
-    description: 'AI gateway proxy for 100+ LLM providers.',
-    baseUrl: 'http://localhost:4000/v1',
-    authKind: 'api_key',
-    backendKind: 'ai-sdk',
-    fallbackModels: [],
-    status: 'ready',
-    protocol: 'openai',
-    runtimeAdapter: { kind: 'openai-compatible', name: 'provider' },
-    modelDiscovery: { kind: 'protocol' },
-    category: 'custom',
-    catalogGroup: 'aggregators',
-    catalogBadge: 'Gateway',
-    signupUrl: 'https://docs.litellm.ai/',
-    catalogOrder: 12,
-  },
   ollama: {
     label: 'Ollama',
     description: 'Local models from Ollama on localhost.',

packages/headless/src/provider-env.ts

@@ -17,7 +17,6 @@ const PROVIDER_CREDENTIAL_ENV = {
   MiniMax: env('MINIMAX', ['MINIMAX_BASE_URL']),
   'MiniMax-cn': env('MINIMAX', ['MINIMAX_BASE_URL']),
   siliconflow: env('SILICONFLOW', ['SILICONFLOW_BASE_URL']),
-  litellm: env('OPENAI', ['OPENAI_BASE_URL']),
   'openai-compatible': env('OPENAI', ['OPENAI_BASE_URL']),
   'claude-subscription': env('ANTHROPIC'),
 } satisfies Partial<Record<ProviderType, ProviderCredentialEnv>>;

packages/runtime/src/__tests__/model-factory-thinking.test.ts

@@ -77,39 +77,6 @@ describe('buildProviderOptions: thinking level', () => {
   });
 });
 
-describe('getAIModel: litellm provider', () => {
-  test('creates an OpenAI-compatible client with provider name "litellm" and correct modelId', () => {
-    const model = getAIModel({
-      connection: conn('litellm', 'litellm-gateway'),
-      apiKey: 'sk-test-key',
-      modelId: 'gpt-4o',
-    });
-    assert.equal(model.modelId, 'gpt-4o');
-    assert.equal(model.provider, 'litellm.chat');
-  });
-
-  test('uses the connection baseUrl override when provided', () => {
-    const model = getAIModel({
-      connection: { ...conn('litellm', 'litellm-custom'), baseUrl: 'https://litellm.company.internal/v1' },
-      apiKey: 'sk-test-key',
-      modelId: 'claude-sonnet-4-5-20250929',
-    });
-    assert.equal(model.modelId, 'claude-sonnet-4-5-20250929');
-    assert.equal(model.provider, 'litellm.chat');
-  });
-
-  test('falls back to localhost:4000 default when no baseUrl override is set', () => {
-    const model = getAIModel({
-      connection: conn('litellm', 'litellm-default'),
-      apiKey: 'sk-test-key',
-      modelId: 'gpt-4o-mini',
-    });
-    // The model is created successfully with default base URL
-    assert.equal(model.provider, 'litellm.chat');
-    assert.equal(model.modelId, 'gpt-4o-mini');
-  });
-});
-
 describe('getAIModel: models.dev registry providers', () => {
   test('routes SiliconFlow through the shared OpenAI-compatible adapter without rewriting model ids', () => {
     const model = getAIModel({
@@ -123,15 +90,6 @@ describe('getAIModel: models.dev registry providers', () => {
   });
 });
 
-describe('buildProviderOptions: litellm falls through to default (empty options)', () => {
-  test('litellm returns empty options regardless of thinking level', () => {
-    assert.deepEqual(buildProviderOptions(conn('litellm'), 'gpt-4o'), {});
-    assert.deepEqual(buildProviderOptions(conn('litellm'), 'gpt-4o', 'high'), {});
-    assert.deepEqual(buildProviderOptions(conn('litellm'), 'gpt-4o', 'off'), {});
-    assert.deepEqual(buildProviderOptions(conn('litellm'), 'claude-sonnet-4-5-20250929'), {});
-  });
-});
-
 describe('buildProviderOptions: openai-compatible namespace', () => {
   test('zai-coding-plan emits reasoningEffort under the raw dashed namespace', () => {
     assert.deepEqual(buildProviderOptions(conn('zai-coding-plan', 'zai-coding-plan'), 'glm-5.2', 'high'), { 'zai-coding-plan': { reasoningEffort: 'high' } });

packages/ui/src/chat-model-helpers.ts

@@ -53,7 +53,6 @@ const PROVIDER_SHORT_LABEL: Partial<Record<ProviderType, string>> = {
   google: 'Google',
   deepseek: 'DeepSeek',
   moonshot: 'Moonshot',
-  litellm: 'LiteLLM',
   ollama: 'Ollama',
   'kimi-coding-plan': 'Kimi',
   'zai-coding-plan': 'Z.AI',