Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active
TypeScript aianthropicartifactsawsazurechatgptchatgpt-cloneclaudeclonedeepseekgeminigooglegpt-5librechatmcpo1openairesponses-apivisionwebui
import { Providers } from '@librechat/agents';
import { EModelEndpoint } from 'librechat-data-provider';
import type { AppConfig } from '@librechat/data-schemas';
import {
getProviderConfig,
providerConfigMap,
resolveTitleConfig,
resolveTitleTiming,
} from './providers';
const buildAppConfig = (
customEndpoints: Array<{ name: string; baseURL?: string; apiKey?: string; provider?: string }>,
): AppConfig =>
({
endpoints: {
[EModelEndpoint.custom]: customEndpoints,
},
}) as unknown as AppConfig;
describe('getProviderConfig', () => {
it('resolves the existing google (API key) path to initializeGoogle', () => {
// Regression guard: the API-key path uses `Providers.GOOGLE === 'google'`,
// which has always mapped via `EModelEndpoint.google`. Adding the
// `Providers.VERTEXAI` entry must not perturb this.
const result = getProviderConfig({
provider: Providers.GOOGLE,
appConfig: buildAppConfig([]),
});
expect(result.overrideProvider).toBe(Providers.GOOGLE);
expect(result.getOptions).toBe(providerConfigMap[EModelEndpoint.google]);
expect(result.customEndpointConfig).toBeUndefined();
});
it('vertexai resolves to the same initializer as google (issue #13006 follow-up)', () => {
const result = getProviderConfig({
provider: Providers.VERTEXAI,
appConfig: buildAppConfig([]),
});
expect(result.overrideProvider).toBe(Providers.VERTEXAI);
expect(result.getOptions).toBe(providerConfigMap[EModelEndpoint.google]);
expect(result.customEndpointConfig).toBeUndefined();
});
it('falls back case-insensitively when only a CamelCase match exists', () => {
// Agent runtime resolved provider to lowercase `"openrouter"`, but the
// user's `librechat.yaml` declared `name: "OpenRouter"`.
const appConfig = buildAppConfig([
{ name: 'OpenRouter', baseURL: 'https://openrouter.ai/api/v1', apiKey: 'sk-test' },
]);
const result = getProviderConfig({ provider: 'openrouter', appConfig });
expect(result.overrideProvider).toBe(Providers.OPENROUTER);
expect(result.customEndpointConfig?.name).toBe('OpenRouter');
});
it('prefers the exact-case match when both casings exist (preserves case-sensitive identity)', () => {
// Two distinct custom endpoints differing only in case is supported by
// `loadCustomEndpointsConfig` (the keys are case-preserving). Direct
// exact-case lookup should win — case-insensitive fallback must not
// shadow the user's intent.
const appConfig = buildAppConfig([
{ name: 'OpenRouter', baseURL: 'https://prod.example/v1', apiKey: 'prod' },
{ name: 'openrouter', baseURL: 'https://staging.example/v1', apiKey: 'staging' },
]);
const result = getProviderConfig({ provider: 'openrouter', appConfig });
expect(result.customEndpointConfig?.baseURL).toBe('https://staging.example/v1');
});
it('resolves an exact-case lowercase entry', () => {
const appConfig = buildAppConfig([
{ name: 'openrouter', baseURL: 'https://openrouter.ai/api/v1', apiKey: 'sk-test' },
]);
const result = getProviderConfig({ provider: 'openrouter', appConfig });
expect(result.customEndpointConfig?.name).toBe('openrouter');
});
it('throws on ambiguous case-insensitive matches when no exact-case entry exists (codex review)', () => {
// User has two distinct entries differing only in case, both
// non-lowercase. The agent runtime resolves provider to lowercase
// "openrouter" — neither matches case-sensitively, and silently
// picking array-first could route requests to the wrong baseURL/apiKey.
const appConfig = buildAppConfig([
{ name: 'OpenRouter', baseURL: 'https://prod.example/v1', apiKey: 'prod' },
{ name: 'OPENROUTER', baseURL: 'https://canary.example/v1', apiKey: 'canary' },
]);
expect(() => getProviderConfig({ provider: 'openrouter', appConfig })).toThrow(
/ambiguous.*OpenRouter.*OPENROUTER/i,
);
});
it('throws when openrouter has no matching custom endpoint at all', () => {
expect(() =>
getProviderConfig({ provider: 'openrouter', appConfig: buildAppConfig([]) }),
).toThrow('Provider openrouter not supported');
});
it('routes a custom endpoint with provider:anthropic to the Anthropic client', () => {
const appConfig = buildAppConfig([
{
name: 'Claude-Compatible',
baseURL: 'https://gateway.example.com',
apiKey: 'sk-ant',
provider: EModelEndpoint.anthropic,
},
]);
const result = getProviderConfig({ provider: 'Claude-Compatible', appConfig });
expect(result.overrideProvider).toBe(Providers.ANTHROPIC);
expect(result.customEndpointConfig?.provider).toBe(EModelEndpoint.anthropic);
});
it('defaults a custom endpoint without provider to the OpenAI-compatible client', () => {
const appConfig = buildAppConfig([
{ name: 'My-LLM', baseURL: 'https://api.example.com/v1', apiKey: 'sk-test' },
]);
const result = getProviderConfig({ provider: 'My-LLM', appConfig });
expect(result.overrideProvider).toBe(Providers.OPENAI);
expect(result.customEndpointConfig?.name).toBe('My-LLM');
});
it('applies provider:anthropic even when the endpoint name collides with a known custom provider', () => {
// `openrouter` resolves via `providerConfigMap` first (skipping the generic
// custom branch); the override must still be re-applied from the config so
// overrideProvider-derived values (token/context budget) use the Anthropic map.
const appConfig = buildAppConfig([
{
name: 'openrouter',
baseURL: 'https://gateway.example.com',
apiKey: 'sk-ant',
provider: EModelEndpoint.anthropic,
},
]);
const result = getProviderConfig({ provider: 'openrouter', appConfig });
expect(result.overrideProvider).toBe(Providers.ANTHROPIC);
});
});
describe('resolveTitleTiming', () => {
const withEndpoints = (endpoints: Record<string, unknown>): AppConfig =>
({ endpoints }) as unknown as AppConfig;
it("defaults to 'immediate' when no config is provided", () => {
expect(resolveTitleTiming({})).toBe('immediate');
});
it("defaults to 'immediate' when endpoints is missing", () => {
expect(resolveTitleTiming({ appConfig: {} as AppConfig })).toBe('immediate');
});
it("defaults to 'immediate' when no titleTiming is set", () => {
const appConfig = withEndpoints({ [EModelEndpoint.agents]: { titleConvo: true } });
expect(resolveTitleTiming({ appConfig, endpoint: EModelEndpoint.agents })).toBe('immediate');
});
it("returns 'final' from the global `all` config", () => {
const appConfig = withEndpoints({ all: { titleTiming: 'final' } });
expect(resolveTitleTiming({ appConfig, endpoint: EModelEndpoint.agents })).toBe('final');
});
it("returns 'final' from the per-endpoint config when `all` is unset", () => {
const appConfig = withEndpoints({ [EModelEndpoint.agents]: { titleTiming: 'final' } });
expect(resolveTitleTiming({ appConfig, endpoint: EModelEndpoint.agents })).toBe('final');
});
it('lets `all` take precedence over the per-endpoint value', () => {
const appConfig = withEndpoints({
all: { titleTiming: 'immediate' },
[EModelEndpoint.agents]: { titleTiming: 'final' },
});
expect(resolveTitleTiming({ appConfig, endpoint: EModelEndpoint.agents })).toBe('immediate');
});
it('does not let unrelated `all` config block a per-endpoint value', () => {
const appConfig = withEndpoints({
all: { titleConvo: true },
[EModelEndpoint.agents]: { titleTiming: 'final' },
});
expect(resolveTitleTiming({ appConfig, endpoint: EModelEndpoint.agents })).toBe('final');
});
it('checks endpoint candidates in order before provider fallback', () => {
const appConfig = withEndpoints({
[EModelEndpoint.agents]: { titleTiming: 'final' },
[EModelEndpoint.openAI]: { titleTiming: 'immediate' },
});
expect(
resolveTitleTiming({
appConfig,
endpoint: [EModelEndpoint.agents, EModelEndpoint.openAI],
}),
).toBe('final');
});
it('falls back to backing provider timing when agents has no titleTiming', () => {
const appConfig = withEndpoints({
[EModelEndpoint.agents]: { titleConvo: true },
[EModelEndpoint.openAI]: { titleTiming: 'final' },
});
expect(
resolveTitleTiming({
appConfig,
endpoint: [EModelEndpoint.agents, EModelEndpoint.openAI],
}),
).toBe('final');
});
it("returns 'immediate' for an endpoint with no override and no `all`", () => {
const appConfig = withEndpoints({ [EModelEndpoint.openAI]: { titleTiming: 'final' } });
expect(resolveTitleTiming({ appConfig, endpoint: EModelEndpoint.agents })).toBe('immediate');
});
it("resolves 'final' from a custom endpoint config (endpoints.custom[])", () => {
const appConfig = withEndpoints({
[EModelEndpoint.custom]: [{ name: 'MyProvider', titleTiming: 'final' }],
});
expect(resolveTitleTiming({ appConfig, endpoint: 'MyProvider' })).toBe('final');
});
it('resolves a normalized custom provider name (openrouter -> OpenRouter)', () => {
const appConfig = withEndpoints({
[EModelEndpoint.custom]: [
{ name: 'OpenRouter', baseURL: 'https://openrouter.ai/api/v1', titleTiming: 'final' },
],
});
expect(resolveTitleTiming({ appConfig, endpoint: 'openrouter' })).toBe('final');
});
});
describe('resolveTitleConfig', () => {
const withEndpoints = (endpoints: Record<string, unknown>): AppConfig =>
({ endpoints }) as unknown as AppConfig;
const openAITitles = {
titleConvo: true,
titleModel: 'gpt-4o-mini',
titleMethod: 'structured',
titlePrompt: 'OpenAI prompt',
titlePromptTemplate: 'OpenAI template: {convo}',
} as const;
it('returns no settings when nothing is configured', () => {
expect(resolveTitleConfig({ endpoint: EModelEndpoint.openAI })).toEqual({});
expect(
resolveTitleConfig({ appConfig: {} as AppConfig, endpoint: EModelEndpoint.openAI }),
).toEqual({});
});
it('keeps the endpoint title settings when `all` sets only other keys', () => {
const appConfig = withEndpoints({
all: { streamRate: 1, headers: { 'x-gateway': 'a' }, activityLabel: true },
[EModelEndpoint.openAI]: openAITitles,
});
expect(resolveTitleConfig({ appConfig, endpoint: EModelEndpoint.openAI })).toEqual(
openAITitles,
);
});
it('lets each `all` field win on its own, keeping the endpoint value for the rest', () => {
const appConfig = withEndpoints({
all: { titlePrompt: 'Global prompt', titleMethod: 'completion' },
[EModelEndpoint.openAI]: openAITitles,
});
expect(resolveTitleConfig({ appConfig, endpoint: EModelEndpoint.openAI })).toEqual({
...openAITitles,
titlePrompt: 'Global prompt',
titleMethod: 'completion',
});
});
it('takes titleEndpoint and titleModel together from the first layer that sets either', () => {
const allSetsEndpoint = withEndpoints({
all: { titleEndpoint: EModelEndpoint.anthropic },
[EModelEndpoint.openAI]: openAITitles,
});
const allSetsModel = withEndpoints({
all: { titleModel: 'claude-haiku' },
[EModelEndpoint.openAI]: { titleEndpoint: EModelEndpoint.anthropic, titleModel: 'gpt-4o' },
});
expect(
resolveTitleConfig({ appConfig: allSetsEndpoint, endpoint: EModelEndpoint.openAI }),
).toMatchObject({ titleEndpoint: EModelEndpoint.anthropic, titleModel: undefined });
expect(
resolveTitleConfig({ appConfig: allSetsModel, endpoint: EModelEndpoint.openAI }),
).toMatchObject({ titleEndpoint: undefined, titleModel: 'claude-haiku' });
});
it('treats an explicit current_model in `all` as setting the model choice', () => {
const appConfig = withEndpoints({
all: { titleModel: 'current_model' },
[EModelEndpoint.openAI]: { titleEndpoint: EModelEndpoint.anthropic, titleModel: 'x' },
});
expect(resolveTitleConfig({ appConfig, endpoint: EModelEndpoint.openAI })).toMatchObject({
titleEndpoint: undefined,
titleModel: 'current_model',
});
});
it('disables titles everywhere through `all.titleConvo: false`', () => {
const appConfig = withEndpoints({
all: { titleConvo: false },
[EModelEndpoint.openAI]: openAITitles,
});
expect(resolveTitleConfig({ appConfig, endpoint: EModelEndpoint.openAI }).titleConvo).toBe(
false,
);
});
it('honors an endpoint `titleConvo: false` next to an unrelated `all` block', () => {
const appConfig = withEndpoints({
all: { streamRate: 1 },
[EModelEndpoint.openAI]: { titleConvo: false },
});
expect(resolveTitleConfig({ appConfig, endpoint: EModelEndpoint.openAI }).titleConvo).toBe(
false,
);
});
it('falls back to the custom endpoint config, still behind `all`', () => {
const appConfig = withEndpoints({
all: { titlePrompt: 'Global prompt' },
[EModelEndpoint.custom]: [
{
name: 'Gateway',
baseURL: 'https://gateway.example/v1',
apiKey: 'gateway-key',
titleModel: 'gateway-small',
titleMethod: 'completion',
titlePrompt: 'Gateway prompt',
},
],
});
const { customEndpointConfig } = getProviderConfig({ provider: 'Gateway', appConfig });
expect(resolveTitleConfig({ appConfig, endpoint: 'Gateway', customEndpointConfig })).toEqual({
titleModel: 'gateway-small',
titleMethod: 'completion',
titlePrompt: 'Global prompt',
});
});
it('prefers the named endpoint block over the custom config field by field', () => {
const appConfig = withEndpoints({
[EModelEndpoint.openAI]: { titlePrompt: 'Named prompt' },
});
expect(
resolveTitleConfig({
appConfig,
endpoint: EModelEndpoint.openAI,
customEndpointConfig: { titlePrompt: 'Custom prompt', titleMethod: 'functions' },
}),
).toEqual({ titlePrompt: 'Named prompt', titleMethod: 'functions' });
});
});